init

2026-07-30 08:45:38 +08:00
commit 25b24ac4cd
1333 changed files with 170092 additions and 0 deletions
@@ -0,0 +1,27 @@
+# Domain Plugins
+
+每个子目录是一个完整的领域交付单元。领域插件只能依赖 `cognitive_agent.sdk` 和 `contracts/`，不得依赖其他领域或中央编排实现。
+
+新领域从 `_template/` 复制，目录名和 `spec.domain` 必须使用相同的 ASCII snake_case 标识。插件注册由 `manifest.yaml` 驱动，不在中央代码中硬编码领域列表。
+
+领域目录固定结构：
+
+```text
+<domain>/
+├── README.md
+├── OWNERS.yaml
+├── manifest.yaml
+├── config/
+│   ├── capabilities.yaml
+│   └── presets/
+├── schemas/
+│   └── function-parameters.schema.json
+├── can/
+│   └── signal-mapping.yaml
+├── src/<domain>_domain/
+└── tests/
+    ├── unit/
+    ├── contract/
+    └── integration/
+```
+
@@ -0,0 +1,7 @@
+domain: example_domain
+primary_owner: TBD
+reviewers:
+  - integration_owner
+paths:
+  - domains/example_domain/**
+
@@ -0,0 +1,12 @@
+# Domain Template
+
+复制本目录后完成以下替换：
+
+1. 把 `example_domain` 替换为稳定领域标识。
+2. 把 `src/domain_plugin` 重命名为 `<domain>_domain`，同步修改 manifest entrypoint。
+3. 填写真实负责人、能力、参数 Schema、preset 和信号映射。
+4. `signal-mapping.yaml` 在通过 HIL 前保持 `status: draft`。
+5. 补齐 Generator、Validator、Compiler、Telemetry Decoder 和测试。
+
+领域插件不得保存 DBC 副本或直接发送 CAN。Compiler 返回符合公共 Schema 的 `CommandPlan`，公共 Runtime 负责安全门禁和发送。
+
@@ -0,0 +1,18 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: example-domain-reference-mapping
+  name: Example domain reference vehicle mapping
+  version: 0.1.0
+spec:
+  domain: example_domain
+  platform: reference_vehicle
+  dbc_ref: ../../../platforms/reference_vehicle/can/dbc/vehicle.dbc
+  status: draft
+  mappings:
+    - capability: example_domain.set_level
+      parameter: level_pct
+      compiler_handler: compile_set_level
+      calibration_ref: ../../../platforms/reference_vehicle/calibration/example_domain.yaml
+      unit: percent
+
@@ -0,0 +1,23 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: CapabilityCatalog
+metadata:
+  id: example-domain-capabilities
+  name: Example domain capabilities
+  version: 0.1.0
+spec:
+  domain: example_domain
+  capabilities:
+    - name: example_domain.set_level
+      description: Replace with a user-facing semantic capability description.
+      target_devices:
+        - example_device
+      target_scopes:
+        - vehicle
+      parameters:
+        - name: level_pct
+          type: number
+          unit: percent
+          required: true
+          description: Replace with the physical meaning and range source.
+      risk_level: medium
+      confirmation_policy: on_ambiguity
@@ -0,0 +1,24 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: example-domain-default
+  name: Example preset
+  version: 0.1.0
+spec:
+  domain: example_domain
+  operations:
+    - id: set_example_level
+      capability: example_domain.set_level
+      target:
+        device: example_device
+      parameters:
+        level_pct: 50
+      transition:
+        duration_ms: 1000
+        profile: device_default
+      timeout_ms: 5000
+      failure_policy: abort
+  locked_parameters: []
+  applicability:
+    platform: reference_vehicle
+
@@ -0,0 +1,16 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: DomainManifest
+metadata:
+  id: example-domain-plugin
+  name: Example domain plugin
+  version: 0.1.0
+spec:
+  domain: example_domain
+  entrypoint: domain_plugin.plugin:create_plugin
+  function_schema: schemas/function-parameters.schema.json
+  capabilities: config/capabilities.yaml
+  signal_mapping: can/signal-mapping.yaml
+  supported_platforms:
+    - reference_vehicle
+  telemetry_required: true
+
@@ -0,0 +1,32 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://cognitive-agent.local/domains/example_domain/function-parameters.schema.json",
+  "title": "ExampleDomainFunctionParameters",
+  "type": "object",
+  "properties": {
+    "spec": {
+      "type": "object",
+      "properties": {
+        "domain": {"const": "example_domain"},
+        "operations": {
+          "type": "array",
+          "items": {
+            "type": "object",
+            "properties": {
+              "capability": {"const": "example_domain.set_level"},
+              "parameters": {
+                "type": "object",
+                "additionalProperties": false,
+                "required": ["level_pct"],
+                "properties": {
+                  "level_pct": {"type": "number", "minimum": 0, "maximum": 100}
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
@@ -0,0 +1,12 @@
+# Plugin Source
+
+建议实现以下模块，具体接口以 `cognitive_agent.sdk` 为准：
+
+- `plugin.py`：创建并注册插件能力。
+- `generator.py`：`IntentIR -> FunctionConfig`。
+- `validator.py`：领域组合约束和上下文校验。
+- `compiler.py`：`FunctionConfig + PlatformContext -> CommandPlan`。
+- `telemetry.py`：CAN 回读到领域状态。
+
+模块只返回数据或领域错误，不直接访问 LLM、数据库、HTTP 或 CAN socket。
+
@@ -0,0 +1,8 @@
+# Domain Tests
+
+- `unit/`：Generator、Validator、Compiler、Telemetry Decoder 的纯单元测试。
+- `contract/`：manifest、能力目录、preset、Schema、映射和 SDK 插件契约。
+- `integration/`：通过公共模拟 CAN Gateway 验证发送计划、回读、超时和失败行为。
+
+硬件用例必须显式标记 `pytest.mark.hardware`，默认 CI 不运行。
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1,8 @@
+domain: ambient_light
+primary_owner: ambient_light_owner
+reviewers:
+  - integration_owner
+  - can_owner
+paths:
+  - domains/ambient_light/**
+
@@ -0,0 +1,7 @@
+# Ambient Light Domain
+
+氛围灯领域已根据根目录 `空调-天幕-氛围灯CAN信号矩阵表.xlsx` 的“氛围灯”工作表完成 dry-run 闭环。
+
+当前能力包括总开关、精确 `#RRGGBB` 颜色、0–100% 亮度，以及露营、小憩、观影、冥想和浪漫等矩阵原生模式。颜色和亮度使用 `0x32B/0x2B` 中的 IVI 1680 万色 RGB 信号，无需虚构颜色索引；总开关使用 `0x064/0x02`，场景模式使用 `0x496/0x04`。
+
+动态效果不会由 Agent 逐帧生成。源表缺少完整时序、计数器、校验和和车型适用性，当前严格禁止实车发送。
@@ -0,0 +1,50 @@
+version: 1.0.0
+source:
+  file: 空调-天幕-氛围灯CAN信号矩阵表.xlsx
+  sheet: 氛围灯
+scope: normalized_ambient_light_subset
+hardware_execution_allowed: false
+limitations:
+  - Source workbook is partial and omits bus timing, counter, checksum, and variant applicability.
+  - Exact RGB control uses the matrix-defined IVI 16.8-million-color request.
+messages:
+  - name: Ambient_Power_Control_064_02
+    frame_id: 100
+    minimum_payload_bytes: 8
+    fixed_signals: {Sub_ID: 2}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Atomic_atmosphere_light_switch_S, start_bit: 56, bit_length: 2}
+  - name: Ambient_Mode_Control_496_04
+    frame_id: 1174
+    minimum_payload_bytes: 4
+    fixed_signals: {Sub_ID: 4}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: atmosphere_Light_Request_Command_496_S, start_bit: 27, bit_length: 5}
+  - name: Ambient_RGB_Control_32B_2B
+    frame_id: 811
+    minimum_payload_bytes: 5
+    fixed_signals: {Sub_ID: 43}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: IVI_1680Atmosphere_Color_R_Setting_S, start_bit: 8, bit_length: 8}
+      - {name: IVI_1680Atmosphere_Color_G_Setting_S, start_bit: 16, bit_length: 8}
+      - {name: IVI_1680Atmosphere_Color_B_Setting_S, start_bit: 24, bit_length: 8}
+      - {name: IVI_1680Atmosphere_Brightness_Setting_S, start_bit: 32, bit_length: 7}
+  - name: Ambient_Power_Status_3F3
+    frame_id: 1011
+    minimum_payload_bytes: 8
+    signals:
+      - {name: Atmosphere_Light_Switch_Fdback_S, start_bit: 62, bit_length: 2}
+  - name: Ambient_RGB_Status_1F1
+    frame_id: 497
+    minimum_payload_bytes: 48
+    signals:
+      - {name: LFDoor_Flowing_Ambient_Light_Brightness_S, start_bit: 40, bit_length: 7}
+      - {name: Front_Atmspr_Light_Color_R_S, start_bit: 336, bit_length: 8}
+      - {name: Front_Atmspr_Light_Color_G_S, start_bit: 344, bit_length: 8}
+      - {name: Front_Atmspr_Light_Color_B_S, start_bit: 352, bit_length: 8}
+      - {name: Rear_Atmspr_Light_Color_R_S, start_bit: 360, bit_length: 8}
+      - {name: Rear_Atmspr_Light_Color_G_S, start_bit: 368, bit_length: 8}
+      - {name: Rear_Atmspr_Light_Color_B_S, start_bit: 376, bit_length: 8}
@@ -0,0 +1,55 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: ambient-light-reference-vehicle-mapping
+  name: Ambient-light mapping from supplied CAN matrix
+  version: 1.0.0
+  description: Virtual-only mapping using matrix-defined power, mode, RGB and brightness signals.
+spec:
+  domain: ambient_light
+  platform: reference_vehicle
+  dbc_ref: ../../../platforms/reference_vehicle/can/dbc/vehicle.dbc
+  matrix_ref: can-matrix.yaml
+  status: draft
+  mappings:
+    - capability: ambient_light.set_power
+      parameter: state
+      target_zone: vehicle
+      compiler_handler: compile_enum
+      value_map: {"on": 1, "off": 2}
+      bus: cabin_can
+      message: Ambient_Power_Control_064_02
+      signal: Atomic_atmosphere_light_switch_S
+      feedback_message: Ambient_Power_Status_3F3
+      feedback_signal: Atmosphere_Light_Switch_Fdback_S
+      feedback_handler: compile_enum
+      feedback_value_map: {"on": 2, "off": 1}
+    - capability: ambient_light.set_illumination
+      parameter: color_hex
+      target_zone: vehicle
+      compiler_handler: compile_hex_rgb
+      bus: cabin_can
+      message: Ambient_RGB_Control_32B_2B
+      signals:
+        - IVI_1680Atmosphere_Color_R_Setting_S
+        - IVI_1680Atmosphere_Color_G_Setting_S
+        - IVI_1680Atmosphere_Color_B_Setting_S
+    - capability: ambient_light.set_illumination
+      parameter: brightness_pct
+      target_zone: vehicle
+      compiler_handler: compile_integer
+      bus: cabin_can
+      message: Ambient_RGB_Control_32B_2B
+      signal: IVI_1680Atmosphere_Brightness_Setting_S
+      feedback_message: Ambient_RGB_Status_1F1
+      feedback_signal: LFDoor_Flowing_Ambient_Light_Brightness_S
+      feedback_handler: compile_integer
+      unit: percent
+    - capability: ambient_light.set_mode
+      parameter: mode
+      target_zone: vehicle
+      compiler_handler: compile_enum
+      value_map: {camping: 1, rest: 2, cinema: 3, meditation: 4, romantic: 5, "off": 6, "on": 7}
+      bus: cabin_can
+      message: Ambient_Mode_Control_496_04
+      signal: atmosphere_Light_Request_Command_496_S
@@ -0,0 +1,51 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: CapabilityCatalog
+metadata:
+  id: ambient-light-capabilities
+  name: Ambient light capabilities
+  version: 1.0.0
+spec:
+  domain: ambient_light
+  capabilities:
+    - name: ambient_light.set_power
+      description: Turn the vehicle ambient-light system on or off.
+      target_devices: [ambient_light]
+      target_scopes: [vehicle]
+      parameters:
+        - name: state
+          type: string
+          required: true
+          allowed_values: ["on", "off"]
+          description: Absolute ambient-light power state.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: ambient_light.set_illumination
+      description: Set exact 24-bit RGB color and brightness for the cabin ambient light.
+      target_devices: [ambient_light]
+      target_scopes: [vehicle]
+      parameters:
+        - name: color_hex
+          type: string
+          required: true
+          description: RGB color encoded as #RRGGBB; infer a suitable color for semantic requests.
+        - name: brightness_pct
+          type: integer
+          unit: percent
+          minimum: 0
+          maximum: 100
+          required: true
+          description: Absolute brightness percentage.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: ambient_light.set_mode
+      description: Select a matrix-defined ambient-light scene mode.
+      target_devices: [ambient_light]
+      target_scopes: [vehicle]
+      parameters:
+        - name: mode
+          type: string
+          required: true
+          allowed_values: [camping, rest, cinema, meditation, romantic, "off", "on"]
+          description: Absolute scene mode.
+      risk_level: low
+      confirmation_policy: on_ambiguity
@@ -0,0 +1,26 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: ambient-light-relax
+  name: Ambient light relaxation preset
+  version: 1.0.0
+  description: Virtual visual preset using the matrix-defined 24-bit RGB request.
+spec:
+  domain: ambient_light
+  operations:
+    - id: set_relax_lighting
+      capability: ambient_light.set_illumination
+      target:
+        device: ambient_light
+        zone: vehicle
+      parameters:
+        color_hex: "#2E6F9E"
+        brightness_pct: 30
+      transition:
+        duration_ms: 1200
+        profile: ease_in_out
+      timeout_ms: 5000
+      failure_policy: continue
+  locked_parameters: []
+  applicability:
+    platform: reference_vehicle
@@ -0,0 +1,16 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: DomainManifest
+metadata:
+  id: ambient-light-domain-plugin
+  name: Ambient light domain plugin
+  version: 0.1.0
+spec:
+  domain: ambient_light
+  entrypoint: ambient_light_domain.plugin:create_plugin
+  function_schema: schemas/function-parameters.schema.json
+  capabilities: config/capabilities.yaml
+  signal_mapping: can/signal-mapping.yaml
+  supported_platforms:
+    - reference_vehicle
+  telemetry_required: true
+
@@ -0,0 +1,59 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://cognitive-agent.local/domains/ambient_light/function-parameters.schema.json",
+  "title": "AmbientLightFunctionParameters",
+  "type": "object",
+  "properties": {
+    "spec": {
+      "type": "object",
+      "properties": {
+        "domain": {"const": "ambient_light"},
+        "operations": {
+          "type": "array",
+          "items": {
+            "oneOf": [
+              {
+                "properties": {
+                  "capability": {"const": "ambient_light.set_power"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["state"],
+                    "properties": {"state": {"enum": ["on", "off"]}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "ambient_light.set_illumination"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["color_hex", "brightness_pct"],
+                    "properties": {
+                      "color_hex": {"type": "string", "pattern": "^#[0-9A-Fa-f]{6}$"},
+                      "brightness_pct": {"type": "integer", "minimum": 0, "maximum": 100}
+                    }
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "ambient_light.set_mode"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["mode"],
+                    "properties": {
+                      "mode": {"enum": ["camping", "rest", "cinema", "meditation", "romantic", "off", "on"]}
+                    }
+                  }
+                }
+              }
+            ]
+          }
+        }
+      }
+    }
+  }
+}
@@ -0,0 +1,4 @@
+# Ambient Light Plugin Source
+
+实现 `plugin.py`、`generator.py`、`validator.py`、`compiler.py` 和 `telemetry.py`。颜色、亮度和效果必须转换为 ECU 支持的稳定命令；不在插件内实现无限制逐帧 CAN 动画。
+
@@ -0,0 +1,4 @@
+# Ambient Light Tests
+
+覆盖颜色格式、亮度边界、灯区支持、效果降级、亮度曲线、回读和超时。动态效果测试需验证命令速率上限。
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1,8 @@
+domain: climate
+primary_owner: climate_owner
+reviewers:
+  - integration_owner
+  - can_owner
+paths:
+  - domains/climate/**
+
@@ -0,0 +1,7 @@
+# Climate Domain
+
+空调领域已根据根目录 `空调-天幕-氛围灯CAN信号矩阵表.xlsx` 的“空调”工作表完成 dry-run 闭环。
+
+当前能力包括分区电源、18–32℃ 温度、整车 0–7 档风量、A/C、内外循环、分区吹面/吹脚和前除霜。`can/can-matrix.yaml` 保留来源信号名、帧 ID、Sub ID、Start Bit 和 Bit Length；`can/signal-mapping.yaml` 负责把稳定语义参数确定性转换为矩阵原始值。
+
+命令主要使用 `0x064` 的 `0x03`–`0x06` 子帧，虚拟反馈使用 `0x0DB` 和 `0x404`。源表未提供完整周期、计数器、校验和、总线通道和车型适用性，`hardware_execution_allowed` 固定为 `false`。
@@ -0,0 +1,77 @@
+version: 1.0.0
+source:
+  file: 空调-天幕-氛围灯CAN信号矩阵表.xlsx
+  sheet: 空调
+scope: normalized_climate_subset
+hardware_execution_allowed: false
+limitations:
+  - Source workbook is a partial matrix without bus timing, counters, checksums, or vehicle variant applicability.
+  - Sub IDs are encoded as the first payload byte for virtual execution.
+messages:
+  - name: Climate_Control_064_03
+    frame_id: 100
+    minimum_payload_bytes: 8
+    fixed_signals: {Sub_ID: 3}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Atomic_AC_S, start_bit: 36, bit_length: 2}
+      - {name: Atomic_cycle_mode_S, start_bit: 38, bit_length: 2}
+      - {name: Atomic_defrosting_S, start_bit: 40, bit_length: 2}
+      - {name: Atomic_vehicle_air_conditioning_switch_S, start_bit: 50, bit_length: 2}
+      - {name: Atomic_front_air_conditioning_switch_S, start_bit: 52, bit_length: 2}
+      - {name: Atomic_rear_air_conditioning_switch_S, start_bit: 54, bit_length: 2}
+      - {name: Atomic_driver_air_conditioning_switch_S, start_bit: 56, bit_length: 2}
+      - {name: Atomic_passenger_air_conditioning_switch_S, start_bit: 58, bit_length: 2}
+      - {name: Atomic_left_rear_air_conditioning_switch_S, start_bit: 60, bit_length: 2}
+      - {name: Atomic_right_rear_air_conditioning_switch_S, start_bit: 62, bit_length: 2}
+  - name: Climate_Control_064_04
+    frame_id: 100
+    minimum_payload_bytes: 8
+    fixed_signals: {Sub_ID: 4}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Atomic_vehicle_air_volume_S, start_bit: 8, bit_length: 4}
+      - {name: Atomic_vehicle_air_conditioning_temperature_S, start_bit: 40, bit_length: 8}
+      - {name: Atomic_front_air_conditioning_temperature_S, start_bit: 48, bit_length: 8}
+      - {name: Atomic_rear_air_conditioning_temperature_S, start_bit: 56, bit_length: 8}
+  - name: Climate_Control_064_05
+    frame_id: 100
+    minimum_payload_bytes: 8
+    fixed_signals: {Sub_ID: 5}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Atomic_driver_air_conditioning_temperature_S, start_bit: 8, bit_length: 8}
+      - {name: Atomic_passenger_air_conditioning_temperature_S, start_bit: 16, bit_length: 8}
+      - {name: Atomic_left_rear_air_conditioning_temperature_S, start_bit: 24, bit_length: 8}
+      - {name: Atomic_right_rear_air_conditioning_temperature_S, start_bit: 32, bit_length: 8}
+      - {name: Atomic_driver_air_conditioning_outlet_mode_S, start_bit: 56, bit_length: 4}
+      - {name: Atomic_left_rear_air_conditioning_outlet_mode_S, start_bit: 60, bit_length: 4}
+  - name: Climate_Control_064_06
+    frame_id: 100
+    minimum_payload_bytes: 4
+    fixed_signals: {Sub_ID: 6}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Atomic_co_pilot_air_conditioning_outlet_mode_S, start_bit: 8, bit_length: 4}
+      - {name: Atomic_right_rear_air_conditioning_outlet_mode_S, start_bit: 16, bit_length: 4}
+  - name: Climate_Status_0DB
+    frame_id: 219
+    minimum_payload_bytes: 8
+    signals:
+      - {name: ACHMC_AC_Open_Rslt_Fdbck_Infor_S, start_bit: 0, bit_length: 2}
+      - {name: AV_Close_Results_Feedback_S, start_bit: 2, bit_length: 2}
+      - {name: AC_Cycle_Mode_0DB_S, start_bit: 4, bit_length: 2}
+      - {name: Medium_AC_Switch_S, start_bit: 6, bit_length: 2}
+      - {name: ACHMC_Mast_Dri_Temp_Set_Rslt_S, start_bit: 8, bit_length: 5}
+      - {name: Co_Pilot_AC_Temp_Set_S, start_bit: 16, bit_length: 5}
+      - {name: AC_Wind_Model_0DB_S, start_bit: 28, bit_length: 4}
+      - {name: IKEY_AC_Air_Gear_S, start_bit: 32, bit_length: 4}
+      - {name: ACHMC_AC_Front_Defrosting_Fdbck_S, start_bit: 41, bit_length: 2}
+  - name: Climate_Status_404
+    frame_id: 1028
+    minimum_payload_bytes: 4
+    signals:
+      - {name: ACHMC_AC_On_Status_S, start_bit: 8, bit_length: 2}
+      - {name: AC_Cycle_Mode_404_S, start_bit: 12, bit_length: 2}
+      - {name: AC_Wind_Model_404_S, start_bit: 16, bit_length: 4}
+      - {name: Ac_wind_rnk_S, start_bit: 20, bit_length: 4}
@@ -0,0 +1,36 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: climate-reference-vehicle-mapping
+  name: Climate mapping from supplied CAN matrix
+  version: 1.0.0
+  description: Virtual-only mapping normalized from the supplied climate worksheet.
