Keywords AI LLM Proxy (Chat Completions)

openapi: 3.0.1
info:
  title: Keywords AI API
  description: >-
    Keywords AI is an LLM observability and gateway platform. This specification
    documents the publicly documented REST surface served from
    https://api.keywordsai.co/api: the OpenAI-compatible chat completions proxy,
    the asynchronous request-logging endpoint, prompt and prompt-version
    management, threads, traces, evaluators, users (customers), datasets, and
    experiments. All endpoints authenticate with a Bearer API key. (Keywords AI
    is rebranding to Respan; the keywordsai.co host and API remain active.)
  termsOfService: https://www.keywordsai.co/terms-of-service
  contact:
    name: Keywords AI Support
    url: https://www.keywordsai.co
    email: team@keywordsai.co
  version: '1.0'
servers:
  - url: https://api.keywordsai.co/api
    description: Keywords AI production API base.
security:
  - bearerAuth: []
tags:
  - name: Gateway
    description: OpenAI-compatible LLM proxy (chat completions).
  - name: Logging
    description: Asynchronous request / span logging.
  - name: Prompts
    description: Prompt and prompt-version management.
  - name: Threads
    description: Multi-turn conversation threads.
  - name: Traces
    description: OpenTelemetry-aligned distributed traces.
  - name: Evaluators
    description: Output-scoring evaluators.
  - name: Users
    description: End-user (customer) analytics.
  - name: Datasets
    description: Datasets and dataset logs.
  - name: Experiments
    description: Prompt / model experiments.
paths:
  /chat/completions:
    post:
      operationId: createChatCompletion
      tags:
        - Gateway
      summary: Create a chat completion
      description: >-
        OpenAI-compatible chat completions through the Keywords AI gateway.
        Routes to 250+ models behind a single endpoint, with optional streaming
        (Server-Sent Events when `stream` is true), tool calling, structured
        outputs, fallbacks, load balancing, caching, and automatic logging.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ChatCompletionRequest'
            example:
              model: gpt-4o-mini
              messages:
                - role: user
                  content: Reply with exactly ok.
              max_tokens: 16
              temperature: 0
      responses:
        '200':
          description: A chat completion (or, when stream=true, an SSE stream of chunks).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ChatCompletionResponse'
            text/event-stream:
              schema:
                type: string
                description: 'data-only Server-Sent Events terminated by a [DONE] sentinel.'
        '401':
          $ref: '#/components/responses/Unauthorized'
        '429':
          $ref: '#/components/responses/RateLimited'
  /request-logs/:
    post:
      operationId: createRequestLog
      tags:
        - Logging
      summary: Create a request log (span)
      description: >-
        Asynchronously ingest telemetry for an LLM call that was not routed
        through the gateway proxy. The legacy alias `/request-logs/create/`
        remains supported.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/RequestLog'
            example:
              model: gpt-4o
              prompt_messages:
                - role: user
                  content: Hello
              completion_message:
                role: assistant
                content: Hi there!
              prompt_tokens: 8
              completion_tokens: 4
              cost: 0.00012
              latency: 0.42
              customer_identifier: user_123
      responses:
        '200':
          description: Log accepted.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/SimpleMessage'
        '401':
          $ref: '#/components/responses/Unauthorized'
  /prompts/:
    post:
      operationId: createPrompt
      tags:
        - Prompts
      summary: Create a prompt
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PromptCreate'
            example:
              name: customer_support
      responses:
        '201':
          description: Prompt created.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Prompt'
        '401':
          $ref: '#/components/responses/Unauthorized'
    get:
      operationId: listPrompts
      tags:
        - Prompts
      summary: List prompts
      responses:
        '200':
          description: A list of prompts.
          content:
            application/json:
              schema:
                type: object
                properties:
                  results:
                    type: array
                    items:
                      $ref: '#/components/schemas/Prompt'
  /prompts/{prompt_id}/:
    parameters:
      - $ref: '#/components/parameters/PromptId'
    get:
      operationId: retrievePrompt
      tags:
        - Prompts
      summary: Retrieve a prompt
      responses:
        '200':
          description: A prompt.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Prompt'
        '404':
          $ref: '#/components/responses/NotFound'
    patch:
      operationId: updatePrompt
      tags:
        - Prompts
      summary: Update a prompt
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PromptCreate'
      responses:
        '200':
          description: Prompt updated.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Prompt'
    delete:
      operationId: deletePrompt
      tags:
        - Prompts
      summary: Delete a prompt
      responses:
        '204':
          description: Prompt deleted.
