> ## Documentation Index > Fetch the complete documentation index at: https://docs.corti.ai/llms.txt > Use this file to discover all available pages before exploring further. # List feedback for a task > Returns all feedback resources submitted for the task by the authenticated user, newest-first. The task must exist, belong to the supplied context, and belong to the authenticated customer. Feedback is scoped to the calling user via row-level security, so the response contains only that user's feedback. ## OpenAPI ````yaml /agentic/auto-generated-openapi-v2.yml get /v2/agentic/contexts/{contextId}/tasks/{taskId}/feedback openapi: 3.1.0 info: title: Corti Agent API v2 version: 2.0.0 summary: Manage agents and converse with them over the A2A v1.0 protocol. description: > Version 2 of the Corti Agents REST API. ## What's new in v2 - **Unified `connectors`** — a single, flat, discriminated array replaces the v1 split between `experts`, `mcpServers`, and sub-agents. The envelope is extensible: new connector kinds (`a2a`, `openapi`, `custom`) slot in behind the `type` discriminator without breaking changes. - **Clean CRUD verbs** — `POST` creates, `GET` fetches/lists, `PATCH` performs a *true* partial update, `DELETE` removes. v1's PATCH was effectively a PUT. - **First-class metadata** — `visibility`, `model`, `lifecycle` (now a body field, not a query param), and free-form `labels`. ## A2A v1.0 only The conversational surface speaks **A2A protocol version `1.0` exclusively** (both the `JSONRPC` and `HTTP+JSON` bindings). The deprecated v0.3 binding from v1 is intentionally not carried forward. Per A2A §3.6, clients MUST send the `A2A-Version` header (`Major.Minor`, e.g. `1.0`) on every request to the `/a2a/*` surface, or supply it as the `A2A-Version` query parameter. This surface implements `1.0` only; an absent header is treated as `1.0`. Patch versions MUST NOT be sent. The server echoes the negotiated version in the `A2A-Version` response header. ## Partial updates (PATCH) `PATCH` uses **JSON Merge Patch** semantics (RFC 7386), served under the `application/merge-patch+json` media type: - **Omit a field** → left unchanged. - **`null`** → cleared / reset to its default. - **Arrays** (e.g. `connectors`) → *replaced wholesale*, never merged. To mutate a single connector without replacing the whole array, use the `/connectors` sub-resource endpoints. ## Streaming (Server-Sent Events) The streaming A2A surfaces — `message/stream`, `tasks/subscribe`, and the streaming JSON-RPC methods — respond with `text/event-stream`. The `text/event-stream` media type carries a sequence of Server-Sent Events; the attached schema describes a single event frame. Every event's `data` field is a string carrying a JSON document (`contentMediaType: application/json`); `contentSchema` declares the parsed payload (`StreamResponse` for the `HTTP+JSON` binding, `JSONRPCResponse` for the `JSONRPC` binding). The optional `id` field carries the SSE event id used for resumption via the `Last-Event-ID` request header. contact: name: Corti API Support url: https://corti.ai email: support@corti.ai license: name: Proprietary url: https://corti.ai/terms termsOfService: https://corti.ai/terms servers: - url: https://api.{env}.corti.app description: Corti regional API gateway variables: env: default: eu enum: - eu - us description: Deployment region. security: - bearerAuth: [] tenantHeader: [] tags: - name: Agents description: Create, read, update, and delete agents. - name: Agent Card description: A2A-compliant agent discovery cards. - name: A2A description: Converse with an agent over the A2A v1.0 protocol. - name: Connectors description: Manage the connectors attached to an agent. - name: Usage description: Bucketed invocation history for an agent. - name: Contexts description: Conversational contexts and their tasks. - name: Registry description: Discoverable, pre-built connectors offered by the platform. - name: Artifacts description: Retrieve artifacts produced by tasks. - name: Feedback description: Collect human or automated feedback on tasks and messages. paths: /v2/agentic/contexts/{contextId}/tasks/{taskId}/feedback: parameters: - $ref: '#/components/parameters/ContextId' - $ref: '#/components/parameters/TaskId' get: tags: - Feedback summary: List feedback for a task description: > Returns all feedback resources submitted for the task by the authenticated user, newest-first. The task must exist, belong to the supplied context, and belong to the authenticated customer. Feedback is scoped to the calling user via row-level security, so the response contains only that user's feedback. operationId: FeedbackList responses: '200': description: A list of feedback resources. content: application/json: schema: $ref: '#/components/schemas/FeedbackListResponse' example: feedbacks: - id: fb.0192f4c8-9a3b-7e2f-9c4d-5a6b7c8d9e0f taskId: task.0192f4c8-4e7c-7d50-b13c-5eaf8a4d9c62 rating: scale: binary value: 1 normalizedScore: 1 labels: [] createdAt: '2026-05-19T12:00:00Z' '401': $ref: '#/components/responses/Unauthorized' '404': $ref: '#/components/responses/NotFound' components: parameters: ContextId: name: contextId in: path required: true description: Context identifier (prefixed UUIDv7). schema: $ref: '#/components/schemas/CommonContextIDValue' TaskId: name: taskId in: path required: true description: Task identifier (prefixed UUIDv7). schema: $ref: '#/components/schemas/CommonTaskIDValue' schemas: FeedbackListResponse: type: object description: >- All feedback resources for a task, newest-first. Feedback is scoped to the authenticated user via row-level security. required: - feedbacks properties: feedbacks: type: array items: $ref: '#/components/schemas/FeedbackResponse' description: Feedback resources for the task. CommonContextIDValue: type: string format: context-id description: >- Context identifier. Accepts `ctx.` or a bare UUIDv7 on input; always returned prefixed. pattern: ^(ctx\.)?[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$ examples: - ctx.0192f4c8-3d6b-7c4f-a02b-4d9e7f3c8b51 CommonTaskIDValue: type: string format: task-id description: >- Task identifier. Accepts `task.` or a bare UUIDv7 on input; always returned prefixed. pattern: ^(task\.)?[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$ examples: - task.0192f4c8-4e7c-7d50-b13c-5eaf8a4d9c62 FeedbackResponse: type: object description: A stored feedback resource. required: - id - taskId - rating - normalizedScore - labels - createdAt properties: id: $ref: '#/components/schemas/FeedbackIDValue' taskId: $ref: '#/components/schemas/CommonTaskIDValue' rating: $ref: '#/components/schemas/FeedbackRating' normalizedScore: type: number format: double minimum: 0 maximum: 1 description: | Corti-derived internal score between 0 and 1. The original scale and value are always retained alongside this score. example: 1 labels: type: array items: $ref: '#/components/schemas/FeedbackLabel' description: Structured observations about the result. examples: - - correct - helpful reason: type: string maxLength: 1000 description: Free-text explanation of the rating or labels. examples: - The response stated the patient had diabetes - but this was not present in the available data. target: $ref: '#/components/schemas/FeedbackTarget' metadata: $ref: '#/components/schemas/FeedbackMetadata' createdAt: type: string format: date-time readOnly: true description: When the feedback was created. example: '2026-05-19T12:00:00Z' CommonErrorResponse: type: object description: > Corti management-plane error envelope, used by all non-A2A endpoints. - **Standard** — when the error chain contains at least one `PublicError`, `code` and `message` come from the outermost `PublicError` and `details` is merged across the whole chain (outer values take precedence). - **Fallback** — when the chain contains no `PublicError`, the response is a generic `500` carrying a `requestId` for support reference. - **Validation** — a single `PublicError` whose `details.validationErrors` lists the offending fields. Field names use camelCase on the wire (e.g. `requestId`, `howToFix`). The free-form `details` object may carry arbitrary caller-defined keys. Rate limiting (HTTP 429) is not yet implemented; the server does not emit a 429 response. required: - error properties: error: type: object required: - code - message properties: code: type: string description: Stable, machine-readable, SCREAMING_SNAKE_CASE error code. examples: - ASSIGNMENT_CONFLICT - VALIDATION_FAILED - INTERNAL_ERROR message: type: string description: Human-readable explanation. howToFix: type: string description: Optional guidance for the caller to resolve the error. details: type: object description: | Structured context, merged from every `PublicError` in the chain (outer values win). Omitted on the generic fallback response. additionalProperties: true properties: validationErrors: type: array description: Present when `code` is `VALIDATION_FAILED`. items: type: object required: - field - reason properties: field: type: string description: The field that failed validation. reason: type: string description: Why the field failed validation. requestId: type: string description: > Correlation ID from request middleware. Included only on the generic `500` fallback so consumers can quote it in support requests. description: The error object with code, message, and optional details. examples: - error: code: ASSIGNMENT_CONFLICT message: could not complete assignment details: assignment_id: asg_99 expert_id: exp_42 - error: code: VALIDATION_FAILED message: validation failed details: validationErrors: - field: email reason: invalid format - field: name reason: required - error: code: INTERNAL_ERROR message: internal server error requestId: req_abc123 FeedbackIDValue: type: string format: feedback-id description: >- Feedback identifier. Accepts `fb.