+spec:
+  domain: climate
+  platform: reference_vehicle
+  dbc_ref: ../../../platforms/reference_vehicle/can/dbc/vehicle.dbc
+  matrix_ref: can-matrix.yaml
+  status: draft
+  mappings:
+    - {capability: climate.set_power, parameter: state, target_zone: vehicle, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_vehicle_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}, feedback_message: Climate_Status_404, feedback_signal: ACHMC_AC_On_Status_S, feedback_handler: compile_enum, feedback_value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: front_row, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_front_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: rear_row, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_rear_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: driver, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_driver_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: passenger, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_passenger_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: rear_left, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_left_rear_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_power, parameter: state, target_zone: rear_right, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_right_rear_air_conditioning_switch_S, value_map: {"on": 1, "off": 2}}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: vehicle, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_04, signal: Atomic_vehicle_air_conditioning_temperature_S, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: front_row, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_04, signal: Atomic_front_air_conditioning_temperature_S, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: rear_row, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_04, signal: Atomic_rear_air_conditioning_temperature_S, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: driver, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_driver_air_conditioning_temperature_S, feedback_message: Climate_Status_0DB, feedback_signal: ACHMC_Mast_Dri_Temp_Set_Rslt_S, feedback_handler: compile_temperature_feedback, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: passenger, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_passenger_air_conditioning_temperature_S, feedback_message: Climate_Status_0DB, feedback_signal: Co_Pilot_AC_Temp_Set_S, feedback_handler: compile_temperature_feedback, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: rear_left, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_left_rear_air_conditioning_temperature_S, unit: celsius}
+    - {capability: climate.set_temperature, parameter: temperature_c, target_zone: rear_right, compiler_handler: compile_integer, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_right_rear_air_conditioning_temperature_S, unit: celsius}
+    - {capability: climate.set_fan_level, parameter: level, target_zone: vehicle, compiler_handler: compile_level_plus_one, bus: cabin_can, message: Climate_Control_064_04, signal: Atomic_vehicle_air_volume_S, feedback_message: Climate_Status_404, feedback_signal: Ac_wind_rnk_S, feedback_handler: compile_integer}
+    - {capability: climate.set_ac, parameter: state, target_zone: vehicle, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_AC_S, value_map: {"on": 1, "off": 2}, feedback_message: Climate_Status_0DB, feedback_signal: Medium_AC_Switch_S, feedback_handler: compile_enum, feedback_value_map: {"on": 2, "off": 1}}
+    - {capability: climate.set_cycle_mode, parameter: mode, target_zone: vehicle, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_cycle_mode_S, value_map: {internal: 1, external: 2}, feedback_message: Climate_Status_0DB, feedback_signal: AC_Cycle_Mode_0DB_S, feedback_handler: compile_enum, feedback_value_map: {internal: 2, external: 1}}
+    - {capability: climate.set_front_defrost, parameter: state, target_zone: vehicle, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_03, signal: Atomic_defrosting_S, value_map: {"on": 1, "off": 2}, feedback_message: Climate_Status_0DB, feedback_signal: ACHMC_AC_Front_Defrosting_Fdbck_S, feedback_handler: compile_enum, feedback_value_map: {"on": 2, "off": 1}}
+    - {capability: climate.set_outlet_mode, parameter: mode, target_zone: driver, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_driver_air_conditioning_outlet_mode_S, value_map: {face: 1, feet: 2}, feedback_message: Climate_Status_0DB, feedback_signal: AC_Wind_Model_0DB_S, feedback_handler: compile_enum, feedback_value_map: {face: 1, feet: 3}}
+    - {capability: climate.set_outlet_mode, parameter: mode, target_zone: rear_left, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_05, signal: Atomic_left_rear_air_conditioning_outlet_mode_S, value_map: {face: 1, feet: 2}}
+    - {capability: climate.set_outlet_mode, parameter: mode, target_zone: passenger, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_06, signal: Atomic_co_pilot_air_conditioning_outlet_mode_S, value_map: {face: 1, feet: 2}, feedback_message: Climate_Status_0DB, feedback_signal: AC_Wind_Model_0DB_S, feedback_handler: compile_enum, feedback_value_map: {face: 1, feet: 3}}
+    - {capability: climate.set_outlet_mode, parameter: mode, target_zone: rear_right, compiler_handler: compile_enum, bus: cabin_can, message: Climate_Control_064_06, signal: Atomic_right_rear_air_conditioning_outlet_mode_S, value_map: {face: 1, feet: 2}}
@@ -0,0 +1,96 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: CapabilityCatalog
+metadata:
+  id: climate-capabilities
+  name: Climate capabilities
+  version: 1.0.0
+spec:
+  domain: climate
+  capabilities:
+    - name: climate.set_power
+      description: Turn HVAC power on or off for the requested cabin zone.
+      target_devices: [hvac]
+      target_scopes: [driver, passenger, front_row, rear_row, rear_left, rear_right, vehicle]
+      parameters:
+        - name: state
+          type: string
+          required: true
+          allowed_values: ["on", "off"]
+          description: Absolute HVAC power state.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: climate.set_temperature
+      description: Set an absolute HVAC temperature from 18 to 32 degrees Celsius.
+      target_devices: [hvac]
+      target_scopes: [driver, passenger, front_row, rear_row, rear_left, rear_right, vehicle]
+      parameters:
+        - name: temperature_c
+          type: number
+          unit: celsius
+          minimum: 18
+          maximum: 32
+          required: true
+          description: Absolute requested temperature in degrees Celsius.
+      risk_level: medium
+      confirmation_policy: on_ambiguity
+    - name: climate.set_fan_level
+      description: Set whole-cabin airflow to an absolute integer level from 0 to 7.
+      target_devices: [hvac]
+      target_scopes: [vehicle]
+      parameters:
+        - name: level
+          type: integer
+          minimum: 0
+          maximum: 7
+          required: true
+          description: Absolute fan level, where 0 is off and 7 is maximum.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: climate.set_ac
+      description: Turn the air-conditioning compressor request on or off.
+      target_devices: [hvac]
+      target_scopes: [vehicle]
+      parameters:
+        - name: state
+          type: string
+          required: true
+          allowed_values: ["on", "off"]
+          description: Absolute compressor request state.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: climate.set_cycle_mode
+      description: Select internal or external cabin air circulation.
+      target_devices: [hvac]
+      target_scopes: [vehicle]
+      parameters:
+        - name: mode
+          type: string
+          required: true
+          allowed_values: [internal, external]
+          description: Absolute circulation mode.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: climate.set_outlet_mode
+      description: Direct a zone outlet toward the face or feet.
+      target_devices: [hvac]
+      target_scopes: [driver, passenger, rear_left, rear_right]
+      parameters:
+        - name: mode
+          type: string
+          required: true
+          allowed_values: [face, feet]
+          description: Absolute outlet direction.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: climate.set_front_defrost
+      description: Turn front windshield defrost on or off.
+      target_devices: [hvac]
+      target_scopes: [vehicle]
+      parameters:
+        - name: state
+          type: string
+          required: true
+          allowed_values: ["on", "off"]
+          description: Absolute front defrost state.
+      risk_level: medium
+      confirmation_policy: on_ambiguity
@@ -0,0 +1,36 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: climate-comfort
+  name: Climate comfort preset
+  version: 1.0.0
+  description: Virtual comfort values backed by the supplied partial matrix.
+spec:
+  domain: climate
+  operations:
+    - id: set_cabin_temperature
+      capability: climate.set_temperature
+      target:
+        device: hvac
+        zone: vehicle
+      parameters:
+        temperature_c: 23
+      transition:
+        profile: device_default
+      timeout_ms: 5000
+      failure_policy: continue
+    - id: set_cabin_fan
+      capability: climate.set_fan_level
+      target:
+        device: hvac
+        zone: vehicle
+      parameters:
+        level: 3
+      transition:
+        duration_ms: 1500
+        profile: ease_in_out
+      timeout_ms: 5000
+      failure_policy: continue
+  locked_parameters: []
+  applicability:
+    platform: reference_vehicle
@@ -0,0 +1,16 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: DomainManifest
+metadata:
+  id: climate-domain-plugin
+  name: Climate domain plugin
+  version: 0.1.0
+spec:
+  domain: climate
+  entrypoint: climate_domain.plugin:create_plugin
+  function_schema: schemas/function-parameters.schema.json
+  capabilities: config/capabilities.yaml
+  signal_mapping: can/signal-mapping.yaml
+  supported_platforms:
+    - reference_vehicle
+  telemetry_required: true
+
@@ -0,0 +1,76 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://cognitive-agent.local/domains/climate/function-parameters.schema.json",
+  "title": "ClimateFunctionParameters",
+  "type": "object",
+  "properties": {
+    "spec": {
+      "type": "object",
+      "properties": {
+        "domain": {"const": "climate"},
+        "operations": {
+          "type": "array",
+          "items": {
+            "oneOf": [
+              {
+                "properties": {
+                  "capability": {"enum": ["climate.set_power", "climate.set_ac", "climate.set_front_defrost"]},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["state"],
+                    "properties": {"state": {"enum": ["on", "off"]}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "climate.set_temperature"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["temperature_c"],
+                    "properties": {"temperature_c": {"type": "number", "minimum": 18, "maximum": 32}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "climate.set_fan_level"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["level"],
+                    "properties": {"level": {"type": "integer", "minimum": 0, "maximum": 7}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "climate.set_cycle_mode"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["mode"],
+                    "properties": {"mode": {"enum": ["internal", "external"]}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "climate.set_outlet_mode"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["mode"],
+                    "properties": {"mode": {"enum": ["face", "feet"]}}
+                  }
+                }
+              }
+            ]
+          }
+        }
+      }
+    }
+  }
+}
@@ -0,0 +1,4 @@
+# Climate Plugin Source
+
+实现 `plugin.py`、`generator.py`、`validator.py`、`compiler.py` 和 `telemetry.py`。分区、摄氏温度、百分比风量必须在 Compiler 中转换为平台定义的 message/signal，不直接发送 CAN。
+
@@ -0,0 +1,4 @@
+# Climate Tests
+
+覆盖温度/风量边界、不同 zone、自动模式冲突、平台不支持、发送超时和状态回读。`integration/` 使用模拟 CAN Gateway；HIL/实车用例单独标记。
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1,8 @@
+domain: roof
+primary_owner: roof_owner
+reviewers:
+  - integration_owner
+  - can_owner
+paths:
+  - domains/roof/**
+
@@ -0,0 +1,7 @@
+# Roof Domain
+
+天幕领域已根据根目录 `空调-天幕-氛围灯CAN信号矩阵表.xlsx` 的“天幕”工作表完成 dry-run 闭环。
+
+当前能力包括调光天幕 0–100% 透光率、后天幕遮阳帘 0–100% 绝对位置和自动调光开关。透光率使用 `0x4C1/0x13`，遮阳帘使用 `0x3E6/0x04`；虚拟反馈分别使用 `0x4BB` 和 `0x47F`，并保留初始化、霍尔和热保护信号作为后续安全接入点。
+
+本文约定遮阳帘 `position_pct` 为 0% 完全关闭、100% 完全打开。源表缺少车型适用性和完整安全互锁，当前严格禁止实车发送。
@@ -0,0 +1,43 @@
+version: 1.0.0
+source:
+  file: 空调-天幕-氛围灯CAN信号矩阵表.xlsx
+  sheet: 天幕
+scope: normalized_roof_subset
+hardware_execution_allowed: false
+limitations:
+  - Source workbook is partial and vehicle-variant applicability is not supplied.
+  - Anti-pinch and motor fault signals are feedback-only safety evidence in this dry-run.
+messages:
+  - name: Roof_Glass_Control_4C1_13
+    frame_id: 1217
+    minimum_payload_bytes: 4
+    fixed_signals: {Sub_ID: 19}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Adjust_The_Skylight_Transmittance_Command_S, start_bit: 16, bit_length: 8}
+      - {name: Canopy_Automatic_Dimming_0x4C1_S, start_bit: 24, bit_length: 2}
+  - name: Roof_Shade_Control_3E6_04
+    frame_id: 998
+    minimum_payload_bytes: 5
+    fixed_signals: {Sub_ID: 4}
+    signals:
+      - {name: Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Rear_Awning_Shade_Screen_Position_Percentage_S, start_bit: 32, bit_length: 8}
+  - name: Roof_Glass_Status_4BB
+    frame_id: 1211
+    minimum_payload_bytes: 57
+    signals:
+      - {name: Current_Transmittance_Of_Skylight_S, start_bit: 444, bit_length: 4}
+      - {name: Current_Status_Of_Dimming_Sunroof_S, start_bit: 448, bit_length: 3}
+      - {name: Canopy_Automatic_Dimming_0x4BB_S, start_bit: 452, bit_length: 2}
+  - name: Roof_Shade_Status_47F
+    frame_id: 1151
+    minimum_payload_bytes: 7
+    signals:
+      - {name: Percentage_Of_Rear_Canopy_Position_47F_S, start_bit: 0, bit_length: 8}
+      - {name: The_Back_Canopy_Closed_State_47F_S, start_bit: 8, bit_length: 2}
+      - {name: Back_Canopy_Initialization_State_47F_S, start_bit: 10, bit_length: 2}
+      - {name: The_Current_Action_Status_Of_The_Back_Sky_47F_S, start_bit: 16, bit_length: 3}
+      - {name: Hall_Signal_Of_Rear_Canopy_Motor_A47F_S, start_bit: 21, bit_length: 1}
+      - {name: Hall_Signal_Of_Rear_Canopy_Motor_B47F_S, start_bit: 22, bit_length: 1}
+      - {name: Thermal_Protection_Status_Of_Rear_Canopy_Motor_47F_S, start_bit: 48, bit_length: 1}
@@ -0,0 +1,48 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: roof-reference-vehicle-mapping
+  name: Roof mapping from supplied CAN matrix
+  version: 1.0.0
+  description: Virtual-only mapping normalized from the supplied roof worksheet.
+spec:
+  domain: roof
+  platform: reference_vehicle
+  dbc_ref: ../../../platforms/reference_vehicle/can/dbc/vehicle.dbc
+  matrix_ref: can-matrix.yaml
+  status: draft
+  mappings:
+    - capability: roof.set_shade_position
+      parameter: position_pct
+      target_zone: vehicle
+      compiler_handler: compile_integer
+      bus: cabin_can
+      message: Roof_Shade_Control_3E6_04
+      signal: Rear_Awning_Shade_Screen_Position_Percentage_S
+      feedback_message: Roof_Shade_Status_47F
+      feedback_signal: Percentage_Of_Rear_Canopy_Position_47F_S
+      feedback_handler: compile_integer
+      unit: percent
+    - capability: roof.set_transmittance
+      parameter: transmittance_pct
+      target_zone: vehicle
+      compiler_handler: compile_integer
+      bus: cabin_can
+      message: Roof_Glass_Control_4C1_13
+      signal: Adjust_The_Skylight_Transmittance_Command_S
+      feedback_message: Roof_Glass_Status_4BB
+      feedback_signal: Current_Transmittance_Of_Skylight_S
+      feedback_handler: compile_transmittance_bucket
+      unit: percent
+    - capability: roof.set_automatic_dimming
+      parameter: state
+      target_zone: vehicle
+      compiler_handler: compile_enum
+      value_map: {"on": 1, "off": 2}
+      bus: cabin_can
+      message: Roof_Glass_Control_4C1_13
+      signal: Canopy_Automatic_Dimming_0x4C1_S
+      feedback_message: Roof_Glass_Status_4BB
+      feedback_signal: Canopy_Automatic_Dimming_0x4BB_S
+      feedback_handler: compile_enum
+      feedback_value_map: {"on": 1, "off": 2}
@@ -0,0 +1,49 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: CapabilityCatalog
+metadata:
+  id: roof-capabilities
+  name: Roof capabilities
+  version: 1.0.0
+spec:
+  domain: roof
+  capabilities:
+    - name: roof.set_shade_position
+      description: Set the rear panoramic roof shade to an absolute position, 0 closed and 100 open.
+      target_devices: [roof_shade]
+      target_scopes: [vehicle]
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          minimum: 0
+          maximum: 100
+          required: true
+          description: Absolute shade opening percentage.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: roof.set_transmittance
+      description: Set dimming canopy light transmittance, 0 darkest and 100 clearest.
+      target_devices: [roof_glass]
+      target_scopes: [vehicle]
+      parameters:
+        - name: transmittance_pct
+          type: integer
+          unit: percent
+          minimum: 0
+          maximum: 100
+          required: true
+          description: Absolute smart-glass transmittance percentage.
+      risk_level: medium
+      confirmation_policy: on_ambiguity
+    - name: roof.set_automatic_dimming
+      description: Turn automatic canopy dimming on or off.
+      target_devices: [roof_glass]
+      target_scopes: [vehicle]
+      parameters:
+        - name: state
+          type: string
+          required: true
+          allowed_values: ["on", "off"]
+          description: Absolute automatic dimming state.
+      risk_level: low
+      confirmation_policy: on_ambiguity
@@ -0,0 +1,24 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: roof-open-view
+  name: Roof open-view preset
+  version: 1.0.0
+  description: Example preset pending platform safety approval.