  /prompts/{prompt_id}/versions/:
    parameters:
      - $ref: '#/components/parameters/PromptId'
    post:
      operationId: createPromptVersion
      tags:
        - Prompts
      summary: Create a prompt version
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PromptVersion'
            example:
              model: gpt-4o
              messages:
                - role: system
                  content: You are a helpful {{role}} assistant.
              temperature: 0.7
      responses:
        '201':
          description: Prompt version created.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PromptVersion'
    get:
      operationId: listPromptVersions
      tags:
        - Prompts
      summary: List prompt versions
      responses:
        '200':
          description: A list of prompt versions.
          content:
            application/json:
              schema:
                type: object
                properties:
                  results:
                    type: array
                    items:
                      $ref: '#/components/schemas/PromptVersion'
  /log_threads/:
    post:
      operationId: listThreads
      tags:
        - Threads
      summary: List threads
      description: Retrieve conversation threads matching the supplied filters, with pagination.
      requestBody:
        required: false
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ListFilter'
      responses:
        '200':
          description: A paginated list of threads.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PaginatedList'
  /traces/list/:
    post:
      operationId: listTraces
      tags:
        - Traces
      summary: List traces
      description: Retrieve traces (OpenTelemetry-aligned span trees) with filtering and pagination.
      requestBody:
        required: false
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ListFilter'
      responses:
        '200':
          description: A paginated list of traces.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PaginatedList'
  /traces/{trace_id}/:
    parameters:
      - name: trace_id
        in: path
        required: true
        schema:
          type: string
    get:
      operationId: retrieveTrace
      tags:
        - Traces
      summary: Retrieve a trace
      responses:
        '200':
          description: A trace with its spans.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Trace'
        '404':
          $ref: '#/components/responses/NotFound'
    delete:
      operationId: deleteTrace
      tags:
        - Traces
      summary: Delete a trace
      responses:
        '204':
          description: Trace deleted.
  /evaluators/:
    post:
      operationId: createEvaluator
      tags:
        - Evaluators
      summary: Create an evaluator
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Evaluator'
      responses:
        '201':
          description: Evaluator created.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Evaluator'
  /evaluators/list/:
    post:
      operationId: listEvaluators
      tags:
        - Evaluators
      summary: List evaluators
      description: Retrieve evaluators matching the supplied filters.
      requestBody:
        required: false
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ListFilter'
      responses:
        '200':
          description: A paginated list of evaluators.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PaginatedList'
  /evaluators/{evaluator_id}/run/:
    parameters:
      - name: evaluator_id
        in: path
        required: true
        schema:
          type: string
    post:
      operationId: runEvaluator
      tags:
        - Evaluators
      summary: Run an evaluator
      description: Execute an evaluator against supplied inputs/logs and return scores.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              properties:
                input:
                  type: string
                output:
                  type: string
                metadata:
                  type: object
                  additionalProperties: true
      responses:
        '200':
          description: Evaluation result with score(s).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/EvaluationResult'
  /users/list/:
    post:
      operationId: listUsers
      tags:
        - Users
      summary: List users (customers)
      description: >-
        Retrieve end-user (customer) records in your organization with analytics
        and budget fields. Supports filtering by customer_identifier, name,
        email, number_of_requests, total_tokens, total_cost, and active_days.
      parameters:
        - name: page
          in: query
          schema:
            type: integer
            default: 1
        - name: page_size
          in: query
          schema:
            type: integer
            default: 50
            maximum: 1000
        - name: environment
          in: query
          schema:
            type: string
            enum:
              - prod
              - test
      requestBody:
        required: false
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ListFilter'
      responses:
        '200':
          description: A paginated list of users (customers).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PaginatedList'
  /datasets/:
    post:
      operationId: createDataset
      tags:
        - Datasets
      summary: Create a dataset
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Dataset'
      responses:
        '201':
          description: Dataset created.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Dataset'
    get:
      operationId: listDatasets
      tags:
        - Datasets
      summary: List datasets
      responses:
        '200':
          description: A list of datasets.
          content:
            application/json:
              schema:
                type: object
                properties:
                  results:
                    type: array
                    items:
                      $ref: '#/components/schemas/Dataset'
  /experiments/:
    post:
      operationId: createExperiment
      tags:
        - Experiments
      summary: Create an experiment
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Experiment'
      responses:
        '201':
          description: Experiment created.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Experiment'
components:
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: 'Keywords AI API key supplied as the Authorization Bearer header (Authorization: Bearer <KEYWORDSAI_API_KEY>).'
  parameters:
    PromptId:
      name: prompt_id
      in: path
      required: true
      schema:
        type: string
      description: The prompt identifier (id or slug).
  responses:
    Unauthorized:
      description: Missing or invalid API key.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    NotFound:
      description: Resource not found.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
    RateLimited:
      description: Too many requests.
