Pieces

Pieces Models API

Enumerate and manage the LLMs available to Pieces OS - list models, download/load/unload local models (e.g. Llama, Mistral via on-device runtime), track download progress, and delete model caches. Local models run fully on-device; cloud models are routed through the same local API surface.

GitHub OpenAPI

Documentation

📖
Documentation
https://docs.pieces.app/products/core-dependencies/pieces-os
📖
APIReference
https://github.com/pieces-app/pieces-os-client-openapi-spec

Specifications

⚙
OpenAPI
https://raw.githubusercontent.com/api-evangelist/pieces/refs/heads/main/openapi/pieces-openapi.yml

Other Resources

🔗
PostmanCollection
https://raw.githubusercontent.com/api-evangelist/pieces/refs/heads/main/collections/pieces.postman_collection.json
🔗
OpenCollection
https://raw.githubusercontent.com/api-evangelist/pieces/refs/heads/main/collections/pieces.opencollection.json
🔗
GitHub
https://github.com/pieces-app

OpenAPI Specification

pieces-openapi.yml Raw ↑ 
openapi: 3.0.1
info:
  title: Pieces OS Local API
  description: >-
    The Pieces OS local REST API. Pieces OS is an on-device process that runs on
    the developer's machine and serves this API over the loopback interface at
    http://localhost:1000. It backs the Pieces Copilot, saved snippets (assets),
    local and cloud model management, and workspace context, and is the source of
    the official OpenAPI-generated SDKs. This document captures the core
    documented resources (assets, formats, copilot/QGPT, conversations, models,
    applications, user, and well-known). For the complete machine-generated
    specification see the upstream pieces-os-client-openapi-spec repository.
  termsOfService: https://pieces.app/legal/terms
  contact:
    name: Pieces Support
    url: https://docs.pieces.app
  version: '1.0'
servers:
  - url: http://localhost:1000
    description: Default on-device Pieces OS port (loopback only).
  - url: http://localhost:5323
    description: Alternate on-device Pieces OS port.
tags:
  - name: Well Known
    description: Health and version of the local Pieces OS instance.
  - name: Assets
    description: Saved snippets stored in the local Pieces OS database.
  - name: Formats
    description: Fragment representations that back each asset.
  - name: QGPT
    description: Pieces Copilot generative engine (question, relevance, stream).
  - name: Conversations
    description: Copilot conversations and their messages.
  - name: Models
    description: Local and cloud LLMs available to Pieces OS.
  - name: Applications
    description: Application registration and sessions.
  - name: User
    description: Local user context.
paths:
  /.well-known/health:
    get:
      operationId: getWellKnownHealth
      tags:
        - Well Known
      summary: Health check for the local Pieces OS instance.
      responses:
        '200':
          description: OK
          content:
            text/plain:
              schema:
                type: string
  /.well-known/version:
    get:
      operationId: getWellKnownVersion
      tags:
        - Well Known
      summary: Returns the running Pieces OS version.
      responses:
        '200':
          description: OK
          content:
            text/plain:
              schema:
                type: string
  /assets:
    get:
      operationId: getAssetsSnapshot
      tags:
        - Assets
      summary: Get a snapshot of all saved assets.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Assets'
  /assets/create:
    post:
      operationId: createAsset
      tags:
        - Assets
      summary: Create (save) a new asset from a seed.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Seed'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Asset'
  /assets/search:
    post:
      operationId: searchAssets
      tags:
        - Assets
      summary: Search saved assets by query.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
              properties:
                query:
                  type: string
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/SearchedAssets'
  /asset/{asset}:
    get:
      operationId: getAsset
      tags:
        - Assets
      summary: Get a single asset by identifier.
      parameters:
        - $ref: '#/components/parameters/AssetId'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Asset'
  /asset/update:
    post:
      operationId: updateAsset
      tags:
        - Assets
      summary: Update an existing asset.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Asset'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Asset'
  /assets/{asset}/delete:
    post:
      operationId: deleteAsset
      tags:
        - Assets
      summary: Delete an asset by identifier.
      parameters:
        - $ref: '#/components/parameters/AssetId'
      responses:
        '200':
          description: OK
  /formats:
    get:
      operationId: getFormatsSnapshot
      tags:
        - Formats
      summary: Get a snapshot of all formats.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Formats'
  /format/{format}:
    get:
      operationId: getFormat
      tags:
        - Formats
      summary: Get a single format by identifier.
      parameters:
        - name: format
          in: path
          required: true
          description: The UUID of the format.
          schema:
            type: string
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Format'
  /qgpt/question:
    post:
      operationId: qgptQuestion
      tags:
        - QGPT
      summary: Ask the Copilot a question grounded in relevant snippets.
      description: >-
        Processes relevant code snippets or UUIDs (typically returned from
        /qgpt/relevance) along with a question query to produce scored answers.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/QGPTQuestionInput'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/QGPTQuestionOutput'
        '401':
          description: Invalid authentication.
        '429':
          description: Too many requests (rate limit or quota exceeded).
        '500':
          description: Internal server error.
        '503':
          description: Service unavailable (engine overloaded).
  /qgpt/relevance:
    post:
      operationId: qgptRelevance
      tags:
        - QGPT
      summary: Compute the snippets most relevant to a query.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/QGPTRelevanceInput'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/QGPTRelevanceOutput'
  /qgpt/reprompt:
    post:
      operationId: qgptReprompt
      tags:
        - QGPT
      summary: Reprompt the Copilot within an ongoing exchange.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                type: object
  /qgpt/stream:
    get:
      operationId: qgptStream
      tags:
        - QGPT
      summary: Streamed Copilot answers over a WebSocket connection.
      description: >-
        Provides a WebSocket connection (ws://localhost:1000/qgpt/stream) that
        streams inputs to the QGPT model and returns incremental answer chunks.
        Handles relevance and questions and manages multiple conversations. This
        is a WebSocket upgrade, not a plain HTTP request; the HTTP method is
        documented here for tooling completeness. See the AsyncAPI document for