+spec:
+  domain: roof
+  operations:
+    - id: open_roof_shade
+      capability: roof.set_shade_position
+      target:
+        device: roof_shade
+        zone: vehicle
+      parameters:
+        position_pct: 100
+      transition:
+        profile: device_default
+      timeout_ms: 15000
+      failure_policy: abort
+  locked_parameters: []
+  applicability:
+    vehicle_speed_state: stationary
@@ -0,0 +1,16 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: DomainManifest
+metadata:
+  id: roof-domain-plugin
+  name: Roof domain plugin
+  version: 0.1.0
+spec:
+  domain: roof
+  entrypoint: roof_domain.plugin:create_plugin
+  function_schema: schemas/function-parameters.schema.json
+  capabilities: config/capabilities.yaml
+  signal_mapping: can/signal-mapping.yaml
+  supported_platforms:
+    - reference_vehicle
+  telemetry_required: true
+
@@ -0,0 +1,54 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://cognitive-agent.local/domains/roof/function-parameters.schema.json",
+  "title": "RoofFunctionParameters",
+  "type": "object",
+  "properties": {
+    "spec": {
+      "type": "object",
+      "properties": {
+        "domain": {"const": "roof"},
+        "operations": {
+          "type": "array",
+          "items": {
+            "oneOf": [
+              {
+                "properties": {
+                  "capability": {"const": "roof.set_shade_position"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["position_pct"],
+                    "properties": {"position_pct": {"type": "integer", "minimum": 0, "maximum": 100}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "roof.set_transmittance"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["transmittance_pct"],
+                    "properties": {"transmittance_pct": {"type": "integer", "minimum": 0, "maximum": 100}}
+                  }
+                }
+              },
+              {
+                "properties": {
+                  "capability": {"const": "roof.set_automatic_dimming"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["state"],
+                    "properties": {"state": {"enum": ["on", "off"]}}
+                  }
+                }
+              }
+            ]
+          }
+        }
+      }
+    }
+  }
+}
@@ -0,0 +1,4 @@
+# Roof Plugin Source
+
+实现 `plugin.py`、`generator.py`、`validator.py`、`compiler.py` 和 `telemetry.py`。插件必须区分遮阳帘与调光玻璃，执行前检查平台能力和防夹等安全条件，并只返回 `CommandPlan`。
+
@@ -0,0 +1,4 @@
+# Roof Tests
+
+覆盖位置方向、百分比边界、平台能力缺失、防夹/车速/温度互锁、校准缺失、执行超时和回读不一致。
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
@@ -0,0 +1,8 @@
+domain: seat
+primary_owner: seat_owner
+reviewers:
+  - integration_owner
+  - can_owner
+paths:
+  - domains/seat/**
+
@@ -0,0 +1,25 @@
+# Seat Domain
+
+负责座椅单领域能力。通风和加热基于根目录 `座椅部分功能CAN矩阵表.xlsx` 建立可审计的归一化矩阵；位置电机使用单独的虚构矩阵，仅验证模型理解、配置生成、CAN 编译和反馈闭环。
+
+## 首个交付目标
+
+- 支持显式指令以及“主驾有点热，有点闷”等舒适反馈。
+- GPT 在同一 `conversation_id` 下获取当前虚拟执行状态和前序成功动作，自行解析跟进、因果和作用域。
+- GPT 返回领域无关的通用能力请求；座椅适配器只校验并转换已注册能力，不用本地舒适或位置词条覆盖模型结果。
+- 主副驾分别支持水平、高度、坐垫俯仰、靠背、腿托升降和腿托伸缩六个虚构电机轴，会话状态保留每轴的 `0-100%` 绝对位置。
+- 混合请求中可执行的座椅动作独立 dry-run，未接入的其他域保留为 `deferred_requests`，不会阻断座椅子请求。
+- 生成只含 `domain: seat` 的 `FunctionConfig`。
+- 1–3 档目标进入 `FunctionConfig` 和会话状态；`0x42A` 仍只编译开关，档位由 `0x4FA` dry-run 模拟。
+- 使用模拟 CAN Gateway 测试正常执行、冲突、缺失能力和回读不一致。
+
+## 待负责人补齐
+
+- 0x42A 的正式报文名称、发送节点、周期、DLC、字节序、counter/checksum 和总线所有权。
+- 档位控制信号；当前“设置三档”只是认知目标和虚拟回读，不得视为已产生实车档位命令。
+- 行车状态、乘员状态和防夹等安全互锁。
+- DBC message/signal 以及命令保持、释放和真实回读策略。
+
+`can/seat-thermal-matrix.yaml` 和 `can/seat-motion-virtual-matrix.yaml` 中 `hardware_execution_allowed: false` 是强制边界。后者的帧号、信号布局、时序和反馈行为全部为虚构，当前闭环只能 dry-run。
+
+真实 T2G 台架使用独立的 `signal-mapping.hardware.yaml`、`seat-thermal-hardware-matrix.yaml` 和 `seat-motion-hardware-matrix.yaml`，不会解除上述虚拟矩阵的边界。仓库内 `firmware/t2g/Seat_Controller_demo` 是唯一硬件固件接口，真实路径只接受它的 `combined` profile。当前主驾与副驾都是 `t2g_primary_seat` 的逻辑别名：两种指令使用同一组实体电机、加热和通风，成功反馈会同步到两侧会话状态。完整接线、自动烧录、授权和反馈要求见 [T2G 座椅真实硬件台架接入](../../docs/workflows/t2g-seat-hardware-bench.md)。
@@ -0,0 +1,71 @@
+version: 3.0.0
+source:
+  project: Seat_Move_demo
+  workbook: Seat_Move_demo/Docs/outputs/seat_move_four_motor_20260724/座椅四电机移动功能CAN矩阵表.xlsx
+  test_report: Seat_Move_demo/Docs/TEST_REPORT_2026-07-24.md
+  verified_on: 2026-07-24
+scope: t2g_driver_seat_four_motor_protocol
+hardware_execution_allowed: true
+bit_numbering: lsb0
+byte_order: intel_little_endian
+limitations:
+  - Motor 0 through 2 were hardware verified; Motor 3 is disabled until it is wired and calibrated.
+  - Cushion tilt has no actuator in Seat_Move_demo and is never eligible for hardware transmission.
+  - Mechanical direction labels still require a final visual sign-off on the assembled seat.
+messages:
+  - name: T2G_Seat_Move_Command_300
+    frame_id: 0x300
+    role: command
+    minimum_payload_bytes: 8
+    hardware_send_allowed: true
+    signals:
+      - {name: Opcode, start_bit: 0, bit_length: 8}
+      - {name: Motor_Selector, start_bit: 8, bit_length: 2}
+      - {name: Direction, start_bit: 10, bit_length: 2}
+      - {name: Protocol_Version, start_bit: 12, bit_length: 2}
+      - {name: Sequence, start_bit: 16, bit_length: 16}
+      - {name: Param_A, start_bit: 32, bit_length: 16}
+      - {name: Param_B, start_bit: 48, bit_length: 16}
+  - name: T2G_Seat_Move_System_Status_301
+    frame_id: 0x301
+    role: feedback
+    minimum_payload_bytes: 8
+    signals:
+      - {name: Ack, start_bit: 0, bit_length: 3}
+      - {name: Reject_Reason, start_bit: 3, bit_length: 3}
+      - {name: Shutdown_Inhibit, start_bit: 6, bit_length: 1}
+      - {name: Safety_Inhibit, start_bit: 7, bit_length: 1}
+      - {name: Last_Opcode, start_bit: 8, bit_length: 4}
+      - {name: Active_Mask, start_bit: 12, bit_length: 4}
+      - {name: Last_Sequence, start_bit: 16, bit_length: 16}
+      - {name: Retry_Mask, start_bit: 32, bit_length: 4}
+      - {name: Record_Clean, start_bit: 36, bit_length: 1}
+      - {name: Watchdog_Reset, start_bit: 37, bit_length: 1}
+      - {name: Record_Valid, start_bit: 38, bit_length: 1}
+      - {name: Safety_Fault, start_bit: 40, bit_length: 8}
+      - {name: Config_Sequence, start_bit: 48, bit_length: 16}
+  - name: T2G_Seat_Move_Motor_Status_302
+    frame_id: 0x302
+    role: feedback
+    minimum_payload_bytes: 8
+    signals:
+      - {name: Motor_Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Task_State, start_bit: 8, bit_length: 4}
+      - {name: Result, start_bit: 12, bit_length: 4}
+      - {name: Direction, start_bit: 16, bit_length: 2}
+      - {name: Active, start_bit: 18, bit_length: 1}
+      - {name: Retried, start_bit: 19, bit_length: 1}
+      - {name: Position_Valid, start_bit: 20, bit_length: 1}
+      - {name: Motor_Sequence, start_bit: 24, bit_length: 16}
+      - {name: Value_X100, start_bit: 40, bit_length: 16}
+      - {name: Duty_Percent, start_bit: 56, bit_length: 8}
+  - name: T2G_Seat_Move_Diagnostic_303
+    frame_id: 0x303
+    role: feedback
+    minimum_payload_bytes: 8
+    signals:
+      - {name: Motor_Sub_ID, start_bit: 0, bit_length: 8}
+      - {name: Last_Motor_Event, start_bit: 8, bit_length: 4}
+      - {name: Hall_Count, start_bit: 16, bit_length: 16}
+      - {name: Max_Hall_Count, start_bit: 32, bit_length: 16}
+      - {name: Max_Speed_Pps_X100, start_bit: 48, bit_length: 16}
@@ -0,0 +1,55 @@
+version: 0.1.0
+source:
+  kind: synthetic_demo
+  file: seat-motion-virtual-matrix.yaml
+  designed_on: 2026-07-23
+scope: driver_and_front_passenger_position_motors
+hardware_execution_allowed: false
+synthetic: true
+bit_numbering: lsb0
+byte_order: intel_little_endian
+limitations:
+  - All frame IDs, signal layouts, timing and feedback behavior in this file are fictional.
+  - Values are normalized positions, not physical millimeters or degrees.
+  - No vehicle-state, occupant, anti-pinch, current-sensing or collision interlock is implemented.
+  - These frames must never be sent to a vehicle, bench ECU or HIL network.
+messages:
+  - name: Driver_Seat_Motion_Command_550
+    frame_id: 0x550
+    role: synthetic_command
+    minimum_payload_bytes: 8
+    hardware_send_allowed: false
+    signals: &motion_command_signals
+      - {name: Longitudinal_Target_Pct, description: "Fore-aft absolute target", start_bit: 0, bit_length: 7}
+      - {name: Cushion_Tilt_Target_Pct, description: "Cushion pitch absolute target", start_bit: 7, bit_length: 7}
+      - {name: Backrest_Recline_Target_Pct, description: "Backrest recline absolute target", start_bit: 14, bit_length: 7}
+      - {name: Legrest_Height_Target_Pct, description: "Leg-rest lift absolute target", start_bit: 21, bit_length: 7}
+      - {name: Legrest_Extension_Target_Pct, description: "Leg-rest extension absolute target", start_bit: 28, bit_length: 7}
+      - {name: Motion_Enable_Mask, description: "One enable bit per position axis", start_bit: 35, bit_length: 5}
+      - {name: Command_Counter, description: "Synthetic per-plan command sequence", start_bit: 40, bit_length: 4}
+      - {name: Command_Mode, description: "0=no request, 1=absolute position", start_bit: 44, bit_length: 2}
+      - {name: Command_Checksum, description: "Synthetic additive checksum", start_bit: 48, bit_length: 8}
+  - name: Passenger_Seat_Motion_Command_551
+    frame_id: 0x551
+    role: synthetic_command
+    minimum_payload_bytes: 8
+    hardware_send_allowed: false
+    signals: *motion_command_signals
+  - name: Driver_Seat_Motion_Status_558
+    frame_id: 0x558
+    role: synthetic_feedback
+    minimum_payload_bytes: 8
+    signals: &motion_status_signals
+      - {name: Longitudinal_Actual_Pct, description: "Fore-aft measured position", start_bit: 0, bit_length: 7}
+      - {name: Cushion_Tilt_Actual_Pct, description: "Cushion pitch measured position", start_bit: 7, bit_length: 7}
+      - {name: Backrest_Recline_Actual_Pct, description: "Backrest recline measured position", start_bit: 14, bit_length: 7}
+      - {name: Legrest_Height_Actual_Pct, description: "Leg-rest lift measured position", start_bit: 21, bit_length: 7}
+      - {name: Legrest_Extension_Actual_Pct, description: "Leg-rest extension measured position", start_bit: 28, bit_length: 7}
+      - {name: Moving_Mask, description: "One moving-state bit per position axis", start_bit: 35, bit_length: 5}
+      - {name: Fault_Mask, description: "One synthetic motor-fault bit per position axis", start_bit: 40, bit_length: 5}
+      - {name: Position_Valid_Mask, description: "One valid-feedback bit per position axis", start_bit: 45, bit_length: 5}
+  - name: Passenger_Seat_Motion_Status_559
+    frame_id: 0x559
+    role: synthetic_feedback
+    minimum_payload_bytes: 8
+    signals: *motion_status_signals
@@ -0,0 +1,43 @@
+version: 1.0.0
+source:
+  projects:
+    - Seat_Backrest_Heater_demo
+    - Seat_Fan_demo
+  workbooks:
+    - 座椅加热功能CAN矩阵表.xlsx
+    - 座椅通风功能CAN矩阵表.xlsx
+  verified_on: 2026-07-23
+scope: t2g_driver_seat_heating_and_ventilation
+hardware_execution_allowed: true
+bit_numbering: lsb0
+byte_order: intel_little_endian
+limitations:
+  - Only the driver-seat backrest heater and ventilation fan are implemented by the verified demos.
+  - The two demos are standalone firmware images; use one board per profile or a reviewed combined image.
+  - Product deployment still requires command timeout, E2E protection and calibrated thermal/current limits.
+messages:
+  - name: T2G_Seat_Thermal_Control_29E
+    frame_id: 0x29E
+    role: command
+    minimum_payload_bytes: 8
+    hardware_send_allowed: true
+    signals:
+      - {name: Driver_Seat_Ventilation_On, start_bit: 0, bit_length: 2}
+      - {name: Driver_Seat_Ventilation_Mode, start_bit: 2, bit_length: 2}
+      - {name: Driver_Seat_Heating_On, start_bit: 4, bit_length: 2}
+      - {name: Driver_Seat_Ventilation_Gear, start_bit: 16, bit_length: 4}
+      - {name: Driver_Seat_Heating_Gear, start_bit: 20, bit_length: 4}
+  - name: T2G_Seat_Thermal_Mode_3CE
+    frame_id: 0x3CE
+    role: command
+    minimum_payload_bytes: 8
+    hardware_send_allowed: true
+    signals:
+      - {name: Driver_Seat_Thermal_Mode, start_bit: 4, bit_length: 2}
+  - name: T2G_Seat_Thermal_Status_4FA
+    frame_id: 0x4FA
+    role: feedback
+    minimum_payload_bytes: 8
+    signals:
+      - {name: Driver_Seat_Ventilation_Status, start_bit: 0, bit_length: 3}
+      - {name: Driver_Seat_Heating_Status, start_bit: 3, bit_length: 3}
@@ -0,0 +1,189 @@
+version: 1.0.0
+source:
+  file: ../../../座椅部分功能CAN矩阵表.xlsx
+  sheet: Sheet1
+  normalized_on: 2026-07-22
+scope: driver_and_front_passenger_heating_ventilation
+hardware_execution_allowed: false
+bit_numbering: lsb0_inferred_from_start_bit_and_bit_columns
+byte_order: intel_little_endian_inferred_not_confirmed
+limitations:
+  - Message names are local aliases because the spreadsheet only provides frame IDs.
+  - Transmitter, receiver, cycle time, DLC, byte order, counter and checksum are absent.
+  - Payload lengths below are minimum lengths inferred from the highest occupied bit.
+  - Frame 0x42A is treated as a control candidate only for virtual execution.
+  - Gear signals are observable states; no gear command signal is identified.
+messages:
+  - name: Seat_Thermal_Status_4FA
+    frame_id: 0x4FA
+    role: feedback
+    minimum_payload_bytes: 2
+    signals:
+      - name: Pilot_Seat_Vtilat_Stat_S
+        description: 主驾座椅通风工作状态
+        start_bit: 0
+        bit_length: 3
+        values:
+          0: invalid
+          1: "off"
+          2: low
+          3: high
+          4: level_3
+      - name: Pilot_Seat_Heat_Stat_S
+        description: 主驾座椅加热工作状态
+        start_bit: 3
+        bit_length: 3
+        values:
+          0: invalid
+          1: "off"
+          2: low
+          3: high
+          4: level_3
+      - name: Copilot_Seat_Vtilat_Stat_4FA_S
+        description: 副驾座椅通风工作状态
+        start_bit: 8
+        bit_length: 3
+        values:
+          0: invalid
+          1: "off"
+          2: low
+          3: high
+          4: level_3
+      - name: Copilot_Seat_Heat_Stat_4FA_S
+        description: 副驾座椅加热工作状态
+        start_bit: 11
+        bit_length: 3
+        values:
+          0: invalid
+          1: "off"
+          2: low
+          3: high
+          4: level_3
+
+  - name: Driver_Seat_Thermal_Status_29E
+    frame_id: 0x29E
+    role: feedback
+    minimum_payload_bytes: 3
+    signals:
+      - name: Driver_Seat_Ventilation_On_S
+        description: 主驾座椅通风开启状态
+        start_bit: 0
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Driver_Seat_Ventilation_Control_Mode_S
+        description: 主驾座椅通风控制模式
+        start_bit: 2
+        bit_length: 2
+        values: {0: invalid, 1: manual, 2: automatic, 3: reserved}
+      - name: The_Driver_Seat_Heating_Is_On_S
+        description: 主驾座椅加热开启状态
+        start_bit: 4
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Driver_Seat_Ventilation_Gear_S
+        description: 主驾座椅通风档位
+        start_bit: 16
+        bit_length: 4
+        values: {0: invalid, 1: level_1, 2: level_2, 3: level_3, 4: level_4, 5: level_5}
+      - name: Heated_Driver_Seat_S
+        description: 主驾座椅加热档位
+        start_bit: 20
+        bit_length: 4
+        values: {0: invalid, 1: level_1, 2: level_2, 3: level_3, 4: level_4, 5: level_5}
+
+  - name: Driver_Seat_Thermal_Mode_3CE
+    frame_id: 0x3CE
+    role: feedback
+    minimum_payload_bytes: 1
+    signals:
+      - name: Main_Driver_Seat_Ven_And_Hea_Control_Mode_S
+        description: 主驾座椅通风加热控制模式
+        start_bit: 4
+        bit_length: 2
+        values: {0: invalid, 1: manual, 2: automatic, 3: reserved}
+
+  - name: Passenger_Seat_Thermal_Status_3D3
+    frame_id: 0x3D3
+    role: feedback
+    minimum_payload_bytes: 3
+    signals:
+      - name: Front_Pass_Seat_Ventilation_On_S
+        description: 副驾座椅通风开启状态
+        start_bit: 0
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Pas_Seat_Heating_On_S
+        description: 副驾座椅加热开启状态
+        start_bit: 2
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Front_Pass_Seat_Ven_And_Hea_Control_Mode_S
+        description: 副驾座椅通风加热控制模式
+        start_bit: 4
+        bit_length: 2
+        values: {0: invalid, 1: manual, 2: automatic, 3: reserved}
+      - name: Passenger_Seat_Ventilation_Gear_3D3_S
+        description: 副驾座椅通风档位
+        start_bit: 12
+        bit_length: 4
+        values: {0: invalid, 1: level_1, 2: level_2, 3: level_3, 4: level_4, 5: level_5}
+      - name: Heated_Passenger_Seat_3D3_S
+        description: 副驾座椅加热档位
+        start_bit: 16
+        bit_length: 4
+        values: {0: invalid, 1: level_1, 2: level_2, 3: level_3, 4: level_4, 5: level_5}
+
+  - name: Seat_Thermal_Configuration_2AA_01
+    frame_id: 0x2AA
+    sub_id: 0x01
+    role: configuration_status
+    minimum_payload_bytes: 3
+    signals:
+      - name: Copilot_Seat_Three_Speed_Ventilat_Config_S
+        description: 副驾座椅通风配置
+        start_bit: 8
+        bit_length: 3
+        values: {0: invalid, 1: level_1_config, 2: level_2_config, 3: level_3_config, 4: level_5_split_config}
+      - name: Copilot_Seat_Three_Speed_Heat_Config_S
+        description: 副驾座椅加热配置
+        start_bit: 11
+        bit_length: 3
+        values: {0: invalid, 1: level_1_config, 2: level_2_config, 3: level_3_config, 4: level_5_split_config}
+      - name: Driver_Seat_Three_Speed_Ventilat_Config_S
+        description: 主驾座椅通风配置
+        start_bit: 16
+        bit_length: 3
+        values: {0: invalid, 1: level_1_config, 2: level_2_config, 3: level_3_config, 4: level_5_split_config}
+      - name: Driver_Seat_Three_Speed_Heat_Config_S
+        description: 主驾座椅加热配置
+        start_bit: 19
+        bit_length: 3
+        values: {0: invalid, 1: level_1_config, 2: level_2_config, 3: level_3_config, 4: level_5_split_config}
+
+  - name: Seat_Thermal_Control_42A
+    frame_id: 0x42A
+    role: command_candidate
+    minimum_payload_bytes: 2
+    hardware_send_allowed: false
+    signals:
+      - name: Driver_Seat_Ventilation_S
+        description: 主驾通风开启/关闭
+        start_bit: 0
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Driver_Seat_Heating_S
+        description: 主驾加热开启/关闭
+        start_bit: 2
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Front_Passenger_Seat_Ventilation_S
+        description: 副驾通风开启/关闭
+        start_bit: 12
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+      - name: Front_Passenger_Seat_Heating_S
+        description: 副驾加热开启/关闭
+        start_bit: 14
+        bit_length: 2
+        values: {0: invalid, 1: "on", 2: "off", 3: reserved}
+
@@ -0,0 +1,150 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: seat-t2g-bench-hardware-mapping
+  name: Verified T2G seat bench mapping
+  version: 1.1.0
+  description: Mapping implemented by the unified Seat_Controller_demo IAR firmware.
