> ## Documentation Index > Fetch the complete documentation index at: https://docs.co-mind.ai/llms.txt > Use this file to discover all available pages before exploring further. # Create response > OpenAI-Responses-compatible endpoint. The request body is a passthrough to the Inference Engine — every field the caller sends flows through, so any Responses API field the backend model accepts (input content parts, instructions, `response_format`, tools, streaming, temperature, `max_output_tokens`, …) can be used. Tenant model-policy applies: models not on the caller's tenant allowlist return `409` with error code `model_blocked`. Tenant-owned provider keys (`OpenAI`, `Anthropic`, `Gemini`, `Groq`, `SambaNova`, `Cohere`) are auto-resolved and injected on the outbound call per the standard resolution order (user → sub-org → org → system); explicit `X-User-*-Key` request headers still take precedence. ## OpenAPI ````yaml /openapi.yaml post /v1/responses openapi: 3.1.0 info: title: Co-mind.ai Private AI Platform API version: 1.2.4 description: > Co-mind.ai Private AI Platform API. ## Features - OpenAI-compatible endpoints - Tool/function calling support - Streaming responses - Multiple backend support - Personal Access Tokens (PAT) for programmatic access ## Authentication The API supports two authentication methods: **1. JWT Authentication** — Login with email/password to get short-lived access tokens. Use for interactive sessions (web apps, Postman). **2. Personal Access Tokens (PAT)** — Long-lived tokens for programmatic/API access. Create via `POST /v1/api-tokens` after authenticating with JWT. PAT format: `cmnd_.` Both methods use the `Authorization: Bearer ` header. contact: email: support@co-mind.ai servers: - url: http://co-mind-platform-host description: Co-mind.ai AI Platform security: - BearerAuth: [] tags: - name: Discovery description: List available models - name: Authentication description: >- JWT login, refresh, logout, SSO, registration, password reset, and `/v1/auth/me` - name: API Tokens description: Personal Access Token (PAT) management - name: Completions description: Text completion endpoints (single-shot, no conversation state) - name: Chat description: > OpenAI-compatible chat completions. **Stateless** — the client sends the full conversation history on every request; the server does not persist messages. - name: Responses description: > OpenAI-Responses-compatible endpoint. Wire-shape passthrough to the Inference Engine — every field the caller sends flows through, so the caller can use the full Responses API surface. The server does not persist responses server-side; use the `/v1/responses/{resp_id}/continue` variant to extend a prior response by id. - name: Embeddings description: > Text-embedding endpoint. Runs against the platform's configured embedding model (see `GET /v1/models`); the resulting vectors are returned to the caller and are not persisted server-side. - name: Translation description: > Text translation between languages. Ships with `POST /v1/translate/text`; async document translation is on the roadmap for a future release. - name: Knowledge Base description: > Knowledge base management (create / update / delete KBs, upload files, query for context) and **stateless** retrieval-augmented chat (`/v1/knowledgebase/chat/completions`). The RAG chat endpoint is stateless — the client sends the full conversation history on every request; the server injects retrieved KB context and returns the completion. - name: Chat Sessions (Preview) description: > **Preview — not currently enabled.** **Stateful** chat sessions. The server manages conversation history and supports tool orchestration (knowledge base retrieval, research, agents). These endpoints are documented for integrator preparation but are not exposed on any deployment today; they will be enabled in an upcoming release. Request and response schemas may change before general availability. - name: Researcher description: Web search, research sessions, analysis, synthesis, and source credibility - name: Quota description: Usage quota tracking paths: /v1/responses: post: tags: - Responses summary: Create response description: > OpenAI-Responses-compatible endpoint. The request body is a passthrough to the Inference Engine — every field the caller sends flows through, so any Responses API field the backend model accepts (input content parts, instructions, `response_format`, tools, streaming, temperature, `max_output_tokens`, …) can be used. Tenant model-policy applies: models not on the caller's tenant allowlist return `409` with error code `model_blocked`. Tenant-owned provider keys (`OpenAI`, `Anthropic`, `Gemini`, `Groq`, `SambaNova`, `Cohere`) are auto-resolved and injected on the outbound call per the standard resolution order (user → sub-org → org → system); explicit `X-User-*-Key` request headers still take precedence. operationId: createResponse requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/ResponsesRequest' responses: '200': description: Successful response content: application/json: schema: $ref: '#/components/schemas/ResponsesResponse' '400': $ref: '#/components/responses/BadRequest' '401': $ref: '#/components/responses/Unauthorized' '409': description: Model refused by the caller's tenant model-policy allowlist content: application/json: schema: type: object properties: error: type: object properties: message: type: string type: type: string example: policy_error code: type: string example: model_blocked '429': $ref: '#/components/responses/RateLimitExceeded' components: schemas: ResponsesRequest: type: object required: - input - model additionalProperties: true properties: model: type: string example: gpt-4o-mini input: description: > The prompt or conversation input. Accepts either a plain string or an array of typed content parts (text, input_image, input_file, …) per the Responses API shape supported by the backend model. oneOf: - type: string - type: array items: type: object additionalProperties: true instructions: type: string description: Optional system-style instructions applied to the response. max_output_tokens: type: integer minimum: 1 temperature: type: number minimum: 0 maximum: 2 top_p: type: number minimum: 0 maximum: 1 stream: type: boolean default: false previous_response_id: type: string description: >- Id of a prior response to continue. Callers can also POST to `/v1/responses/{resp_id}/continue`. ResponsesResponse: type: object additionalProperties: true properties: id: type: string example: resp_abc123 object: type: string example: response created: type: integer model: type: string output: description: > Model output. Shape depends on the backend and on whether `stream` was requested; for non-streaming calls this is typically an array of content parts (or a compact `output_text` string alongside them). oneOf: - type: string - type: array items: type: object additionalProperties: true usage: $ref: '#/components/schemas/Usage' Usage: type: object properties: prompt_tokens: type: integer completion_tokens: type: integer total_tokens: type: integer Error: type: object properties: error: type: object properties: message: type: string type: type: string code: type: string responses: BadRequest: description: Bad request content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: Invalid request parameters type: invalid_request_error code: bad_request Unauthorized: description: Authentication required or invalid token content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: Invalid authentication credentials type: permission_error code: unauthorized RateLimitExceeded: description: Rate limit exceeded headers: Retry-After: schema: type: integer description: Seconds to wait before retrying x-ratelimit-limit-requests: schema: type: integer x-ratelimit-remaining-requests: schema: type: integer x-ratelimit-reset-requests: schema: type: integer content: application/json: schema: $ref: '#/components/schemas/Error' example: error: message: Rate limit exceeded type: rate_limit_exceeded code: rate_limit_exceeded securitySchemes: BearerAuth: type: http scheme: bearer bearerFormat: JWT or PAT description: > Bearer token authentication. Supports two token types: - **JWT Access Token** — obtained via `POST /v1/auth/login` - **Personal Access Token (PAT)** — created via `POST /v1/api-tokens`, format: `cmnd_.` ````