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
  schemas:
    Message:
      type: object
      required:
        - role
        - content
      properties:
        role:
          type: string
          enum:
            - system
            - user
            - assistant
            - tool
        content:
          type: string
        name:
          type: string
        tool_calls:
          type: array
          items:
            type: object
            additionalProperties: true
    ChatCompletionRequest:
      type: object
      required:
        - model
        - messages
      properties:
        model:
          type: string
          description: Model identifier (any of the 250+ supported models).
        messages:
          type: array
          items:
            $ref: '#/components/schemas/Message'
        stream:
          type: boolean
          description: Stream the response token-by-token via Server-Sent Events.
          default: false
        temperature:
          type: number
          minimum: 0
          maximum: 2
        max_tokens:
          type: integer
        top_p:
          type: number
        frequency_penalty:
          type: number
        presence_penalty:
          type: number
        tools:
          type: array
          items:
            type: object
            additionalProperties: true
        response_format:
          type: object
          additionalProperties: true
        customer_identifier:
          type: string
          description: End-user identifier for per-user analytics and logging.
        fallback_models:
          type: array
          items:
            type: string
        metadata:
          type: object
          additionalProperties: true
    ChatCompletionResponse:
      type: object
      properties:
        id:
          type: string
        object:
          type: string
          example: chat.completion
        created:
          type: integer
        model:
          type: string
        choices:
          type: array
          items:
            type: object
            properties:
              index:
                type: integer
              message:
                $ref: '#/components/schemas/Message'
              finish_reason:
                type: string
        usage:
          type: object
          properties:
            prompt_tokens:
              type: integer
            completion_tokens:
              type: integer
            total_tokens:
              type: integer
    RequestLog:
      type: object
      required:
        - model
      properties:
        model:
          type: string
        prompt_messages:
          type: array
          items:
            $ref: '#/components/schemas/Message'
        completion_message:
          $ref: '#/components/schemas/Message'
        input:
          description: Universal input (string, object, or array).
        output:
          description: Universal output (string, object, or array).
        prompt_tokens:
          type: integer
        completion_tokens:
          type: integer
        cost:
          type: number
          description: Request cost in USD; auto-calculated if omitted.
        latency:
          type: number
          description: Total request latency in seconds.
        customer_identifier:
          type: string
        status:
          type: string
          enum:
            - success
            - error
        timestamp:
          type: string
          format: date-time
        metadata:
          type: object
          additionalProperties: true
    PromptCreate:
      type: object
      required:
        - name
      properties:
        name:
          type: string
        description:
          type: string
    Prompt:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        slug:
          type: string
        description:
          type: string
        current_version:
          type: integer
        created_at:
          type: string
          format: date-time
        updated_at:
          type: string
          format: date-time
    PromptVersion:
      type: object
      required:
        - model
        - messages
      properties:
        model:
          type: string
        messages:
          type: array
          items:
            $ref: '#/components/schemas/Message'
        description:
          type: string
        temperature:
          type: number
        max_tokens:
          type: integer
        top_p:
          type: number
        frequency_penalty:
          type: number
        presence_penalty:
          type: number
        variables:
          type: object
          additionalProperties: true
        fallback_models:
          type: array
          items:
            type: string
        load_balance_models:
          type: array
          items:
            type: object
            additionalProperties: true
        tools:
          type: array
          items:
            type: object
            additionalProperties: true
        response_format:
          type: object
          additionalProperties: true
        deploy:
          type: boolean
    Trace:
      type: object
      properties:
        trace_id:
          type: string
        name:
          type: string
        start_time:
          type: string
          format: date-time
        end_time:
          type: string
          format: date-time
        spans:
          type: array
          items:
            type: object
            additionalProperties: true
    Evaluator:
      type: object
      required:
        - name
      properties:
        id:
          type: string
        name:
          type: string
        type:
          type: string
          description: Evaluator type (e.g. llm, code, human).
        description:
          type: string
        definition:
          type: object
          additionalProperties: true
    EvaluationResult:
      type: object
      properties:
        evaluator_id:
          type: string
        score:
          type: number
        passed:
          type: boolean
        reason:
          type: string
    Dataset:
      type: object
      required:
        - name
      properties:
        id:
          type: string
        name:
          type: string
        description:
          type: string
        created_at:
          type: string
          format: date-time
    Experiment:
      type: object
      required:
        - name
      properties:
        id:
          type: string
        name:
          type: string
        description:
          type: string
        columns:
          type: array
          items:
            type: object
            additionalProperties: true
    ListFilter:
      type: object
      properties:
        filters:
          type: object
          additionalProperties: true
        page:
          type: integer
          default: 1
        page_size:
          type: integer
          default: 50
        sort_by:
          type: string
    PaginatedList:
      type: object
      properties:
        count:
          type: integer
        next:
          type: string
          nullable: true
        previous:
          type: string
          nullable: true
        results:
          type: array
          items:
            type: object
            additionalProperties: true
    SimpleMessage:
      type: object
      properties:
        message:
          type: string
    Error:
      type: object
      properties:
        error:
          type: object
          properties:
            message:
              type: string
            type:
              type: string
            code:
              type: string