+spec:
+  domain: seat
+  platform: t2g_seat_bench
+  dbc_ref: ../../../platforms/t2g_seat_bench/can/dbc/seat.dbc
+  matrix_ref: seat-thermal-hardware-matrix.yaml
+  status: hil_verified
+  mappings:
+    - capability: seat.set_ventilation
+      parameter: state
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_thermal
+      bus: seat_can
+      message: T2G_Seat_Thermal_Control_29E
+      signals: [Driver_Seat_Ventilation_On, Driver_Seat_Ventilation_Mode, Driver_Seat_Ventilation_Gear]
+      feedback_message: T2G_Seat_Thermal_Status_4FA
+      feedback_signal: Driver_Seat_Ventilation_Status
+    - capability: seat.set_heating
+      parameter: state
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_thermal
+      bus: seat_can
+      message: T2G_Seat_Thermal_Control_29E
+      signals: [Driver_Seat_Heating_On, Driver_Seat_Heating_Gear]
+      feedback_message: T2G_Seat_Thermal_Status_4FA
+      feedback_signal: Driver_Seat_Heating_Status
+    - capability: seat.set_longitudinal_position
+      parameter: position_pct
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_backrest_recline
+      parameter: position_pct
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_legrest_height
+      parameter: position_pct
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_legrest_extension
+      parameter: position_pct
+      target_zone: front_left
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_ventilation
+      parameter: state
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_thermal
+      bus: seat_can
+      message: T2G_Seat_Thermal_Control_29E
+      signals: [Driver_Seat_Ventilation_On, Driver_Seat_Ventilation_Mode, Driver_Seat_Ventilation_Gear]
+      feedback_message: T2G_Seat_Thermal_Status_4FA
+      feedback_signal: Driver_Seat_Ventilation_Status
+    - capability: seat.set_heating
+      parameter: state
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_thermal
+      bus: seat_can
+      message: T2G_Seat_Thermal_Control_29E
+      signals: [Driver_Seat_Heating_On, Driver_Seat_Heating_Gear]
+      feedback_message: T2G_Seat_Thermal_Status_4FA
+      feedback_signal: Driver_Seat_Heating_Status
+    - capability: seat.set_longitudinal_position
+      parameter: position_pct
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_backrest_recline
+      parameter: position_pct
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_legrest_height
+      parameter: position_pct
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
+    - capability: seat.set_legrest_extension
+      parameter: position_pct
+      target_zone: front_right
+      physical_target_alias: t2g_primary_seat
+      compiler_handler: compile_t2g_absolute_position
+      bus: seat_can
+      message: T2G_Seat_Move_Command_300
+      signal: Param_A
+      feedback_message: T2G_Seat_Move_Motor_Status_302
+      feedback_signal: Value_X100
+      calibration_ref: ../../../platforms/t2g_seat_bench/calibration/seat.yaml
+      unit: percent
@@ -0,0 +1,132 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: SignalMapping
+metadata:
+  id: seat-reference-vehicle-mapping
+  name: Seat mapping for reference vehicle
+  version: 1.1.0
+  description: Supplied thermal matrix plus explicitly synthetic motion mappings; virtual execution only.
+spec:
+  domain: seat
+  platform: reference_vehicle
+  dbc_ref: ../../../platforms/reference_vehicle/can/dbc/vehicle.dbc
+  matrix_ref: seat-thermal-matrix.yaml
+  status: draft
+  mappings:
+    - capability: seat.set_ventilation
+      parameter: state
+      target_zone: front_left
+      compiler_handler: compile_binary_control
+      bus: body_can
+      message: Seat_Thermal_Control_42A
+      signal: Driver_Seat_Ventilation_S
+    - capability: seat.set_heating
+      parameter: state
+      target_zone: front_left
+      compiler_handler: compile_binary_control
+      bus: body_can
+      message: Seat_Thermal_Control_42A
+      signal: Driver_Seat_Heating_S
+    - capability: seat.set_ventilation
+      parameter: state
+      target_zone: front_right
+      compiler_handler: compile_binary_control
+      bus: body_can
+      message: Seat_Thermal_Control_42A
+      signal: Front_Passenger_Seat_Ventilation_S
+    - capability: seat.set_heating
+      parameter: state
+      target_zone: front_right
+      compiler_handler: compile_binary_control
+      bus: body_can
+      message: Seat_Thermal_Control_42A
+      signal: Front_Passenger_Seat_Heating_S
+    - capability: seat.set_longitudinal_position
+      parameter: position_pct
+      target_zone: front_left
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Driver_Seat_Motion_Command_550
+      signal: Longitudinal_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_cushion_tilt
+      parameter: position_pct
+      target_zone: front_left
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Driver_Seat_Motion_Command_550
+      signal: Cushion_Tilt_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_backrest_recline
+      parameter: position_pct
+      target_zone: front_left
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Driver_Seat_Motion_Command_550
+      signal: Backrest_Recline_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_legrest_height
+      parameter: position_pct
+      target_zone: front_left
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Driver_Seat_Motion_Command_550
+      signal: Legrest_Height_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_legrest_extension
+      parameter: position_pct
+      target_zone: front_left
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Driver_Seat_Motion_Command_550
+      signal: Legrest_Extension_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_longitudinal_position
+      parameter: position_pct
+      target_zone: front_right
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Passenger_Seat_Motion_Command_551
+      signal: Longitudinal_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_cushion_tilt
+      parameter: position_pct
+      target_zone: front_right
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Passenger_Seat_Motion_Command_551
+      signal: Cushion_Tilt_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_backrest_recline
+      parameter: position_pct
+      target_zone: front_right
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Passenger_Seat_Motion_Command_551
+      signal: Backrest_Recline_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_legrest_height
+      parameter: position_pct
+      target_zone: front_right
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Passenger_Seat_Motion_Command_551
+      signal: Legrest_Height_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
+    - capability: seat.set_legrest_extension
+      parameter: position_pct
+      target_zone: front_right
+      compiler_handler: compile_position_control
+      bus: body_can
+      message: Passenger_Seat_Motion_Command_551
+      signal: Legrest_Extension_Target_Pct
+      calibration_ref: seat-motion-virtual-matrix.yaml
+      unit: percent
@@ -0,0 +1,211 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: CapabilityCatalog
+metadata:
+  id: seat-capabilities
+  name: Seat capabilities
+  version: 1.2.0
+spec:
+  domain: seat
+  capabilities:
+    - name: seat.set_ventilation
+      description: >-
+        Move air through one seat surface to reduce local occupant heat and moisture; this does not
+        control general cabin airflow. Turn it on or off for the driver or front passenger seat.
+      target_devices:
+        - seat
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: state
+          type: string
+          required: true
+          description: Requested binary ventilation state.
+          allowed_values:
+            - "on"
+            - "off"
+        - name: level
+          type: integer
+          required: false
+          description: Desired 1-3 comfort level; simulated until a level command signal is supplied.
+          allowed_values:
+            - 1
+            - 2
+            - 3
+      risk_level: medium
+      confirmation_policy: on_ambiguity
+    - name: seat.set_heating
+      description: >-
+        Warm one seat surface for local occupant thermal comfort. Turn it on or off for the driver
+        or front passenger seat.
+      target_devices:
+        - seat
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: state
+          type: string
+          required: true
+          description: Requested binary heating state.
+          allowed_values:
+            - "on"
+            - "off"
+        - name: level
+          type: integer
+          required: false
+          description: Desired 1-3 comfort level; simulated until a level command signal is supplied.
+          allowed_values:
+            - 1
+            - 2
+            - 3
+      risk_level: medium
+      confirmation_policy: on_ambiguity
+    - name: seat.set_longitudinal_position
+      description: >-
+        Set one front seat's fore-aft rail position. position_pct is an absolute target: 0 is the
+        rearmost mechanical limit and 100 is the forwardmost limit. Resolve relative requests such
+        as "move back a little" against the current position supplied in runtime state.
+      target_devices:
+        - seat_longitudinal_motor
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          required: true
+          minimum: 0
+          maximum: 100
+          description: Absolute normalized fore-aft target.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: seat.set_cushion_tilt
+      description: >-
+        Set the seat-cushion pitch for one front seat. position_pct is absolute: 0 places the front
+        edge at its lowest limit and 100 raises the front edge to its highest tilt limit.
+      target_devices:
+        - seat_cushion_tilt_motor
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          required: true
+          minimum: 0
+          maximum: 100
+          description: Absolute normalized cushion front-edge height target.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: seat.set_backrest_recline
+      description: >-
+        Set the backrest recline of one front seat. Use either position_pct or angle_deg.
+        position_pct is absolute: 0 is the most upright limit and 100 is the most reclined limit.
+        angle_deg is the included angle between the seat-cushion surface and backrest centreline,
+        from 95 to 140 degrees. "Backrest backward" means a larger value.
+      target_devices:
+        - seat_backrest_motor
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          required: false
+          minimum: 0
+          maximum: 100
+          description: Absolute normalized backrest-recline target.
+        - name: angle_deg
+          type: number
+          unit: degree
+          required: false
+          minimum: 95
+          maximum: 140
+          description: Exact included angle between the cushion surface and backrest centreline.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: seat.set_legrest_height
+      description: >-
+        Raise or lower one front seat's powered leg rest. Use either position_pct or angle_deg.
+        position_pct is absolute: 0 is fully lowered and 100 is fully raised. angle_deg is the
+        included angle between the cushion surface and leg-rest surface, from 95 to 160 degrees.
+        This axis does not change leg-rest extension.
+      target_devices:
+        - seat_legrest_lift_motor
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          required: false
+          minimum: 0
+          maximum: 100
+          description: Absolute normalized leg-rest lift target.
+        - name: angle_deg
+          type: number
+          unit: degree
+          required: false
+          minimum: 95
+          maximum: 160
+          description: Exact included angle between the cushion surface and leg-rest surface.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: seat.set_legrest_extension
+      description: >-
+        Extend or retract one front seat's powered leg rest. position_pct is absolute: 0 is fully
+        retracted and 100 is fully extended. This axis does not change leg-rest height.
+      target_devices:
+        - seat_legrest_extension_motor
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: position_pct
+          type: integer
+          unit: percent
+          required: true
+          minimum: 0
+          maximum: 100
+          description: Absolute normalized leg-rest extension target.
+      risk_level: high
+      confirmation_policy: on_ambiguity
+    - name: seat.save_posture_preset
+      description: >-
+        Save the current five-axis posture of one front seat as a conversation-owned named preset.
+        This is a control-plane operation and does not emit a CAN command. Use only when the user
+        explicitly asks to save, remember or overwrite the current posture.
+      target_devices:
+        - seat_preset_store
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: preset_name
+          type: string
+          required: true
+          description: User-visible name of the posture preset to create or overwrite.
+      risk_level: low
+      confirmation_policy: on_ambiguity
+    - name: seat.apply_posture_preset
+      description: >-
+        Apply a previously saved five-axis posture preset to one front seat. The runtime expands
+        the stored FunctionPreset into the registered motion capabilities before CAN compilation.
+        Use this capability only when the named preset exists in runtime conversation context.
+      target_devices:
+        - seat_preset_store
+      target_scopes:
+        - driver
+        - passenger
+      parameters:
+        - name: preset_name
+          type: string
+          required: true
+          description: Exact user-visible name of an existing posture preset.
+      risk_level: high
+      confirmation_policy: on_ambiguity
@@ -0,0 +1,7 @@
+# Planned Seat Capabilities
+
+零重力姿态和车型级自定义角度仍不属于实车可执行交付。现有 Excel 只覆盖主/副驾通风和加热相关信号，没有滑轨、靠背、坐垫或腿托控制信号，也没有运动学标定。
+
+当前已使用 `seat-motion-virtual-matrix.yaml` 提供五轴 `0-100%` 虚构位置能力，用于模型理解和端到端 dry-run 验证。靠背/坐垫 `95–140°` 和坐垫/腿托 `95–160°` 采用显式 synthetic 线性标定，仅用于验证精确夹角语义、配置生成和孪生展示。该文件不是 DBC，不提供真实物理毫米/角度转换，也不具备防夹、电流、占位、车速或挡位互锁。
+
+当前支持把某一会话中指定座位的五轴 dry-run 状态保存为命名 `FunctionPreset` 并再次调用；这是用户态控制面数据，不是车型出厂零重力标定。获得完整姿态 CAN 矩阵、坐标系、安全互锁和回读定义后，再交付硬件可执行的 `seat.set_posture` 及官方零重力 preset。
@@ -0,0 +1,24 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: seat-driver-ventilation
+  name: Driver seat ventilation
+  version: 1.0.0
+  description: Binary ventilation command supported by the partial CAN matrix.
+spec:
+  domain: seat
+  operations:
+    - id: driver_ventilation_on
+      capability: seat.set_ventilation
+      target:
+        device: seat
+        zone: front_left
+        instance: driver
+      parameters:
+        state: "on"
+      timeout_ms: 3000
+      failure_policy: abort
+  locked_parameters: []
+  applicability:
+    platform: reference_vehicle
+
@@ -0,0 +1,24 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: FunctionPreset
+metadata:
+  id: seat-passenger-heating
+  name: Front passenger seat heating
+  version: 1.0.0
+  description: Binary heating command supported by the partial CAN matrix.
+spec:
+  domain: seat
+  operations:
+    - id: passenger_heating_on
+      capability: seat.set_heating
+      target:
+        device: seat
+        zone: front_right
+        instance: passenger
+      parameters:
+        state: "on"
+      timeout_ms: 3000
+      failure_policy: abort
+  locked_parameters: []
+  applicability:
+    platform: reference_vehicle
+
@@ -0,0 +1,16 @@
+apiVersion: cognitive-agent/v1alpha1
+kind: DomainManifest
+metadata:
+  id: seat-domain-plugin
+  name: Seat domain plugin
+  version: 1.1.0
+spec:
+  domain: seat
+  entrypoint: seat_domain.plugin:create_plugin
+  function_schema: schemas/function-parameters.schema.json
+  capabilities: config/capabilities.yaml
+  signal_mapping: can/signal-mapping.yaml
+  supported_platforms:
+    - reference_vehicle
+    - t2g_seat_bench
+  telemetry_required: true
@@ -0,0 +1,75 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://cognitive-agent.local/domains/seat/function-parameters.schema.json",
+  "title": "SeatThermalFunctionParameters",
+  "type": "object",
+  "properties": {
+    "spec": {
+      "type": "object",
+      "properties": {
+        "domain": {"const": "seat"},
+        "operations": {
+          "type": "array",
+          "items": {
+            "oneOf": [
+              {
+                "type": "object",
+                "properties": {
+                  "capability": {"const": "seat.set_ventilation"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["state"],
+                    "properties": {
+                      "state": {"type": "string", "enum": ["on", "off"]},
+                      "level": {"type": "integer", "minimum": 1, "maximum": 3}
+                    }
+                  }
+                }
+              },
+              {
+                "type": "object",
+                "properties": {
+                  "capability": {"const": "seat.set_heating"},
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["state"],
+                    "properties": {
+                      "state": {"type": "string", "enum": ["on", "off"]},
+                      "level": {"type": "integer", "minimum": 1, "maximum": 3}
+                    }
+                  }
+                }
+              },
+              {
+                "type": "object",
+                "properties": {
+                  "capability": {
+                    "type": "string",
+                    "enum": [
+                      "seat.set_longitudinal_position",
+                      "seat.set_cushion_tilt",
+                      "seat.set_backrest_recline",
+                      "seat.set_legrest_height",
+                      "seat.set_legrest_extension"
+                    ]
+                  },
+                  "parameters": {
+                    "type": "object",
+                    "additionalProperties": false,
+                    "required": ["position_pct"],
+                    "properties": {
+                      "position_pct": {"type": "integer", "minimum": 0, "maximum": 100},
+                      "angle_deg": {"type": "number", "minimum": 95, "maximum": 160}
+                    }
+                  }
+                }
+              }
+            ]
+          }
+        }
+      }
+    }
+  }
+}
@@ -0,0 +1,3 @@
+# Seat Plugin Source
+
+当前实现包含本地意图解析、配置生成、领域校验、确定性 Compiler、矩阵位编码和虚拟回读 Executor。Compiler 只返回 `CommandPlan`，虚拟 Executor 不打开 CAN socket。获得正式 DBC、发送属性、安全输入和审批前不得增加硬件发送路径。
@@ -0,0 +1,25 @@
+"""Seat thermal domain plugin."""