` or a bare UUIDv7 on input; always returned prefixed. pattern: ^(fb\.)?[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$ examples: - fb.0192f4c8-7e2a-7b3c-9d4e-5f6a7b8c9d01 FeedbackRating: type: object description: The original rating supplied by the customer. required: - scale - value properties: scale: $ref: '#/components/schemas/FeedbackRatingScale' value: type: number format: double description: The rating value on the selected scale. additionalProperties: false examples: - scale: binary value: 1 FeedbackLabel: type: string description: > Structured observation about the result. Positive and negative labels share one taxonomy so customers can represent mixed feedback. - `correct` — factually and contextually correct (positive). - `complete` — included the important expected information (positive). - `helpful` — materially helped the user complete their task (positive). - `wellPresented` — clear, readable, appropriately structured (positive). - `efficient` — reached a useful result without unnecessary content (positive). - `incorrect` — one or more claims, conclusions, or actions were wrong (negative). - `missingInformation` — important or expected information was omitted (negative). - `irrelevant` — included content that did not address the request (negative). - `misunderstoodRequest` — the system responded to the wrong intent (negative). - `unsupportedClaim` — a claim not supported by available information (negative). - `unsafeOrInappropriate` — unsafe, disallowed, or unsuitable (negative). - `poorlyPresented` — difficult to read or unsuitably structured (negative). - `tooVerbose` — substantially more detail than useful (negative). - `other` — another observation described in `reason` (both). enum: - correct - complete - helpful - wellPresented - efficient - incorrect - missingInformation - irrelevant - misunderstoodRequest - unsupportedClaim - unsafeOrInappropriate - poorlyPresented - tooVerbose - other FeedbackTarget: type: object description: | Identifies the specific user-visible response being evaluated. If omitted, the feedback applies to the task as a whole. required: - messageId properties: messageId: $ref: '#/components/schemas/CommonMessageIDValue' additionalProperties: false FeedbackMetadata: type: object description: Customer-provided provenance and correlation information. required: [] properties: collectionMethod: type: string maxLength: 100 description: >- How the customer collected the feedback. Informational only; does not affect rating validation or normalization. examples: - thumbs - survey clientReference: type: string maxLength: 255 description: | Customer-defined reference to correlate the feedback with an object in the customer's own system. Not unique and does not provide idempotency. Should not contain sensitive information. actor: $ref: '#/components/schemas/FeedbackActor' additionalProperties: false FeedbackRatingScale: type: string description: > The scale on which the rating was collected. - `binary` — 0 (negative) or 1 (positive). Planned (not yet available): `likert5` (integer 1–5), `continuous01` (number 0–1). enum: - binary CommonMessageIDValue: type: string format: message-id description: >- Message identifier. Accepts `msg.` or a bare UUIDv7 on input; always returned prefixed. pattern: ^(msg\.)?[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$ examples: - msg.0192f4c8-5f8d-7e61-924d-6fb09b5ead73 FeedbackActor: type: object description: Customer-defined opaque identifier for the feedback submitter. required: - externalId properties: externalId: type: string maxLength: 255 description: | Scoped to the authenticated customer; not globally unique and not independently verified. Should preferably be pseudonymous and must not contain names, emails, national identifiers, or medical record numbers. additionalProperties: false responses: Unauthorized: description: Missing or invalid credentials. content: application/json: schema: $ref: '#/components/schemas/CommonErrorResponse' NotFound: description: The resource does not exist or is not visible to the caller. content: application/json: schema: $ref: '#/components/schemas/CommonErrorResponse' securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: OAuth 2.0 / OIDC bearer token. tenantHeader: type: apiKey in: header name: Tenant-Name description: The tenant the request operates within. ````