        the message-oriented contract.
      responses:
        '101':
          description: Switching Protocols (WebSocket upgrade).
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/QGPTStreamOutput'
  /conversations:
    get:
      operationId: getConversationsSnapshot
      tags:
        - Conversations
      summary: Get a snapshot of all Copilot conversations.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Conversations'
  /conversations/create:
    post:
      operationId: createConversation
      tags:
        - Conversations
      summary: Create a new Copilot conversation.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              type: object
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Conversation'
  /conversation/{conversation}:
    get:
      operationId: getConversation
      tags:
        - Conversations
      summary: Get a single conversation by identifier.
      parameters:
        - name: conversation
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Conversation'
  /conversation/{conversation}/messages:
    get:
      operationId: getConversationMessages
      tags:
        - Conversations
      summary: List the messages in a conversation.
      parameters:
        - name: conversation
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                type: object
  /models:
    get:
      operationId: getModelsSnapshot
      tags:
        - Models
      summary: List the models available to Pieces OS.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Models'
  /model/{model}:
    get:
      operationId: getModel
      tags:
        - Models
      summary: Get a single model by identifier.
      parameters:
        - name: model
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Model'
  /model/{model}/download:
    post:
      operationId: downloadModel
      tags:
        - Models
      summary: Download a local model to the device.
      parameters:
        - name: model
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
  /model/{model}/load:
    post:
      operationId: loadModel
      tags:
        - Models
      summary: Load a downloaded local model into memory.
      parameters:
        - name: model
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
  /model/{model}/unload:
    post:
      operationId: unloadModel
      tags:
        - Models
      summary: Unload a local model from memory.
      parameters:
        - name: model
          in: path
          required: true
          schema:
            type: string
      responses:
        '200':
          description: OK
  /applications:
    get:
      operationId: getApplicationsSnapshot
      tags:
        - Applications
      summary: List registered applications.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Applications'
  /applications/register:
    post:
      operationId: registerApplication
      tags:
        - Applications
      summary: Register an application with Pieces OS.
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/Application'
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Application'
  /applications/session/open:
    post:
      operationId: openApplicationSession
      tags:
        - Applications
      summary: Open an application session.
      responses:
        '200':
          description: OK
  /applications/session/close:
    post:
      operationId: closeApplicationSession
      tags:
        - Applications
      summary: Close an application session.
      responses:
        '200':
          description: OK
  /user:
    get:
      operationId: getUser
      tags:
        - User
      summary: Get the current local user profile.
      responses:
        '200':
          description: OK
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/UserProfile'
components:
  parameters:
    AssetId:
      name: asset
      in: path
      required: true
      description: The UUID of the asset.
      schema:
        type: string
  schemas:
    Asset:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        created:
          type: object
        formats:
          $ref: '#/components/schemas/Formats'
    Assets:
      type: object
      properties:
        iterable:
          type: array
          items:
            $ref: '#/components/schemas/Asset'
    SearchedAssets:
      type: object
      properties:
        iterable:
          type: array
          items:
            type: object
            properties:
              exact:
                type: boolean
              identifier:
                type: string
    Seed:
      type: object
      properties:
        type:
          type: string
        asset:
          type: object
    Format:
      type: object
      properties:
        id:
          type: string
        classification:
          type: object
        fragment:
          type: object
    Formats:
      type: object
      properties:
        iterable:
          type: array
          items:
            $ref: '#/components/schemas/Format'
    QGPTQuestionInput:
      type: object
      properties:
        query:
          type: string
        relevant:
          $ref: '#/components/schemas/QGPTRelevanceOutput'
        model:
          type: string
        application:
          type: string
    QGPTQuestionOutput:
      type: object
      properties:
        answers:
          type: object
          properties:
            iterable:
              type: array
              items:
                type: object
                properties:
                  text:
                    type: string
                  score:
                    type: number
    QGPTRelevanceInput:
      type: object
      properties:
        query:
          type: string
        seeds:
          type: object
    QGPTRelevanceOutput:
      type: object
      properties:
        iterable:
          type: array
          items:
            type: object
            properties:
              seed:
                type: object
              score:
                type: number
    QGPTStreamOutput:
      type: object
      properties:
        question:
          type: object
        status:
          type: string
        conversation:
          type: string
    Conversation:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        type:
          type: string
    Conversations:
      type: object
      properties:
        iterable:
          type: array
          items:
            $ref: '#/components/schemas/Conversation'
    Model:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        cloud:
          type: boolean
        downloaded:
          type: boolean
        loaded:
          type: boolean
    Models:
      type: object
      properties:
        iterable:
          type: array
          items:
            $ref: '#/components/schemas/Model'
    Application:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        version:
          type: string
        platform:
          type: string
    Applications:
      type: object
      properties:
        iterable:
          type: array
          items:
            $ref: '#/components/schemas/Application'
    UserProfile:
      type: object
      properties:
        id:
          type: string
        name:
          type: string
        email:
          type: string
        allocation:
          type: object