+
+from .compiler import SeatCompiler
+from .executor import VirtualSeatExecutor
+from .generator import build_function_config, build_function_preset, build_intent_ir
+from .hardware_compiler import T2GSeatCompiler
+from .hardware_executor import HardwareExecutionError, HardwareSeatExecutor
+from .intent import parse_local_intent, resolve_contextual_intent
+from .matrix import SeatCanMatrix
+from .validator import validate_semantics
+
+__all__ = [
+    "SeatCanMatrix",
+    "SeatCompiler",
+    "T2GSeatCompiler",
+    "HardwareExecutionError",
+    "HardwareSeatExecutor",
+    "VirtualSeatExecutor",
+    "build_function_config",
+    "build_function_preset",
+    "build_intent_ir",
+    "parse_local_intent",
+    "resolve_contextual_intent",
+    "validate_semantics",
+]
@@ -0,0 +1,133 @@
+"""Deterministically compile seat FunctionConfig into a CommandPlan."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+
+STATE_VALUES = {"on": 1, "off": 2}
+MOTION_AXIS_BITS = {
+    "seat.set_longitudinal_position": 0,
+    "seat.set_cushion_tilt": 1,
+    "seat.set_backrest_recline": 2,
+    "seat.set_legrest_height": 3,
+    "seat.set_legrest_extension": 4,
+}
+MOTION_STATUS_MESSAGES = {
+    "front_left": "Driver_Seat_Motion_Status_558",
+    "front_right": "Passenger_Seat_Motion_Status_559",
+}
+
+
+class SeatCompiler:
+    version = "1.1.0"
+
+    def __init__(self, mapping_path: Path):
+        self.mapping = yaml.safe_load(mapping_path.read_text(encoding="utf-8"))
+        self.mapping_version = self.mapping["metadata"]["version"]
+
+    def _mapping_for(self, capability: str, zone: str) -> dict[str, Any]:
+        for mapping in self.mapping["spec"]["mappings"]:
+            if mapping["capability"] == capability and mapping.get("target_zone") == zone:
+                return mapping
+        raise ValueError(f"No signal mapping for capability={capability}, zone={zone}")
+
+    def compile(self, function_config: dict[str, Any]) -> dict[str, Any]:
+        thermal_groups: dict[tuple[str, str], dict[str, Any]] = {}
+        motion_commands: list[dict[str, Any]] = []
+        motion_index = 0
+
+        for operation in function_config["spec"]["operations"]:
+            mapping = self._mapping_for(operation["capability"], operation["target"]["zone"])
+            handler = mapping["compiler_handler"]
+            if handler == "compile_binary_control":
+                key = (mapping["bus"], mapping["message"])
+                group = thermal_groups.setdefault(key, {"signals": {}, "operation_ids": []})
+                group["signals"][mapping["signal"]] = STATE_VALUES[
+                    operation["parameters"]["state"]
+                ]
+                group["operation_ids"].append(operation["id"])
+                continue
+            if handler != "compile_position_control":
+                raise ValueError(f"Unsupported compiler handler: {handler}")
+
+            target = int(operation["parameters"]["position_pct"])
+            axis_bit = MOTION_AXIS_BITS[operation["capability"]]
+            enable_mask = 1 << axis_bit
+            counter = motion_index % 16
+            checksum = (target + enable_mask + counter + 1) & 0xFF
+            feedback_signal = mapping["signal"].replace("_Target_", "_Actual_")
+            motion_commands.append(
+                {
+                    "id": f"move_{operation['id']}",
+                    "operation_ref": operation["id"],
+                    "bus": mapping["bus"],
+                    "message": mapping["message"],
+                    "signals": dict(
+                        sorted(
+                            {
+                                mapping["signal"]: target,
+                                "Motion_Enable_Mask": enable_mask,
+                                "Command_Counter": counter,
+                                "Command_Mode": 1,
+                                "Command_Checksum": checksum,
+                            }.items()
+                        )
+                    ),
+                    "send_offset_ms": motion_index * 50,
+                    "repeat": {"count": 3, "period_ms": 20},
+                    "completion": {
+                        "message": MOTION_STATUS_MESSAGES[operation["target"]["zone"]],
+                        "signal": feedback_signal,
+                        "operator": "within",
+                        "value": target,
+                        "tolerance": 1,
+                        "timeout_ms": int(operation.get("timeout_ms", 8000)),
+                    },
+                }
+            )
+            motion_index += 1
+
+        canonical = json.dumps(function_config, ensure_ascii=False, sort_keys=True).encode("utf-8")
+        config_hash = hashlib.sha256(canonical).hexdigest()
+        metadata = function_config["metadata"]
+        commands = [
+            {
+                "id": "seat_thermal_control" if index == 0 else f"seat_thermal_control_{index}",
+                "operation_ref": ",".join(group["operation_ids"]),
+                "bus": bus,
+                "message": message,
+                "signals": dict(sorted(group["signals"].items())),
+                "send_offset_ms": 0,
+            }
+            for index, ((bus, message), group) in enumerate(thermal_groups.items())
+        ]
+        commands.extend(motion_commands)
+        return {
+            "apiVersion": "cognitive-agent/v1alpha1",
+            "kind": "CommandPlan",
+            "metadata": {
+                "id": f"plan-{metadata['request_id']}",
+                "name": "Seat CAN command plan",
+                "version": "1.0.0",
+                "request_id": metadata["request_id"],
+                "labels": {"execution_scope": "virtual_only"},
+            },
+            "spec": {
+                "domain": "seat",
+                "platform": "reference_vehicle",
+                "source": {
+                    "config_id": metadata["id"],
+                    "config_version": metadata["version"],
+                    "config_sha256": config_hash,
+                    "compiler_version": self.version,
+                    "mapping_version": self.mapping_version,
+                },
+                "commands": commands,
+            },
+        }
@@ -0,0 +1,199 @@
+"""Virtual CAN executor and feedback simulator for the partial matrix."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+from typing import Any
+
+from .matrix import SeatCanMatrix
+
+
+COMMAND_TO_FEEDBACK = {
+    "Driver_Seat_Ventilation_S": "Pilot_Seat_Vtilat_Stat_S",
+    "Driver_Seat_Heating_S": "Pilot_Seat_Heat_Stat_S",
+    "Front_Passenger_Seat_Ventilation_S": "Copilot_Seat_Vtilat_Stat_4FA_S",
+    "Front_Passenger_Seat_Heating_S": "Copilot_Seat_Heat_Stat_4FA_S",
+}
+OPERATION_TO_COMMAND = {
+    ("front_left", "seat.set_ventilation"): "Driver_Seat_Ventilation_S",
+    ("front_left", "seat.set_heating"): "Driver_Seat_Heating_S",
+    ("front_right", "seat.set_ventilation"): "Front_Passenger_Seat_Ventilation_S",
+    ("front_right", "seat.set_heating"): "Front_Passenger_Seat_Heating_S",
+}
+MOTION_CAPABILITY_TO_AXIS = {
+    "seat.set_longitudinal_position": "longitudinal",
+    "seat.set_cushion_tilt": "cushion_tilt",
+    "seat.set_backrest_recline": "backrest_recline",
+    "seat.set_legrest_height": "legrest_height",
+    "seat.set_legrest_extension": "legrest_extension",
+}
+MOTION_AXIS_BITS = {
+    "longitudinal": 0,
+    "cushion_tilt": 1,
+    "backrest_recline": 2,
+    "legrest_height": 3,
+    "legrest_extension": 4,
+}
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat().replace("+00:00", "Z")
+
+
+class VirtualSeatExecutor:
+    def __init__(self, matrix: SeatCanMatrix):
+        self.matrix = matrix
+
+    def execute(
+        self,
+        command_plan: dict[str, Any],
+        function_config: dict[str, Any],
+        actuator_state: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        started_at = _now()
+        operation_by_id = {
+            operation["id"]: operation
+            for operation in function_config["spec"]["operations"]
+        }
+        results = []
+        for command in command_plan["spec"]["commands"]:
+            operation_ids = command["operation_ref"].split(",")
+            operations = [operation_by_id[item] for item in operation_ids]
+            if command["message"] == "Seat_Thermal_Control_42A":
+                results.extend(self._execute_thermal(command, operations))
+            else:
+                if len(operations) != 1:
+                    raise ValueError("A motion command must reference exactly one operation")
+                results.append(
+                    self._execute_motion(command, operations[0], actuator_state or {})
+                )
+
+        request_id = function_config["metadata"]["request_id"]
+        return {
+            "apiVersion": "cognitive-agent/v1alpha1",
+            "kind": "ExecutionReport",
+            "metadata": {
+                "id": f"report-{request_id}",
+                "name": "Seat virtual execution report",
+                "version": "1.0.0",
+                "request_id": request_id,
+                "labels": {"simulation": "true"},
+            },
+            "spec": {
+                "mode": "dry_run",
+                "status": "succeeded",
+                "started_at": started_at,
+                "finished_at": _now(),
+                "results": results,
+            },
+        }
+
+    def _execute_thermal(
+        self, command: dict[str, Any], operations: list[dict[str, Any]]
+    ) -> list[dict[str, Any]]:
+        command_frame = self.matrix.encode(command["message"], command["signals"])
+        operation_by_signal = {
+            OPERATION_TO_COMMAND[(operation["target"]["zone"], operation["capability"])]: operation
+            for operation in operations
+        }
+        feedback_values = {}
+        for signal, value in command["signals"].items():
+            operation = operation_by_signal[signal]
+            level = int(operation["parameters"].get("level", 1))
+            feedback_values[COMMAND_TO_FEEDBACK[signal]] = level + 1 if value == 1 else 1
+        feedback_frame = self.matrix.encode("Seat_Thermal_Status_4FA", feedback_values)
+
+        results = []
+        for operation in operations:
+            command_signal = OPERATION_TO_COMMAND[
+                (operation["target"]["zone"], operation["capability"])
+            ]
+            feedback_signal = COMMAND_TO_FEEDBACK[command_signal]
+            state = operation["parameters"]["state"]
+            requested_level = operation["parameters"].get("level")
+            effective_level = (requested_level or 1) if state == "on" else 0
+            results.append(
+                {
+                    "ref": operation["id"],
+                    "domain": "seat",
+                    "status": "succeeded",
+                    "message": (
+                        f"虚拟回读确认 {operation['target']['instance']} "
+                        f"{operation['capability']}={state}, level={effective_level}"
+                    ),
+                    "telemetry": {
+                        "requested_state": state,
+                        "requested_level": requested_level,
+                        "effective_level": effective_level,
+                        "level_command_simulated": requested_level is not None,
+                        "command_signal": command_signal,
+                        "command_value": command["signals"][command_signal],
+                        "feedback_signal": feedback_signal,
+                        "feedback_value": feedback_values[feedback_signal],
+                        "command_frame": command_frame,
+                        "feedback_frame": feedback_frame,
+                        "simulated": True,
+                    },
+                }
+            )
+        return results
+
+    def _execute_motion(
+        self,
+        command: dict[str, Any],
+        operation: dict[str, Any],
+        actuator_state: dict[str, Any],
+    ) -> dict[str, Any]:
+        capability = operation["capability"]
+        axis = MOTION_CAPABILITY_TO_AXIS[capability]
+        instance = operation["target"]["instance"]
+        target = int(operation["parameters"]["position_pct"])
+        target_angle = operation["parameters"].get("angle_deg")
+        start = int(
+            actuator_state.get(instance, {})
+            .get("position", {})
+            .get(axis, {})
+            .get("position_pct", target)
+        )
+        command_frame = self.matrix.encode(command["message"], command["signals"])
+        axis_mask = 1 << MOTION_AXIS_BITS[axis]
+        completion = command["completion"]
+        feedback_values = {
+            completion["signal"]: target,
+            "Moving_Mask": 0,
+            "Fault_Mask": 0,
+            "Position_Valid_Mask": axis_mask,
+        }
+        feedback_frame = self.matrix.encode(completion["message"], feedback_values)
+        midpoint = round((start + target) / 2)
+        samples = list(dict.fromkeys([start, midpoint, target]))
+        return {
+            "ref": operation["id"],
+            "domain": "seat",
+            "status": "succeeded",
+            "message": (
+                f"虚构电机反馈确认 {instance} {axis} 从 {start}% 到达 {target}%"
+            ),
+            "telemetry": {
+                "motion_axis": axis,
+                "start_position_pct": start,
+                "requested_position_pct": target,
+                "effective_position_pct": target,
+                "requested_angle_deg": target_angle,
+                "effective_angle_deg": target_angle,
+                "position_tolerance_pct": completion.get("tolerance", 0),
+                "motion_samples_pct": samples,
+                "motion_state": "stopped",
+                "moving_mask": 0,
+                "fault_mask": 0,
+                "position_valid_mask": axis_mask,
+                "command_signal": next(
+                    name for name in command["signals"] if name.endswith("_Target_Pct")
+                ),
+                "feedback_signal": completion["signal"],
+                "command_frame": command_frame,
+                "feedback_frame": feedback_frame,
+                "matrix_source": "synthetic_demo",
+                "simulated": True,
+            },
+        }
@@ -0,0 +1,277 @@
+"""Generate public intent and function configuration documents."""
+
+from __future__ import annotations
+
+from datetime import UTC, datetime
+from typing import Any
+
+
+TARGETS = {
+    "driver": {"device": "seat", "zone": "front_left", "instance": "driver"},
+    "passenger": {"device": "seat", "zone": "front_right", "instance": "passenger"},
+}
+CAPABILITIES = {
+    "ventilation": "seat.set_ventilation",
+    "heating": "seat.set_heating",
+}
+MOTION_CAPABILITIES = {
+    "longitudinal": "seat.set_longitudinal_position",
+    "cushion_tilt": "seat.set_cushion_tilt",
+    "backrest_recline": "seat.set_backrest_recline",
+    "legrest_height": "seat.set_legrest_height",
+    "legrest_extension": "seat.set_legrest_extension",
+}
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat().replace("+00:00", "Z")
+
+
+def build_intent_ir(
+    parsed: dict[str, Any], utterance: str, request_id: str, provider: str
+) -> dict[str, Any]:
+    requests = []
+    for action in parsed["actions"]:
+        if action.get("action_type") == "position":
+            parameters = {
+                "axis": action["axis"],
+                "position_pct": action["position_pct"],
+            }
+            if action.get("angle_deg") is not None:
+                parameters["angle_deg"] = action["angle_deg"]
+            requests.append(
+                {
+                    "domain": "seat",
+                    "goal": (
+                        f"Set {action['target']} {action['axis']} position "
+                        f"to {action['position_pct']} percent"
+                    ),
+                    "target": TARGETS[action["target"]],
+                    "parameters": parameters,
+                    "source_span": utterance,
+                }
+            )
+            continue
+        parameters: dict[str, Any] = {
+            "function": action["function"],
+            "state": action["state"],
+        }
+        if action.get("requested_level") is not None:
+            parameters["requested_level"] = action["requested_level"]
+        requests.append(
+            {
+                "domain": "seat",
+                "goal": f"Set {action['target']} {action['function']} {action['state']}",
+                "target": TARGETS[action["target"]],
+                "parameters": parameters,
+                "source_span": utterance,
+            }
+        )
+
+    for preset_action in parsed.get("preset_actions", []):
+        requests.append(
+            {
+                "domain": "seat",
+                "goal": (
+                    f"Save the current {preset_action['target']} posture as "
+                    f"{preset_action['preset_name']}"
+                ),
+                "target": TARGETS[preset_action["target"]],
+                "parameters": {
+                    "capability": "seat.save_posture_preset",
+                    "preset_name": preset_action["preset_name"],
+                },
+                "source_span": utterance,
+            }
+        )
+
+    for preset_event in parsed.get("preset_events", []):
+        requests.append(
+            {
+                "domain": "seat",
+                "goal": (
+                    f"Apply saved posture {preset_event['preset_name']} to "
+                    f"{preset_event['target']}"
+                ),
+                "target": TARGETS[preset_event["target"]],
+                "parameters": {
+                    "capability": "seat.apply_posture_preset",
+                    "preset_name": preset_event["preset_name"],
+                },
+                "source_span": utterance,
+            }
+        )
+
+    for deferred in parsed.get("deferred_requests", []):
+        requests.append(
+            {
+                "domain": deferred["domain"],
+                "goal": f"{deferred['desired_change']} {deferred['capability']}",
+                "target": {
+                    "device": deferred["domain"],
+                    "zone": deferred["target_scope"],
+                },
+                "parameters": {
+                    "capability": deferred["capability"],
+                    "desired_change": deferred["desired_change"],
+                    "requested_value": deferred.get("requested_value"),
+                    "requested_parameters": deferred.get("parameters", {}),
+                    "availability": "catalog_only",
+                },
+                "source_span": utterance,
+            }
+        )
+
+    detected_domains = parsed.get("detected_domains", ["seat"])
+
+    return {
+        "apiVersion": "cognitive-agent/v1alpha1",
+        "kind": "IntentIR",
+        "metadata": {
+            "id": f"intent-{request_id}",
+            "name": "Seat function intent",
+            "version": "1.0.0",
+            "request_id": request_id,
+            "generated_at": _now(),
+            "labels": {"provider": provider},
+        },
+        "spec": {
+            "utterance": utterance,
+            "language": "zh-CN",
+            "mode": "cross_domain" if len(detected_domains) > 1 else "single_domain",
+            "detected_domains": detected_domains,
+            "requests": requests,
+            "ambiguities": parsed["ambiguities"],
+            "confidence": parsed["confidence"],
+            "requires_confirmation": bool(parsed["ambiguities"]),
+        },
+    }
+
+
+def build_function_config(
+    parsed: dict[str, Any], request_id: str, provider: str
+) -> dict[str, Any] | None:
+    if not parsed["actions"]:
+        return None
+
+    operations = []
+    assumptions = [
+        "The supplied matrix is partial; this configuration is eligible for virtual execution only."
+    ]
+    for index, action in enumerate(parsed["actions"], start=1):
+        if action.get("action_type") == "position":
+            operation_id = (
+                f"{action['target']}_{action['axis']}_{action['position_pct']}pct"
+            )
+            if any(item["id"] == operation_id for item in operations):
+                operation_id = f"{operation_id}_{index}"
+            parameters: dict[str, Any] = {"position_pct": action["position_pct"]}
+            if action.get("angle_deg") is not None:
+                parameters["angle_deg"] = action["angle_deg"]
+            operation = {
+                "id": operation_id,
+                "capability": MOTION_CAPABILITIES[action["axis"]],
+                "target": TARGETS[action["target"]],
+                "parameters": parameters,
+                "timeout_ms": 8000,
+                "failure_policy": "abort",
+            }
+            if action.get("motion_intent") is not None:
+                operation["motion"] = dict(action["motion_intent"])
+            operations.append(operation)
+            assumptions.append(
+                f"Operation {operation_id} uses an explicitly synthetic position CAN mapping; "
+                "it is never eligible for hardware transmission."
+            )
+            continue
+        operation_id = f"{action['target']}_{action['function']}_{action['state']}"
+        if any(item["id"] == operation_id for item in operations):
+            operation_id = f"{operation_id}_{index}"
+        parameters: dict[str, Any] = {"state": action["state"]}
+        if action.get("requested_level") is not None:
+            parameters["level"] = action["requested_level"]
+        operations.append(
+            {
+                "id": operation_id,
+                "capability": CAPABILITIES[action["function"]],
+                "target": TARGETS[action["target"]],
+                "parameters": parameters,
+                "timeout_ms": 3000,
+                "failure_policy": "abort",
+            }
+        )
+        if action.get("requested_level") is not None:
+            assumptions.append(
+                f"Operation {operation_id} targets level {action['requested_level']}; "
+                "the level is simulated because the supplied matrix has no level command signal."
+            )
+
+    return {
+        "apiVersion": "cognitive-agent/v1alpha1",
+        "kind": "FunctionConfig",
+        "metadata": {
+            "id": f"seat-{request_id}",
+            "name": "Seat function",
+            "version": "1.0.0",
+            "request_id": request_id,
+            "generated_at": _now(),
+            "labels": {"provider": provider, "execution_scope": "virtual_only"},
+        },
+        "spec": {"domain": "seat", "operations": operations, "assumptions": assumptions},
+    }
+
+
+def build_function_preset(
+    target: str,
+    preset_name: str,
+    device_state: dict[str, Any],
+    request_id: str,
+    provider: str,
+) -> dict[str, Any]:
+    operations = []
+    for axis, capability in MOTION_CAPABILITIES.items():
+        motion_state = device_state["position"][axis]
+        parameters: dict[str, Any] = {
+            "position_pct": int(motion_state["position_pct"])
+        }
+        if motion_state.get("angle_deg") is not None:
+            parameters["angle_deg"] = motion_state["angle_deg"]
+        operations.append(
+            {
+                "id": f"{target}_{axis}_preset",
+                "capability": capability,
+                "target": TARGETS[target],
+                "parameters": parameters,
+                "timeout_ms": 8000,
+                "failure_policy": "abort",
+            }
+        )
+    return {
+        "apiVersion": "cognitive-agent/v1alpha1",
+        "kind": "FunctionPreset",
+        "metadata": {
+            "id": f"seat-preset-{request_id}",
+            "name": preset_name,
+            "version": "1.0.0",
+            "request_id": request_id,
+            "generated_at": _now(),
+            "description": f"User-saved five-axis posture for {target}.",
+            "labels": {
+                "provider": provider,
+                "target": target,
+                "storage_scope": "conversation_durable",
+            },
+        },
+        "spec": {
+            "domain": "seat",
+            "operations": operations,
+            "locked_parameters": [
+                f"operations.{index}.parameters.position_pct"
+                for index in range(len(operations))
+            ],
+            "applicability": {
+                "target": target,
+                "execution_mode": "dry_run",
+            },
+        },
+    }
@@ -0,0 +1,303 @@
+"""Compile seat functions to the CAN protocol implemented by the T2G IAR demos."""
+
+from __future__ import annotations
+
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+
+STATE_VALUES = {"on": 1, "off": 2}
+
+
+class T2GSeatCompiler:
+    """Deterministic compiler for the HIL-verified T2G seat bench protocol."""
+
+    version = "1.3.0"
+
+    def __init__(self, mapping_path: Path, calibration_path: Path):
+        self.mapping = yaml.safe_load(mapping_path.read_text(encoding="utf-8"))
+        self.calibration = yaml.safe_load(calibration_path.read_text(encoding="utf-8"))
+        self.mapping_version = self.mapping["metadata"]["version"]
+        self.axes = self.calibration["axes"]
+        self.protocol = self.calibration["protocol"]
+        self.supported = {
+            (item["capability"], item.get("target_zone"))
+            for item in self.mapping["spec"]["mappings"]
+        }
+
+    def compile(
+        self,
+        function_config: dict[str, Any],
+        actuator_state: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        operations = function_config["spec"]["operations"]
+        self._validate_shared_physical_targets(operations)
+        thermal_operations = [
+            item
+            for item in operations
+            if item["capability"] in {"seat.set_ventilation", "seat.set_heating"}
+        ]
+        motion_operations = [item for item in operations if item not in thermal_operations]
+        commands: list[dict[str, Any]] = []
+
+        if thermal_operations:
+            commands.extend(
+                self._compile_thermal(thermal_operations, actuator_state or {})
+            )
+        commands.extend(self._compile_motion(motion_operations, len(commands)))
+
+        canonical = json.dumps(
+            function_config, ensure_ascii=False, sort_keys=True
+        ).encode("utf-8")
+        metadata = function_config["metadata"]
+        return {
+            "apiVersion": "cognitive-agent/v1alpha1",
+            "kind": "CommandPlan",
+            "metadata": {
+                "id": f"plan-t2g-{metadata['request_id']}",
+                "name": "T2G seat hardware CAN command plan",
+                "version": "1.0.0",
+                "request_id": metadata["request_id"],
+                "labels": {
+                    "execution_scope": "hardware_candidate",
+                    "protocol": "t2g_seat_v1",
+                },
+            },
+            "spec": {
+                "domain": "seat",
+                "platform": "t2g_seat_bench",
+                "source": {
+                    "config_id": metadata["id"],
+                    "config_version": metadata["version"],
+                    "config_sha256": hashlib.sha256(canonical).hexdigest(),
+                    "compiler_version": self.version,
+                    "mapping_version": self.mapping_version,
+                },
+                "commands": commands,
+            },
+        }
+
+    def _compile_thermal(
+        self,
+        operations: list[dict[str, Any]],
+        actuator_state: dict[str, Any],
+    ) -> list[dict[str, Any]]:
+        state = actuator_state.get("driver") or actuator_state.get("passenger", {})
+        desired = {
+            "ventilation": dict(state.get("ventilation", {"state": "off", "level": 0})),
+            "heating": dict(state.get("heating", {"state": "off", "level": 0})),
+        }
+        for operation in operations:
+            self._require_supported(operation)
+            feature = operation["capability"].removeprefix("seat.set_")
+            requested_state = str(operation["parameters"]["state"])
+            previous_level = int(desired[feature].get("level", 0))
+            level = (
+                int(operation["parameters"].get("level", previous_level or 1))
+                if requested_state == "on"
+                else 0
+            )
+            desired[feature] = {"state": requested_state, "level": level}
+
+        operation_ref = ",".join(item["id"] for item in operations)
+        commands: list[dict[str, Any]] = []
+        if desired["heating"]["state"] == "on":
+            commands.append(
+                {
+                    "id": "t2g_seat_thermal_mode",
+                    "operation_ref": operation_ref,
+                    "bus": "seat_can",
+                    "message": "T2G_Seat_Thermal_Mode_3CE",
+                    "signals": {"Driver_Seat_Thermal_Mode": 1},
+                    "send_offset_ms": 0,
+                }
+            )
+        commands.append(
+            {
+                "id": "t2g_seat_thermal_control",
+                "operation_ref": operation_ref,
+                "bus": "seat_can",
+                "message": "T2G_Seat_Thermal_Control_29E",
+                "signals": {
+                    "Driver_Seat_Heating_Gear": int(desired["heating"]["level"]),
+                    "Driver_Seat_Heating_On": STATE_VALUES[desired["heating"]["state"]],
+                    "Driver_Seat_Ventilation_Gear": int(desired["ventilation"]["level"]),
+                    "Driver_Seat_Ventilation_Mode": 1,
+                    "Driver_Seat_Ventilation_On": STATE_VALUES[
+                        desired["ventilation"]["state"]
+                    ],
+                },
+                "send_offset_ms": 20 if commands else 0,
+            }
+        )
+        return commands
+
+    def _compile_motion(
+        self, operations: list[dict[str, Any]], command_offset: int
+    ) -> list[dict[str, Any]]:
+        grouped: list[list[dict[str, Any]]] = []
+        by_capability: dict[str, list[dict[str, Any]]] = {}
+        for operation in operations:
+            self._require_supported(operation)
+            capability = operation["capability"]
+            if capability not in by_capability:
+                by_capability[capability] = []
+                grouped.append(by_capability[capability])
+            by_capability[capability].append(operation)
+
+        commands = []
+        for index, physical_operations in enumerate(grouped):
+            operation = physical_operations[0]
+            capability = operation["capability"]
+            try:
+                axis = self.axes[capability]
+            except KeyError as exc:
+                raise ValueError(
+                    f"T2G IAR firmware has no calibrated actuator for {capability}"
+                ) from exc
+            maximum = int(axis["maximum_value_x100"])
+            position_pct = int(operation["parameters"]["position_pct"])
+            target = round(maximum * position_pct / 100)
+            tolerance = max(1, round(maximum / 100))
+            motion = operation.get("motion")
+            if motion is not None and motion.get("mode") == "relative":
+                delta_pct = int(motion["delta_position_pct"])
+                maximum_delta = int(axis["max_uncalibrated_relative_pct"])
+                if delta_pct <= maximum_delta:
+                    commands.append(
+                        self._compile_uncalibrated_relative(
+                            operation,
+                            physical_operations,
+                            axis,
+                            command_offset + index,
+                        )
+                    )
+                    continue
+                # A large relative request cannot safely be approximated by a
+                # long open-loop duty command or by several chained commands.
+                # Compile its already-resolved target as an absolute move.  The
+                # executor will check the board's live Position_Valid bit before
+                # putting Opcode 1 on the bus.
+            commands.append(
+                {
+                    "id": f"t2g_move_{operation['id']}",
+                    "operation_ref": ",".join(
+                        item["id"] for item in physical_operations
+                    ),
+                    "bus": "seat_can",
+                    "message": "T2G_Seat_Move_Command_300",
+                    "signals": {
+                        "Direction": 0,
+                        "Motor_Selector": int(axis["motor_selector"]),
+                        "Opcode": int(self.protocol["absolute_position_opcode"]),
+                        "Param_A": target,
+                        "Param_B": int(self.protocol["default_duty_percent"]),
+                        "Protocol_Version": int(self.protocol["version"]),
+                        # Allocated immediately before hardware transmission. Zero keeps
+                        # the compiled plan deterministic and is never put on the bus.
+                        "Sequence": 0,
+                    },
+                    "send_offset_ms": (command_offset + index) * 100,
+                    "completion": {
+                        "message": "T2G_Seat_Move_Motor_Status_302",
+                        "signal": "Value_X100",
+                        "operator": "within",
+                        "value": target,
+                        "tolerance": tolerance,
+                        "timeout_ms": int(operation.get("timeout_ms", 8000)),
+                    },
+                }
+            )
+        return commands
+
+    def _compile_uncalibrated_relative(
+        self,
+        operation: dict[str, Any],
+        physical_operations: list[dict[str, Any]],
+        axis: dict[str, Any],
+        command_index: int,
+    ) -> dict[str, Any]:
+        motion = dict(operation["motion"])
+        delta_pct = int(motion["delta_position_pct"])
+        maximum_delta = int(axis["max_uncalibrated_relative_pct"])
+        if delta_pct <= 0:
+            raise ValueError(
+                "relative seat movement resolved to zero travel; no motor command was generated"
+            )
+        if delta_pct > maximum_delta:
+            raise ValueError(
+                "internal compiler error: an oversized relative movement must be "
+                "compiled as a calibrated absolute target"
+            )
+
+        duration_ms = delta_pct * int(axis["timed_duty_ms_per_pct"])
+        duration_ms = max(
+            int(self.protocol["timed_duty_min_duration_ms"]), duration_ms
+        )
+        maximum_duration = int(self.protocol["timed_duty_max_duration_ms"])
+        if duration_ms > maximum_duration:
+            raise ValueError(
+                "uncalibrated relative seat movement would exceed the bounded "
+                f"{maximum_duration} ms timed-motion limit"
+            )
+        direction_name = str(motion["direction"])
+        direction = int(self.protocol[f"{direction_name}_direction"])
+        timeout_ms = max(
+            int(operation.get("timeout_ms", 8000)), duration_ms + 3000
+        )
+        return {
+            "id": f"t2g_move_{operation['id']}",
+            "operation_ref": ",".join(
+                item["id"] for item in physical_operations
+            ),
+            "bus": "seat_can",
+            "message": "T2G_Seat_Move_Command_300",
+            "signals": {
+                "Direction": direction,
+                "Motor_Selector": int(axis["motor_selector"]),
+                "Opcode": int(self.protocol["timed_duty_opcode"]),
+                "Param_A": duration_ms,
+                "Param_B": int(self.protocol["default_duty_percent"]),
+                "Protocol_Version": int(self.protocol["version"]),
+                "Sequence": 0,
+            },
+            "send_offset_ms": command_index * 100,
+            "completion": {
+                "message": "T2G_Seat_Move_Motor_Status_302",
+                "signal": "Task_State",
+                "operator": "eq",
+                "value": 8,
+                "timeout_ms": timeout_ms,
+            },
+        }
+
+    @staticmethod
+    def _validate_shared_physical_targets(
+        operations: list[dict[str, Any]],
+    ) -> None:
+        requested: dict[str, dict[str, Any]] = {}
+        for operation in operations:
+            capability = operation["capability"]
+            request = {
+                "parameters": dict(operation.get("parameters", {})),
+                "motion": dict(operation["motion"]) if operation.get("motion") else None,
+            }
+            previous = requested.get(capability)
+            if previous is not None and previous != request:
+                raise ValueError(
+                    "driver and passenger aliases request conflicting values for "
+                    f"the same physical capability {capability}"
+                )
+            requested[capability] = request
+
+    def _require_supported(self, operation: dict[str, Any]) -> None:
+        key = (operation["capability"], operation["target"]["zone"])
+        if key not in self.supported:
+            raise ValueError(
+                "T2G hardware mapping does not support "
+                f"capability={key[0]}, zone={key[1]}"
+            )
@@ -0,0 +1,713 @@
+"""Hardware-in-the-loop executor for the verified T2G seat CAN protocol."""
+
+from __future__ import annotations
+
+import threading
+import time
+from datetime import UTC, datetime
+from typing import Any, Callable
+
+from cognitive_agent.infrastructure.can.gateway import CanBus, CanFrame
+from cognitive_agent.infrastructure.firmware import T2GFirmwareGuard
+
+from .executor import MOTION_CAPABILITY_TO_AXIS
+from .matrix import SeatCanMatrix
+
+
+MOTOR_RESULT_NAMES = {
+    0: "NONE",
+    1: "RUNNING",
+    2: "OK",
+    3: "STOPPED",
+    4: "STALL",
+    5: "ENDPOINT",
+    6: "TIMEOUT",
+    7: "DRIVER_FAULT",
+    8: "INVALID",
+    9: "CALIBRATION_FAILED",
+}
+
+MOTOR_EVENT_NAMES = {
+    0: "NONE",
+    1: "TARGET_REACHED",
+    2: "NEAR_ENDPOINT",
+    3: "FAR_ENDPOINT",
+    4: "STALL",
+    5: "TIMEOUT",
+    6: "ABORTED",
+    7: "DRIVER_FAULT",
+    8: "CALIBRATION_FAILED",
+}
+
+
+class HardwareExecutionError(RuntimeError):
+    """Raised for rejected, faulted or timed-out physical operations."""
+
+    def __init__(
+        self, message: str, *, telemetry: dict[str, Any] | None = None
+    ) -> None:
+        super().__init__(message)
+        self.telemetry = dict(telemetry or {})
+
+
+def _now() -> str:
+    return datetime.now(UTC).isoformat().replace("+00:00", "Z")
+
+
+class HardwareSeatExecutor:
+    """Serialize access to one adapter and require real firmware feedback."""
+
+    _adapter_lock = threading.Lock()
+    _sequence = 0
+
+    def __init__(
+        self,
+        matrix: SeatCanMatrix,
+        bus_factory: Callable[[], CanBus],
+        *,
+        thermal_timeout_ms: int = 2_000,
+        ack_timeout_ms: int = 1_500,
+        axis_maximum_x100: dict[str, int] | None = None,
+        firmware_guard: T2GFirmwareGuard | None = None,
+    ):
+        self.matrix = matrix
+        self.bus_factory = bus_factory
+        self.thermal_timeout_ms = thermal_timeout_ms
+        self.ack_timeout_ms = ack_timeout_ms
+        self.axis_maximum_x100 = dict(axis_maximum_x100 or {})
+        self.firmware_guard = firmware_guard
+
+    @classmethod
+    def _next_sequence(cls) -> int:
+        cls._sequence = (cls._sequence + 1) & 0xFFFF
+        if cls._sequence == 0:
+            cls._sequence = 1
+        return cls._sequence
+
+    def execute(
+        self,
+        command_plan: dict[str, Any],
+        function_config: dict[str, Any],
+        actuator_state: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        del actuator_state  # physical feedback, not virtual state, is authoritative here
+        started_at = _now()
+        operations = {
+            item["id"]: item for item in function_config["spec"]["operations"]
+        }
+        results: list[dict[str, Any]] = []
+        commands = command_plan["spec"]["commands"]
+        has_motion = any(item["message"] == "T2G_Seat_Move_Command_300" for item in commands)
+        bus: CanBus | None = None
+        motion_armed = False
+
+        with self._adapter_lock:
+            try:
+                if self.firmware_guard is not None:
+                    bus, firmware_telemetry = self.firmware_guard.prepare_bus()
+                    results.append(
+                        {
+                            "ref": "seat-controller-firmware",
+                            "domain": "seat",
+                            "status": "succeeded",
+                            "message": (
+                                "repository Seat_Controller_demo was programmed and verified"
+                                if firmware_telemetry.get("programmed_this_request")
+                                else "repository Seat_Controller_demo identity was verified"
+                            ),
+                            "telemetry": {
+                                **firmware_telemetry,
+                                "simulated": False,
+                            },
+                        }
+                    )
+                else:
+                    bus = self.bus_factory()
+                if has_motion:
+                    self._stop_all(bus, require_ack=True)
+                    motion_armed = True
+                previous_offset = 0
+                for command in commands:
+                    offset = int(command.get("send_offset_ms", 0))
+                    if offset > previous_offset:
+                        time.sleep((offset - previous_offset) / 1000)
+                    previous_offset = offset
+                    operation_ids = command["operation_ref"].split(",")
+                    referenced = [operations[item] for item in operation_ids]
+                    if command["message"] == "T2G_Seat_Thermal_Mode_3CE":
+                        frame = self._send_encoded(bus, command["message"], command["signals"])
+                        results.append(
+                            {
+                                "ref": command["id"],
+                                "domain": "seat",
+                                "status": "succeeded",
+                                "message": "T2G heater manual control mode sent to the physical CAN bus",
+                                "telemetry": {
+                                    "control_step": "thermal_mode",
+                                    "command_frame": frame.public(),
+                                    "simulated": False,
+                                },
+                            }
+                        )
+                    elif command["message"] == "T2G_Seat_Thermal_Control_29E":
+                        results.extend(self._execute_thermal(bus, command, referenced))
+                    elif command["message"] == "T2G_Seat_Move_Command_300":
+                        results.extend(self._execute_motion(bus, command, referenced))
+                    else:
+                        raise HardwareExecutionError(
+                            f"hardware executor refuses unknown message {command['message']}"
+                        )
+            except Exception as exc:  # noqa: BLE001 - normalize transport and protocol failures
+                if bus is not None and motion_armed:
+                    self._best_effort_stop(bus)
+                operation_ref = self._first_unreported_operation(operations, results)
+                results.append(
+                    {
+                        "ref": operation_ref,
+                        "domain": "seat",
+                        "status": "failed",
+                        "message": str(exc),
+                        "telemetry": {
+                            **getattr(exc, "telemetry", {}),
+                            "error_type": type(exc).__name__,
+                            "stop_all_attempted": bool(bus is not None and motion_armed),
+                            "simulated": False,
+                        },
+                    }
+                )
+            finally:
+                if bus is not None:
+                    bus.close()
+
+        succeeded = all(item["status"] == "succeeded" for item in results)
+        request_id = function_config["metadata"]["request_id"]
+        return {
+            "apiVersion": "cognitive-agent/v1alpha1",
+            "kind": "ExecutionReport",
+            "metadata": {
+                "id": f"report-{request_id}",
+                "name": "T2G seat hardware execution report",
+                "version": "1.0.0",
+                "request_id": request_id,
+                "labels": {"simulation": "false", "feedback": "physical_can"},
+            },
+            "spec": {
+                "mode": "hil",
+                "status": "succeeded" if succeeded else "failed",
+                "started_at": started_at,
+                "finished_at": _now(),
+                "results": results,
+            },
+        }
+
+    @staticmethod
+    def _first_unreported_operation(
+        operations: dict[str, dict[str, Any]], results: list[dict[str, Any]]
+    ) -> str:
+        reported = {item["ref"] for item in results}
+        return next((item for item in operations if item not in reported), next(iter(operations)))
+
+    def _send_encoded(
+        self, bus: CanBus, message: str, signals: dict[str, int]
+    ) -> CanFrame:
+        encoded = self.matrix.encode(message, signals)
+        return bus.send(int(encoded["frame_id"], 16), bytes.fromhex(encoded["data_hex"]))
+
+    def _receive_until(
+        self,
+        bus: CanBus,
+        arbitration_id: int,
+        timeout_ms: int,
+        predicate: Callable[[dict[str, int], CanFrame], bool],
+    ) -> tuple[dict[str, int], CanFrame]:
+        deadline = time.monotonic() + timeout_ms / 1000
+        while True:
+            remaining_ms = round((deadline - time.monotonic()) * 1000)
+            if remaining_ms <= 0:
+                raise HardwareExecutionError(
+                    f"timed out after {timeout_ms} ms waiting for CAN 0x{arbitration_id:03X}"
+                )
+            frame = bus.receive(remaining_ms)
+            if frame is None:
+                raise HardwareExecutionError(
+                    f"timed out after {timeout_ms} ms waiting for CAN 0x{arbitration_id:03X}"
+                )
+            if frame.arbitration_id != arbitration_id or len(frame.data) < 8:
+                continue
+            values = self.matrix.decode_by_frame_id(arbitration_id, frame.data)
+            if predicate(values, frame):
+                return values, frame
+
+    def _wait_ack(
+        self, bus: CanBus, sequence: int, opcode: int
+    ) -> tuple[dict[str, int], CanFrame]:
+        def matches(values: dict[str, int], _: CanFrame) -> bool:
+            if values["Last_Sequence"] != sequence or values["Last_Opcode"] != opcode:
+                return False
+            if values["Ack"] != 1:
+                raise HardwareExecutionError(
+                    "T2G movement command rejected: "
+                    f"ack={values['Ack']}, reason={values['Reject_Reason']}, "
+                    f"safety_fault={values['Safety_Fault']}"
+                )
+            if values["Safety_Inhibit"] or values["Safety_Fault"]:
+                raise HardwareExecutionError(
+                    f"T2G movement safety inhibit/fault: {values['Safety_Fault']}"
+                )
+            return True
+
+        return self._receive_until(bus, 0x301, self.ack_timeout_ms, matches)
+
+    def _stop_all(self, bus: CanBus, *, require_ack: bool) -> tuple[CanFrame, CanFrame | None]:
+        sequence = self._next_sequence()
+        signals = {
+            "Opcode": 9,
+            "Motor_Selector": 3,
+            "Direction": 0,
+            "Protocol_Version": 1,
+            "Sequence": sequence,
+            "Param_A": 0,
+            "Param_B": 0,
+        }
+        sent = self._send_encoded(bus, "T2G_Seat_Move_Command_300", signals)
+        if not require_ack:
+            return sent, None
+        _, ack = self._wait_ack(bus, sequence, 9)
+        return sent, ack
+
+    def _best_effort_stop(self, bus: CanBus) -> None:
+        try:
+            self._stop_all(bus, require_ack=False)
+        except Exception:  # noqa: BLE001 - already handling the primary failure
+            pass
+
+    def _execute_thermal(
+        self,
+        bus: CanBus,
+        command: dict[str, Any],
+        operations: list[dict[str, Any]],
+    ) -> list[dict[str, Any]]:
+        command_frame = self._send_encoded(bus, command["message"], command["signals"])
+        expected: dict[str, int] = {}
+        for operation in operations:
+            feature = operation["capability"].removeprefix("seat.set_")
+            signal = (
+                "Driver_Seat_Heating_Status"
+                if feature == "heating"
+                else "Driver_Seat_Ventilation_Status"
+            )
+            state = operation["parameters"]["state"]
+            level = int(operation["parameters"].get("level", 1))
+            expected[signal] = 1 if state == "off" else min(level, 3) + 1
+
+        def matches(values: dict[str, int], _: CanFrame) -> bool:
+            return all(values[name] == value for name, value in expected.items())
+
+        values, feedback = self._receive_until(
+            bus, 0x4FA, self.thermal_timeout_ms, matches
+        )
+        output = []
+        for operation in operations:
+            feature = operation["capability"].removeprefix("seat.set_")
+            signal = (
+                "Driver_Seat_Heating_Status"
+                if feature == "heating"
+                else "Driver_Seat_Ventilation_Status"
+            )
+            raw = values[signal]
+            output.append(
+                {
+                    "ref": operation["id"],
+                    "domain": "seat",
+                    "status": "succeeded",
+                    "message": (
+                        f"physical 0x4FA confirmed shared front-seat {feature} "
+                        f"state={operation['parameters']['state']}, level={max(0, raw - 1)}"
+                    ),
+                    "telemetry": {
+                        "requested_target": operation["target"]["instance"],
+                        "physical_target": "t2g_primary_seat",
+                        "requested_state": operation["parameters"]["state"],
+                        "requested_level": operation["parameters"].get("level"),
+                        "feedback_signal": signal,
+                        "feedback_value": raw,
+                        "command_frame": command_frame.public(),
+                        "feedback_frame": feedback.public(),
+                        "hardware_feedback": True,
+                        "simulated": False,
+                    },
+                }
+            )
+        return output
+
+    def _execute_motion(
+        self,
+        bus: CanBus,
+        command: dict[str, Any],
+        operations: list[dict[str, Any]],
+    ) -> list[dict[str, Any]]:
+        if not operations:
+            raise HardwareExecutionError(
+                "a T2G motion frame must reference at least one operation"
+            )
+        operation = operations[0]
+        if any(
+            item["capability"] != operation["capability"]
+            or item["parameters"] != operation["parameters"]
+            or item.get("motion") != operation.get("motion")
+            for item in operations[1:]
+        ):
+            raise HardwareExecutionError(
+                "one shared T2G motion frame cannot represent conflicting operations"
+            )
+        signals = dict(command["signals"])
+        motor = int(signals["Motor_Selector"])
+        opcode = int(signals["Opcode"])
+        if opcode == 3:
+            return self._execute_timed_motion(
+                bus, command, operations, signals, motor
+            )
+
+        preflight_values, preflight_frame = self._motor_status_snapshot(bus, motor)
+        if not preflight_values["Position_Valid"]:
+            raise HardwareExecutionError(
+                f"T2G motor {motor} absolute-position command blocked before send: "
+                "controller Position_Valid=0. Run the explicit endpoint-calibration "
+                "maintenance procedure with the bench clear and emergency stop ready, "
+                "or use a small relative movement request.",
+                telemetry={
+                    "motor_selector": motor,
+                    "position_valid": False,
+                    "calibration_required": True,
+                    "preflight_frame": preflight_frame.public(),
+                },
+            )
+
+        sequence = self._next_sequence()
+        signals["Sequence"] = sequence
+        command_frame = self._send_encoded(bus, command["message"], signals)
+        ack_values, ack_frame = self._wait_ack(bus, sequence, opcode)
+        target = int(command["completion"]["value"])
+        tolerance = int(command["completion"].get("tolerance", 0))
+
+        def complete(values: dict[str, int], _: CanFrame) -> bool:
+            if values["Motor_Sub_ID"] != motor or values["Motor_Sequence"] != sequence:
+                return False
+            task_state = values["Task_State"]
+            result = values["Result"]
+            if task_state == 9 or result not in {1, 2}:
+                raise HardwareExecutionError(
+                    f"T2G motor {motor} failed: task_state={task_state}, result={result}"
+                )
+            if task_state != 8 or result != 2:
+                return False
+            if not values["Position_Valid"]:
+                raise HardwareExecutionError(
+                    f"T2G motor {motor} completed without a valid calibrated position"
+                )
+            if abs(values["Value_X100"] - target) > tolerance:
+                raise HardwareExecutionError(
+                    f"T2G motor {motor} stopped outside tolerance: "
+                    f"actual={values['Value_X100']}, target={target}, tolerance={tolerance}"
+                )
+            return True
+
+        values, feedback = self._receive_until(
+            bus,
+            0x302,
+            int(command["completion"]["timeout_ms"]),
+            complete,
+        )
+        axis = MOTION_CAPABILITY_TO_AXIS[operation["capability"]]
+        maximum = self.axis_maximum_x100.get(operation["capability"])
+        if maximum is None:
+            requested_pct = int(operation["parameters"]["position_pct"])
+            maximum = max(1, target * 100 // max(1, requested_pct))
+        actual_pct = round(values["Value_X100"] * 100 / maximum)
+        common_telemetry = {
+                "motion_axis": axis,
+                "motor_selector": motor,
+                "physical_target": "t2g_primary_seat",
+                "sequence": sequence,
+                "requested_position_pct": int(operation["parameters"]["position_pct"]),
+                "effective_position_pct": actual_pct,
+                "requested_value_x100": target,
+                "effective_value_x100": values["Value_X100"],
+                "position_tolerance_x100": tolerance,
+                "task_state": values["Task_State"],
+                "motor_result": values["Result"],
+                "position_valid": bool(values["Position_Valid"]),
+                "system_ack": ack_values["Ack"],
+                "preflight_frame": preflight_frame.public(),
+                "command_frame": command_frame.public(),
+                "ack_frame": ack_frame.public(),
+                "feedback_frame": feedback.public(),
+                "hardware_feedback": True,
+                "simulated": False,
+        }
+        return [
+            {
+                "ref": item["id"],
+                "domain": "seat",
+                "status": "succeeded",
+                "message": (
+                    f"physical motor {motor} feedback confirmed shared front-seat "
+                    f"{axis} at {actual_pct}%"
+                ),
+                "telemetry": {
+                    **common_telemetry,
+                    "requested_target": item["target"]["instance"],
+                },
+            }
+            for item in operations
+        ]
+
+    def _motor_status_snapshot(
+        self, bus: CanBus, motor: int
+    ) -> tuple[dict[str, int], CanFrame]:
+        return self._receive_until(
+            bus,
+            0x302,
+            max(self.ack_timeout_ms, 1_000),
+            lambda values, _: values["Motor_Sub_ID"] == motor,
+        )
+
+    def _motor_diagnostic_snapshot(
+        self, bus: CanBus, motor: int
+    ) -> tuple[dict[str, int], CanFrame]:
+        return self._receive_until(
+            bus,
+            0x303,
+            max(self.ack_timeout_ms, 1_000),
+            lambda values, _: values["Motor_Sub_ID"] == motor,
+        )
+
+    def _execute_timed_motion(
+        self,
+        bus: CanBus,
+        command: dict[str, Any],
+        operations: list[dict[str, Any]],
+        signals: dict[str, int],
+        motor: int,
+    ) -> list[dict[str, Any]]:
+        baseline, baseline_frame = self._motor_diagnostic_snapshot(bus, motor)
+        sequence = self._next_sequence()
+        signals["Sequence"] = sequence
+        command_frame = self._send_encoded(bus, command["message"], signals)
+        ack_values, ack_frame = self._wait_ack(bus, sequence, 3)
+
+        def terminal(values: dict[str, int], _: CanFrame) -> bool:
+            if values["Motor_Sub_ID"] != motor or values["Motor_Sequence"] != sequence:
+                return False
+            task_state = values["Task_State"]
+            result = values["Result"]
+            return task_state in {8, 9} or result not in {1, 2}
+
+        values, feedback = self._receive_until(
+            bus,
+            0x302,
+            int(command["completion"]["timeout_ms"]),
+            terminal,
+        )
+        diagnostic: dict[str, int] | None = None
+        diagnostic_frame: CanFrame | None = None
+        diagnostic_error: str | None = None
+        try:
+            diagnostic, diagnostic_frame = self._motor_diagnostic_snapshot(bus, motor)
+        except HardwareExecutionError as exc:
+            diagnostic_error = str(exc)
+
+        hall_before = int(baseline["Hall_Count"])
+        hall_after = (
+            int(diagnostic["Hall_Count"]) if diagnostic is not None else hall_before
+        )
+        direction = int(signals["Direction"])
+        task_state = int(values["Task_State"])
+        result = int(values["Result"])
+        result_name = MOTOR_RESULT_NAMES.get(result, f"UNKNOWN_{result}")
+        last_event = (
+            int(diagnostic["Last_Motor_Event"])
+            if diagnostic is not None
+            else None
+        )
+        last_event_name = (
+            MOTOR_EVENT_NAMES.get(last_event, f"UNKNOWN_{last_event}")
+            if last_event is not None
+            else "UNAVAILABLE"
+        )
+        hall_delta = abs(hall_after - hall_before)
+        max_hall = max(
+            1,
+            int(diagnostic["Max_Hall_Count"])
+            if diagnostic is not None
+            else int(baseline["Max_Hall_Count"]),
+        )
+
+        if task_state != 8 or result != 2:
+            if result == 4 and hall_delta <= 2:
+                failure_category = "stall_no_hall_activity"
+                diagnostic_hint = (
+                    "No sustained Hall activity was observed. The commanded direction may "
+                    "already be at a mechanical endpoint or obstructed, or the Motor Hall "
+                    "power/signal path may be disconnected. With the bench clear, use a "
+                    "small command in the opposite direction to distinguish endpoint from "
+                    "a Hall wiring fault; do not run automatic full-range calibration."
+                )
+            elif result == 4:
+                failure_category = "stall_after_hall_activity"
+                diagnostic_hint = (
+                    "Hall activity was observed before the stall; inspect the mechanism for "
+                    "an endpoint or obstruction in the commanded direction."
+                )
+            elif result == 5 or last_event in {2, 3}:
+                failure_category = "endpoint"
+                diagnostic_hint = (
+                    "The controller reported an endpoint. Verify the requested direction "
+                    "and establish endpoint calibration before absolute positioning."
+                )
+            else:
+                failure_category = result_name.lower()
+                diagnostic_hint = (
+                    "Inspect the controller result, last motor event, driver supply and Hall "
+                    "feedback before retrying the movement."
+                )
+            raise HardwareExecutionError(
+                f"T2G timed motor {motor} {result_name}: task_state={task_state}, "
+                f"Hall={hall_before}->{hall_after}, last_event={last_event_name}. "
+                f"{diagnostic_hint}",
+                telemetry={
+                    "motor_selector": motor,
+                    "motion_axis": MOTION_CAPABILITY_TO_AXIS[
+                        operations[0]["capability"]
+                    ],
+                    "motion_mode": "relative_timed_duty",
+                    "position_valid": bool(values["Position_Valid"]),
+                    "task_state": task_state,
+                    "motor_result": result,
+                    "motor_result_name": result_name,
+                    "last_motor_event": last_event,
+                    "last_motor_event_name": last_event_name,
+                    "failure_category": failure_category,
+                    "diagnostic_hint": diagnostic_hint,
+                    "hall_before": hall_before,
+                    "hall_after": hall_after,
+                    "hall_delta": hall_delta,
+                    "max_hall_count": max_hall,
+                    "direction": direction,
+                    "duration_ms": int(signals["Param_A"]),
+                    "duty_percent": int(signals["Param_B"]),
+                    "command_frame": command_frame.public(),
+                    "ack_frame": ack_frame.public(),
+                    "feedback_frame": feedback.public(),
+                    "diagnostic_before_frame": baseline_frame.public(),
+                    "diagnostic_frame": (
+                        diagnostic_frame.public()
+                        if diagnostic_frame is not None
+                        else None
+                    ),
+                    "diagnostic_read_error": diagnostic_error,
+                },
+            )
+
+        if diagnostic is None or diagnostic_frame is None:
+            raise HardwareExecutionError(
+                f"T2G timed motor {motor} completed but diagnostic Hall feedback "
+                f"was unavailable: {diagnostic_error}",
+                telemetry={
+                    "motor_selector": motor,
+                    "position_valid": bool(values["Position_Valid"]),
+                    "task_state": task_state,
+                    "motor_result": result,
+                    "motor_result_name": result_name,
+                    "hall_before": hall_before,
+                    "command_frame": command_frame.public(),
+                    "ack_frame": ack_frame.public(),
+                    "feedback_frame": feedback.public(),
+                    "diagnostic_before_frame": baseline_frame.public(),
+                    "diagnostic_read_error": diagnostic_error,
+                },
+            )
+
+        direction_verified: bool | None
+        if direction == 1:
+            direction_verified = hall_after > hall_before
+        elif hall_before > 0:
+            direction_verified = hall_after < hall_before
+        else:
+            # At an uncalibrated zero estimate, NEAR movement cannot decrement
+            # the public Hall position. COMPLETE/OK still proves raw Hall edges
+            # because the firmware refuses timed success without one.
+            direction_verified = None
+        if direction_verified is False:
+            raise HardwareExecutionError(
+                f"T2G timed motor {motor} completed but Hall direction did not match "
+                f"the command: before={hall_before}, after={hall_after}, direction={direction}",
+                telemetry={
+                    "motor_selector": motor,
+                    "position_valid": bool(values["Position_Valid"]),
+                    "hall_before": hall_before,
+                    "hall_after": hall_after,
+                    "hall_direction_verified": False,
+                    "feedback_frame": feedback.public(),
+                    "diagnostic_frame": diagnostic_frame.public(),
+                },
+            )
+
+        axis = MOTION_CAPABILITY_TO_AXIS[operations[0]["capability"]]
+        common_telemetry = {
+            "motion_axis": axis,
+            "motion_mode": "relative_timed_duty",
+            "motor_selector": motor,
+            "physical_target": "t2g_primary_seat",
+            "sequence": sequence,
+            "requested_position_pct": int(
+                operations[0]["parameters"]["position_pct"]
+            ),
+            "requested_delta_position_pct": int(
+                operations[0].get("motion", {}).get("delta_position_pct", 0)
+            ),
+            "requested_delta_angle_deg": operations[0]
+            .get("motion", {})
+            .get("delta_angle_deg"),
+            "duration_ms": int(signals["Param_A"]),
+            "duty_percent": int(signals["Param_B"]),
+            "direction": direction,
+            "task_state": task_state,
+            "motor_result": result,
+            "motor_result_name": result_name,
+            "last_motor_event": last_event,
+            "last_motor_event_name": last_event_name,
+            "position_valid": bool(values["Position_Valid"]),
+            "absolute_position_available": bool(values["Position_Valid"]),
+            "hall_before": hall_before,
+            "hall_after": hall_after,
+            "hall_delta": hall_delta,
+            "max_hall_count": max_hall,
+            "observed_delta_position_pct": round(hall_delta * 100 / max_hall),
+            "hall_direction_verified": direction_verified,
+            "system_ack": ack_values["Ack"],
+            "command_frame": command_frame.public(),
+            "ack_frame": ack_frame.public(),
+            "feedback_frame": feedback.public(),
+            "diagnostic_before_frame": baseline_frame.public(),
+            "diagnostic_frame": diagnostic_frame.public(),
+            "hardware_feedback": True,
+            "simulated": False,
+        }
+        return [
+            {
+                "ref": item["id"],
+                "domain": "seat",
+                "status": "succeeded",
+                "message": (
+                    f"physical motor {motor} COMPLETE/OK confirmed relative {axis} "
+                    f"movement with Hall {hall_before}->{hall_after}; absolute position "
+                    "remains unavailable until endpoint calibration"
+                ),
+                "telemetry": {
+                    **common_telemetry,
+                    "requested_target": item["target"]["instance"],
+                },
+            }
+            for item in operations
+        ]
@@ -0,0 +1,433 @@
+"""Deterministic fallback parser for the supported seat thermal vocabulary."""
+
+from __future__ import annotations
+
+from copy import deepcopy
+import re
+from typing import Any
+
+
+TARGET_PATTERN = re.compile(r"主副驾|前排|主驾|驾驶位|司机位|副驾|乘客位")
+FEATURE_PATTERN = re.compile(r"通风|加热")
+STATE_PATTERN = re.compile(r"关闭|关掉|停止|取消|打开|开启|启动|开")
+LEVEL_PATTERN = re.compile(r"([1-5一二三四五])\s*档")
+VENTILATION_FEEDBACK_PATTERN = re.compile(r"(?<!不)(?<!加)热|(?<!不)闷|出汗")
+HEATING_FEEDBACK_PATTERN = re.compile(r"(?<!不)冷|(?<!不)凉")
+MILD_PATTERN = re.compile(r"有点|一点|稍微|微微")
+STRONG_PATTERN = re.compile(r"很|太|非常|特别|难受")
+AIRFLOW_CAUSE_PATTERN = re.compile(r"吹|风太大|风吹")
+FIRST_PERSON_PATTERN = re.compile(r"我|自己")
+VEHICLE_SCOPE_PATTERN = re.compile(r"车上|车内|车里|全车|整车|整个车")
+
+TARGETS = {
+    "主驾": ["driver"],
+    "驾驶位": ["driver"],
+    "司机位": ["driver"],
+    "副驾": ["passenger"],
+    "乘客位": ["passenger"],
+    "主副驾": ["driver", "passenger"],
+    "前排": ["driver", "passenger"],
+}
+FEATURES = {"通风": "ventilation", "加热": "heating"}
+STATES = {
+    "打开": "on",
+    "开启": "on",
+    "启动": "on",
+    "开": "on",
+    "关闭": "off",
+    "关掉": "off",
+    "停止": "off",
+    "取消": "off",
+}
+LEVELS = {"一": 1, "二": 2, "三": 3, "四": 4, "五": 5}
+
+
+def _nearest_preceding_target(clause: str, position: int) -> str | None:
+    candidates = [match for match in TARGET_PATTERN.finditer(clause) if match.start() <= position]
+    if candidates:
+        return candidates[-1].group(0)
+    matches = list(TARGET_PATTERN.finditer(clause))
+    return matches[0].group(0) if len(matches) == 1 else None
+
+
+def _state_for_feature(clause: str, position: int) -> str | None:
+    states = list(STATE_PATTERN.finditer(clause))
+    if not states:
+        return None
+    following = [match for match in states if match.start() >= position]
+    if following and following[0].start() - position <= 8:
+        return STATES[following[0].group(0)]
+    if len(states) == 1:
+        return STATES[states[0].group(0)]
+    preceding = [match for match in states if match.start() < position]
+    return STATES[preceding[-1].group(0)] if preceding else None
+
+
+def _level_for_clause(clause: str) -> int | None:
+    match = LEVEL_PATTERN.search(clause)
+    if not match:
+        return None
+    token = match.group(1)
+    return int(token) if token.isdigit() else LEVELS[token]
+
+
+def _context_device(
+    context: dict[str, Any] | None, target: str, function: str
+) -> dict[str, Any]:
+    if not context:
+        return {"state": "off", "level": 0}
+    return (
+        context.get("devices", {})
+        .get(target, {})
+        .get(function, {"state": "off", "level": 0})
+    )
+
+
+def _feedback_targets(
+    text: str, context: dict[str, Any] | None, parsed: dict[str, Any]
+) -> tuple[list[str], str]:
+    matches = list(TARGET_PATTERN.finditer(text))
+    if matches:
+        targets: list[str] = []
+        for match in matches:
+            for target in TARGETS[match.group(0)]:
+                if target not in targets:
+                    targets.append(target)
+        return targets, "explicit"
+    parsed_targets = list(
+        dict.fromkeys(
+            action.get("target")
+            for action in parsed.get("actions", [])
+            if action.get("target") in {"driver", "passenger"}
+        )
+    )
+    if parsed_targets:
+        return parsed_targets, "model"
+    last_target = context.get("last_target") if context else None
+    if last_target in {"driver", "passenger"}:
+        return [last_target], "history"
+    if FIRST_PERSON_PATTERN.search(text):
+        return ["driver"], "speaker_default"
+    return [], "missing"
+
+
+def _inferred_level(text: str) -> tuple[int, str]:
+    explicit_level = _level_for_clause(text)
+    if explicit_level is not None:
+        return explicit_level, "explicit"
+    if STRONG_PATTERN.search(text):
+        return 3, "strong"
+    if MILD_PATTERN.search(text):
+        return 1, "mild"
+    return 2, "normal"
+
+
+def _decrease_action(
+    target: str, function: str, current: dict[str, Any]
+) -> tuple[dict[str, Any], int, int]:
+    previous_level = int(current.get("level") or 0)
+    if current.get("state") == "on" and previous_level > 1:
+        next_level = previous_level - 1
+        return (
+            {
+                "target": target,
+                "function": function,
+                "state": "on",
+                "requested_level": next_level,
+            },
+            previous_level,
+            next_level,
+        )
+    return (
+        {
+            "target": target,
+            "function": function,
+            "state": "off",
+            "requested_level": None,
+        },
+        previous_level,
+        0,
+    )
+
+
+def resolve_contextual_intent(
+    parsed: dict[str, Any], utterance: str, context: dict[str, Any] | None = None
+) -> dict[str, Any]:
+    """Resolve implicit comfort feedback against the current virtual seat state."""
+
+    result = deepcopy(parsed)
+    text = re.sub(r"\s+", "", utterance.strip())
+    ventilation_cues = list(dict.fromkeys(VENTILATION_FEEDBACK_PATTERN.findall(text)))
+    cold_cues = list(dict.fromkeys(HEATING_FEEDBACK_PATTERN.findall(text)))
+    if not ventilation_cues and not cold_cues:
+        return result
+
+    if VEHICLE_SCOPE_PATTERN.search(text) and not TARGET_PATTERN.search(text):
+        result["actions"] = []
+        result["ambiguities"] = [
+            {
+                "field": "domain",
+                "question": "该请求作用于整车舱空气，需要由空调/新风域执行；当前座椅单域不能代替执行。",
+            }
+        ]
+        result["decision_summary"] = [
+            "“车上/车内”表示整车舱范围，不是主驾或副驾单座位。",
+            "“太闷”属于舱内空气质量/新风需求，应路由到 climate 域。",
+            "当前未接入空调执行器，未将整车请求错误下发到任一座椅。",
+        ]
+        result["detected_domains"] = ["climate"]
+        result["scope"] = "vehicle"
+        result["confidence"] = max(float(result.get("confidence", 0)), 0.96)
+        return result
+
+    targets, target_source = _feedback_targets(text, context, result)
+    if not targets:
+        if not any(item.get("field") == "target" for item in result["ambiguities"]):
+            result["ambiguities"].append(
+                {"field": "target", "question": "该舒适反馈需要作用于主驾还是副驾？"}
+            )
+        return result
+
+    labels = {
+        "driver": "主驾",
+        "passenger": "副驾",
+        "ventilation": "通风",
+        "heating": "加热",
+    }
+    summaries: list[str] = []
+    if target_source == "history":
+        summaries.append(f"本轮省略了座位，沿用上一轮的{labels[targets[0]]}目标。")
+    elif target_source == "speaker_default":
+        summaries.append("当前为座椅单域驾驶员交互，将第一人称“我”解析为主驾。")
+
+    planned_actions: list[dict[str, Any]] = []
+    desired_level, intensity = _inferred_level(text)
+    airflow_is_cause = bool(AIRFLOW_CAUSE_PATTERN.search(text))
+    is_cold = bool(cold_cues)
+    is_hot = any(cue in {"热", "出汗"} for cue in ventilation_cues)
+    for target in targets:
+        ventilation = _context_device(context, target, "ventilation")
+        heating = _context_device(context, target, "heating")
+        cue_text = "、".join(cold_cues if is_cold else ventilation_cues)
+        summaries.append(f"检测到{labels[target]}“{cue_text}”舒适反馈，开始核对当前执行器状态。")
+
+        if cold_cues and ventilation_cues and not airflow_is_cause:
+            balanced_level = min(desired_level, 3)
+            planned_actions.extend(
+                [
+                    {
+                        "target": target,
+                        "function": "heating",
+                        "state": "on",
+                        "requested_level": balanced_level,
+                    },
+                    {
+                        "target": target,
+                        "function": "ventilation",
+                        "state": "on",
+                        "requested_level": balanced_level,
+                    },
+                ]
+            )
+            summaries[-1] = (
+                f"检测到{labels[target]}“{'、'.join(cold_cues + ventilation_cues)}”两个并存的舒适维度，"
+                "不将其判定为语义冲突。"
+            )
+            summaries.append(
+                f"冷感由{labels[target]}加热{balanced_level}档补偿，闷感由"
+                f"{labels[target]}通风{balanced_level}档缓解；生成可审计的组合目标。"
+            )
+            continue
+
+        if is_cold and (airflow_is_cause or ventilation.get("state") == "on"):
+            action, previous_level, next_level = _decrease_action(
+                target, "ventilation", ventilation
+            )
+            planned_actions.append(action)
+            if airflow_is_cause:
+                summaries.append(
+                    f"“{labels[target]}吹得冷”明确将{labels[target]}通风指定为致冷来源，不映射为加热请求。"
+                )
+            if previous_level > 1:
+                summaries.append(
+                    f"当前{labels[target]}通风为{previous_level}档，先降低到{next_level}档，不同时调整加热。"
+                )
+            elif previous_level == 1:
+                summaries.append(
+                    f"当前{labels[target]}通风为1档，冷感更可能由已开启通风造成；"
+                    "先关闭通风，不立即开启加热。"
+                )
+            else:
+                summaries.append(
+                    f"当前状态显示{labels[target]}通风已关闭，保持关闭；不会因“吹”反向调整加热。"
+                )
+            continue
+
+        if is_cold:
+            current_level = int(heating.get("level") or 0)
+            if heating.get("state") == "on" and current_level > 0:
+                next_level = min(max(current_level + 1, desired_level), 3)
+                planned_actions.append(
+                    {
+                        "target": target,
+                        "function": "heating",
+                        "state": "on",
+                        "requested_level": next_level,
+                    }
+                )
+                summaries.append(
+                    f"{labels[target]}通风已关闭，加热已开启{current_level}档但仍感到冷，"
+                    f"将加热提升到{next_level}档。"
+                )
+            else:
+                planned_actions.append(
+                    {
+                        "target": target,
+                        "function": "heating",
+                        "state": "on",
+                        "requested_level": desired_level,
+                    }
+                )
+                summaries.append(
+                    f"{labels[target]}通风已关闭且加热未开启，冷感仍存在；"
+                    f"按{'轻微' if intensity == 'mild' else '当前'}强度开启加热{desired_level}档。"
+                )
+            continue
+
+        if is_hot and heating.get("state") == "on":
+            action, previous_level, next_level = _decrease_action(target, "heating", heating)
+            planned_actions.append(action)
+            summaries.append(
+                f"当前{labels[target]}加热为{previous_level}档，热感优先视为已开启加热的结果；"
+                f"先{'关闭加热' if next_level == 0 else f'降低到{next_level}档'}，不立即开启通风。"
+            )
+            continue
+
+        current_level = int(ventilation.get("level") or 0)
+        if ventilation.get("state") == "on" and current_level > 0:
+            next_level = min(max(current_level + 1, desired_level), 3)
+            summaries.append(
+                f"{labels[target]}通风已开启{current_level}档但舒适问题仍存在，提升到{next_level}档。"
+            )
+        else:
+            next_level = desired_level
+            summaries.append(
+                f"{labels[target]}没有已开启的加热致因，按当前强度开启通风{next_level}档。"
+            )
+        planned_actions.append(
+            {
+                "target": target,
+                "function": "ventilation",
+                "state": "on",
+                "requested_level": next_level,
+            }
+        )
+
+    result["actions"] = planned_actions
+    result["ambiguities"] = [
+        item
+        for item in result["ambiguities"]
+        if item.get("field") not in {"capability", "state", "target"}
+    ]
+    result["decision_summary"] = summaries
+    result["confidence"] = max(float(result.get("confidence", 0)), 0.9)
+    return result
+
+
+def parse_local_intent(
+    utterance: str, context: dict[str, Any] | None = None
+) -> dict[str, Any]:
+    """Parse common Chinese seat commands without making hardware assumptions."""
+
+    text = re.sub(r"\s+", "", utterance.strip())
+    clauses = [part for part in re.split(r"[，,。；;！!？?]", text) if part]
+    actions: list[dict[str, Any]] = []
+    ambiguities: list[dict[str, str]] = []
+
+    for clause in clauses:
+        level = _level_for_clause(clause)
+        for feature_match in FEATURE_PATTERN.finditer(clause):
+            feature_zh = feature_match.group(0)
+            target_zh = _nearest_preceding_target(clause, feature_match.start())
+            state = _state_for_feature(clause, feature_match.end())
+            if target_zh is None:
+                ambiguities.append(
+                    {"field": "target", "question": f"“{feature_zh}”需要作用于主驾还是副驾？"}
+                )
+                continue
+            if state is None:
+                ambiguities.append(
+                    {"field": "state", "question": f"请确认{target_zh}{feature_zh}是打开还是关闭。"}
+                )
+                continue
+            for target in TARGETS[target_zh]:
+                actions.append(
+                    {
+                        "target": target,
+                        "function": FEATURES[feature_zh],
+                        "state": state,
+                        "requested_level": level,
+                    }
+                )
+
+    merged: dict[tuple[str, str], dict[str, Any]] = {}
+    for action in actions:
+        key = (action["target"], action["function"])
+        previous = merged.get(key)
+        if previous and previous["state"] != action["state"]:
+            ambiguities.append(
+                {
+                    "field": f"{key[0]}.{key[1]}.state",
+                    "question": "同一个座椅功能同时出现打开和关闭指令，请确认最终状态。",
+                }
+            )
+        merged[key] = action
+    actions = list(merged.values())
+
+    parsed = resolve_contextual_intent(
+        {
+            "actions": actions,
+            "ambiguities": ambiguities,
+            "decision_summary": [],
+            "confidence": 0.96 if actions and not ambiguities else 0.45,
+        },
+        utterance,
+        context,
+    )
+    actions = parsed["actions"]
+    ambiguities = parsed["ambiguities"]
+
+    if not actions and not ambiguities:
+        ambiguities.append(
+            {
+                "field": "capability",
+                "question": "当前仅支持主/副驾座椅通风或加热的打开与关闭。",
+            }
+        )
+
+    summary = ["识别为座椅热舒适单域请求。", *parsed["decision_summary"]]
+    labels = {
+        "driver": "主驾",
+        "passenger": "副驾",
+        "ventilation": "通风",
+        "heating": "加热",
+        "on": "打开",
+        "off": "关闭",
+    }
+    summary.extend(
+        f"目标：{labels[action['target']]}座椅{labels[action['function']]}{labels[action['state']]}。"
+        for action in actions
+    )
+    if any(action["requested_level"] is not None for action in actions):
+        summary.append("目标档位已写入功能配置；现有 CAN 矩阵仅支持档位虚拟回读。")
+
+    response = {
+        "actions": actions,
+        "ambiguities": ambiguities,
+        "decision_summary": summary,
+        "confidence": max(parsed["confidence"], 0.96) if actions and not ambiguities else 0.45,
+    }
+    for field in ("detected_domains", "scope"):
+        if field in parsed:
+            response[field] = parsed[field]
+    return response
@@ -0,0 +1,115 @@
+"""Load and encode the normalized partial seat CAN matrix."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from collections.abc import Iterable
+from typing import Any
+
+import yaml
+
+
+class MatrixError(ValueError):
+    """Raised when a command cannot be represented by the normalized matrix."""
+
+
+class SeatCanMatrix:
+    def __init__(self, matrix_path: Path | Iterable[Path]):
+        paths = [matrix_path] if isinstance(matrix_path, Path) else list(matrix_path)
+        if not paths:
+            raise MatrixError("At least one matrix document is required")
+        self.path = paths[0]
+        self.paths = paths
+        documents = [yaml.safe_load(path.read_text(encoding="utf-8")) for path in paths]
+        if len(documents) == 1:
+            self.data = documents[0]
+        else:
+            self.data = {
+                "version": "+".join(str(item.get("version", "unknown")) for item in documents),
+                "source": {"files": [str(path) for path in paths]},
+                "scope": "combined_virtual_seat_capabilities",
+                "hardware_execution_allowed": all(
+                    item.get("hardware_execution_allowed") is True for item in documents
+                ),
+                "limitations": [
+                    limitation
+                    for document in documents
+                    for limitation in document.get("limitations", [])
+                ],
+                "messages": [
+                    message
+                    for document in documents
+                    for message in document.get("messages", [])
+                ],
+            }
+        self.messages = {}
+        for item in self.data["messages"]:
+            if item["name"] in self.messages:
+                raise MatrixError(f"Duplicate matrix message: {item['name']}")
+            self.messages[item["name"]] = item
+
+    @property
+    def hardware_execution_allowed(self) -> bool:
+        return self.data.get("hardware_execution_allowed") is True
+
+    def message(self, name: str) -> dict[str, Any]:
+        try:
+            return self.messages[name]
+        except KeyError as exc:
+            raise MatrixError(f"Unknown matrix message: {name}") from exc
+
+    def signal(self, message_name: str, signal_name: str) -> dict[str, Any]:
+        message = self.message(message_name)
+        for signal in message["signals"]:
+            if signal["name"] == signal_name:
+                return signal
+        raise MatrixError(f"Unknown signal {signal_name} in {message_name}")
+
+    def encode(self, message_name: str, values: dict[str, int]) -> dict[str, Any]:
+        message = self.message(message_name)
+        raw_value = 0
+        for signal_name, value in values.items():
+            signal = self.signal(message_name, signal_name)
+            width = int(signal["bit_length"])
+            if not isinstance(value, int) or value < 0 or value >= 1 << width:
+                raise MatrixError(f"Value {value!r} does not fit {signal_name}[{width}]")
+            start_bit = int(signal["start_bit"])
+            mask = ((1 << width) - 1) << start_bit
+            raw_value = (raw_value & ~mask) | (value << start_bit)
+
+        payload_size = int(message["minimum_payload_bytes"])
+        payload = raw_value.to_bytes(payload_size, byteorder="little", signed=False)
+        return {
+            "message": message_name,
+            "frame_id": f"0x{int(message['frame_id']):03X}",
+            "payload_size": payload_size,
+            "data_hex": " ".join(f"{byte:02X}" for byte in payload),
+            "signals": values,
+            "encoding": "lsb0/intel inferred for virtual execution",
+        }
+
+    def decode(self, message_name: str, payload: bytes) -> dict[str, int]:
+        """Decode every declared little-endian lsb0 signal from a CAN payload."""
+        message = self.message(message_name)
+        minimum = int(message["minimum_payload_bytes"])
+        if len(payload) < minimum:
+            raise MatrixError(
+                f"Payload for {message_name} must contain at least {minimum} bytes"
+            )
+        raw_value = int.from_bytes(payload[:minimum], byteorder="little", signed=False)
+        return {
+            signal["name"]: (
+                raw_value >> int(signal["start_bit"])
+            ) & ((1 << int(signal["bit_length"])) - 1)
+            for signal in message["signals"]
+        }
+
+    def decode_by_frame_id(self, frame_id: int, payload: bytes) -> dict[str, int]:
+        matches = [
+            item for item in self.data["messages"] if int(item["frame_id"]) == frame_id
+        ]
+        if len(matches) != 1:
+            raise MatrixError(
+                f"Expected exactly one matrix message for frame 0x{frame_id:03X}"
+            )
+        return self.decode(matches[0]["name"], payload)
@@ -0,0 +1,71 @@
+"""Read the supplied XLSX matrix using the standard OOXML structures."""
+
+from __future__ import annotations
+
+import re
+import zipfile
+from pathlib import Path
+from typing import Any
+from xml.etree import ElementTree
+
+
+MAIN_NS = "http://schemas.openxmlformats.org/spreadsheetml/2006/main"
+CELL_REFERENCE = re.compile(r"([A-Z]+)([0-9]+)")
+
+
+def _column_index(reference: str) -> int:
+    match = CELL_REFERENCE.fullmatch(reference)
+    if not match:
+        raise ValueError(f"Invalid XLSX cell reference: {reference}")
+    result = 0
+    for character in match.group(1):
+        result = result * 26 + ord(character) - ord("A") + 1
+    return result - 1
+
+
+def _shared_strings(archive: zipfile.ZipFile) -> list[str]:
+    root = ElementTree.fromstring(archive.read("xl/sharedStrings.xml"))
+    return [
+        "".join(node.text or "" for node in item.iter(f"{{{MAIN_NS}}}t"))
+        for item in root.findall(f"{{{MAIN_NS}}}si")
+    ]
+
+
+def read_source_signals(path: Path) -> list[dict[str, Any]]:
+    """Return the 23 signal rows needed for source-to-normalized contract checks."""
+
+    with zipfile.ZipFile(path) as archive:
+        strings = _shared_strings(archive)
+        worksheet = ElementTree.fromstring(archive.read("xl/worksheets/sheet1.xml"))
+
+    rows: list[list[str]] = []
+    for row_node in worksheet.iter(f"{{{MAIN_NS}}}row"):
+        values: dict[int, str] = {}
+        for cell in row_node.findall(f"{{{MAIN_NS}}}c"):
+            value_node = cell.find(f"{{{MAIN_NS}}}v")
+            if value_node is None or value_node.text is None:
+                continue
+            value = value_node.text
+            if cell.get("t") == "s":
+                value = strings[int(value)]
+            values[_column_index(str(cell.get("r")))] = value
+        if values:
+            width = max(values) + 1
+            rows.append([values.get(index, "") for index in range(width)])
+
+    records = []
+    for row in rows[1:]:
+        if len(row) < 6 or not row[0]:
+            continue
+        records.append(
+            {
+                "name": row[0],
+                "description": row[1],
+                "frame_id": int(row[2], 16),
+                "sub_id": None if row[3] in {"", "null"} else int(row[3], 16),
+                "start_bit": int(float(row[4])),
+                "bit_length": int(float(row[5])),
+            }
+        )
+    return records
+
@@ -0,0 +1,256 @@
+"""Seat-specific semantic validation beyond JSON Schema."""
+
+from __future__ import annotations
+
+from typing import Any
+
+
+SUPPORTED_TARGETS = {"driver", "passenger"}
+SUPPORTED_FUNCTIONS = {"ventilation", "heating"}
+SUPPORTED_STATES = {"on", "off"}
+SUPPORTED_MOTION_AXES = {
+    "longitudinal",
+    "cushion_tilt",
+    "backrest_recline",
+    "legrest_height",
+    "legrest_extension",
+}
+ANGLE_RANGES = {
+    "backrest_recline": (95, 140),
+    "legrest_height": (95, 160),
+}
+
+
+def validate_semantics(
+    parsed: dict[str, Any], function_config: dict[str, Any] | None
+) -> list[dict[str, str]]:
+    issues: list[dict[str, str]] = []
+
+    detected_domains = set(parsed.get("detected_domains", ["seat"]))
+    if parsed.get("actions") and "seat" not in detected_domains:
+        issues.append(
+            {
+                "severity": "error",
+                "code": "DOMAIN_ACTION_MISMATCH",
+                "path": "detected_domains",
+                "message": "GPT 生成了座椅动作，但 detected_domains 中没有 seat。",
+            }
+        )
+
+    for index, deferred in enumerate(parsed.get("deferred_requests", [])):
+        domain = str(deferred.get("domain"))
+        capability = str(deferred.get("capability"))
+        if domain not in detected_domains or not capability.startswith(f"{domain}."):
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "DEFERRED_REQUEST_MISMATCH",
+                    "path": f"deferred_requests[{index}]",
+                    "message": "暂缓请求的领域、能力名和 detected_domains 不一致。",
+                }
+            )
+        issues.append(
+            {
+                "severity": "warning",
+                "code": "UNAVAILABLE_DOMAIN",
+                "path": f"deferred_requests[{index}]",
+                "message": f"已理解 {capability} 目标，但 {domain} 域当前仅有能力目录，未接入执行器。",
+            }
+        )
+
+    for ambiguity in parsed.get("ambiguities", []):
+        issues.append(
+            {
+                "severity": "error",
+                "code": "AMBIGUOUS_INTENT",
+                "path": ambiguity["field"],
+                "message": ambiguity["question"],
+            }
+        )
+
+    seen: dict[tuple[str, str], str] = {}
+    active_functions: dict[str, set[str]] = {}
+    has_motion = False
+    has_driver_motion = False
+    for index, action in enumerate(parsed.get("actions", [])):
+        path = f"actions[{index}]"
+        if action.get("target") not in SUPPORTED_TARGETS:
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "UNSUPPORTED_TARGET",
+                    "path": f"{path}.target",
+                    "message": "当前只支持主驾和前排副驾。",
+                }
+            )
+        if action.get("action_type") == "position":
+            has_motion = True
+            has_driver_motion = has_driver_motion or action.get("target") == "driver"
+            axis = action.get("axis")
+            position = action.get("position_pct")
+            if axis not in SUPPORTED_MOTION_AXES:
+                issues.append(
+                    {
+                        "severity": "error",
+                        "code": "UNSUPPORTED_MOTION_AXIS",
+                        "path": f"{path}.axis",
+                        "message": "请求的座椅位置轴未在当前能力目录中注册。",
+                    }
+                )
+            if (
+                not isinstance(position, int)
+                or isinstance(position, bool)
+                or not 0 <= position <= 100
+            ):
+                issues.append(
+                    {
+                        "severity": "error",
+                        "code": "UNSUPPORTED_POSITION_RANGE",
+                        "path": f"{path}.position_pct",
+                        "message": "座椅电机绝对目标必须是 0-100 的整数百分比。",
+                    }
+                )
+            angle = action.get("angle_deg")
+            if angle is not None:
+                angle_range = ANGLE_RANGES.get(str(axis))
+                if (
+                    angle_range is None
+                    or not isinstance(angle, (int, float))
+                    or isinstance(angle, bool)
+                    or not angle_range[0] <= float(angle) <= angle_range[1]
+                ):
+                    issues.append(
+                        {
+                            "severity": "error",
+                            "code": "UNSUPPORTED_ANGLE_RANGE",
+                            "path": f"{path}.angle_deg",
+                            "message": "座椅夹角超出当前虚拟标定范围。",
+                        }
+                    )
+            key = (str(action.get("target")), str(axis))
+            previous_position = seen.get(key)
+            if previous_position is not None and previous_position != str(position):
+                issues.append(
+                    {
+                        "severity": "error",
+                        "code": "CONFLICTING_OPERATIONS",
+                        "path": path,
+                        "message": "同一个座椅电机轴不能在一份配置中包含不同目标位置。",
+                    }
+                )
+            seen[key] = str(position)
+            continue
+        if action.get("function") not in SUPPORTED_FUNCTIONS:
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "UNSUPPORTED_CAPABILITY",
+                    "path": f"{path}.function",
+                    "message": "当前只支持座椅通风和加热。",
+                }
+            )
+        if action.get("state") not in SUPPORTED_STATES:
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "UNSUPPORTED_STATE",
+                    "path": f"{path}.state",
+                    "message": "当前控制信号只支持打开或关闭。",
+                }
+            )
+        requested_level = action.get("requested_level")
+        if requested_level is not None and requested_level not in {1, 2, 3}:
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "UNSUPPORTED_LEVEL_RANGE",
+                    "path": f"{path}.requested_level",
+                    "message": "当前舒适推断和 0x4FA 虚拟回读仅支持 1-3 档。",
+                }
+            )
+        elif requested_level is not None:
+            issues.append(
+                {
+                    "severity": "warning",
+                    "code": "LEVEL_COMMAND_SIMULATED",
+                    "path": f"{path}.requested_level",
+                    "message": (
+                        "目标档位已写入 FunctionConfig 和会话状态；"
+                        "0x42A 仅编译开关信号，档位回读仅在 dry-run 中模拟。"
+                    ),
+                }
+            )
+        key = (str(action.get("target")), str(action.get("function")))
+        previous_state = seen.get(key)
+        if previous_state is not None and previous_state != action.get("state"):
+            issues.append(
+                {
+                    "severity": "error",
+                    "code": "CONFLICTING_OPERATIONS",
+                    "path": path,
+                    "message": "同一个座椅功能不能在一份配置中同时打开和关闭。",
+                }
+            )
+        seen[key] = str(action.get("state"))
+        if action.get("state") == "on":
+            active_functions.setdefault(str(action.get("target")), set()).add(
+                str(action.get("function"))
+            )
+
+    if has_motion:
+        issues.append(
+            {
+                "severity": "warning",
+                "code": "SYNTHETIC_MOTION_CAN",
+                "path": "platform.reference_vehicle.seat_motion",
+                "message": "座椅位置控制和反馈 CAN 矩阵完全为演示虚构，仅用于验证模型与配置链路。",
+            }
+        )
+    if has_driver_motion:
+        issues.append(
+            {
+                "severity": "warning",
+                "code": "DRIVER_MOTION_INTERLOCK_NOT_IMPLEMENTED",
+                "path": "platform.reference_vehicle.vehicle_state",
+                "message": "未接入车速、挡位、乘员和防夹状态，主驾移动只能虚拟执行。",
+            }
+        )
+
+    for target, functions in active_functions.items():
+        if functions == {"ventilation", "heating"}:
+            issues.append(
+                {
+                    "severity": "warning",
+                    "code": "COMBINED_THERMAL_SIMULATED",
+                    "path": f"actions.{target}",
+                    "message": (
+                        "同一座位的加热与通风组合目标仅用于 dry-run；"
+                        "实车前必须确认 ECU 是否允许两功能同时开启。"
+                    ),
+                }
+            )
+
+    if (
+        function_config is None
+        and not parsed.get("deferred_requests")
+        and not parsed.get("ambiguities")
+        and not parsed.get("preset_actions")
+    ):
+        issues.append(
+            {
+                "severity": "error",
+                "code": "NO_EXECUTABLE_OPERATION",
+                "path": "spec.operations",
+                "message": "没有生成可执行的座椅操作。",
+            }
+        )
+
+    issues.append(
+        {
+            "severity": "warning",
+            "code": "VIRTUAL_EXECUTION_ONLY",
+            "path": "platform.reference_vehicle",
+            "message": "CAN 元数据不完整，当前仅允许虚拟执行，禁止实车下发。",
+        }
+    )
+    return issues
@@ -0,0 +1,4 @@
+# Seat Tests
+
+`unit/` 覆盖姿态参数和运动学转换，`contract/` 覆盖配置与插件契约，`integration/` 使用模拟 CAN Gateway 覆盖命令、回读、互锁和超时。实车范围需与平台安全标定分别测试，不能只依赖 JSON Schema 的宽范围。
+
@@ -0,0 +1 @@
+
@@ -0,0 +1 @@
+
--- a/Show More
+++ b/Show More