OpenQCoreAI
Get Started

Resources · Docs

Platform Documentation.

Production-ready by design.

OpenQCore documentation for teams building reliable AI products across chat, voice, multimodal generation, and knowledge workflows with scoped runtime control.

Start BuildingAPI Surface
API-first RuntimeWorkspace & Personal ScopesStreaming + Non-StreamingRate-Limit + Dedup + LockVoice Pipeline (STT/LLM/TTS)Console Knowledge APIs

Quick Start

From first request to production path in minutes.

1) Authenticate

Use JWT session auth or API key auth with scope-aware access control.

2) Choose runtime scope

Set personal or workspace context using runtime headers when needed.

3) Call chat completions

Use /api/v1/chat/completions for core assistant execution.

4) Add voice and multimodal

Use /api/v1/voice and Iris endpoints for STT/TTS, image, video, and OCR workflows.

5) Integrate knowledge

Use /api/v1/console/knowledge for upload, ingestion, and retrieval-driven ask.

Base URL

Public API base

/api/v1

All public platform endpoints are mounted under /api/v1.

Authentication & Scope

Context-aware access across personal and workspace execution.

Auth methods

JWT/session or API key

Scope header

X-Scope-Type: personal | workspace

Workspace header

X-Workspace-ID: [uuid] (required in workspace scope)

Agent header

X-Agent-ID: [uuid] (optional)

Validation rules

  • If scope is personal, workspace_id is not allowed.
  • If scope is workspace, workspace_id is required.
  • Invalid scope headers return 422.

Rate Limits

Policy-based throttling by auth context.

API key

60 requests / minute

Bucket key uses API key hash identity.

JWT/session

120 requests / minute

Bucket key uses workspace id when present, otherwise user id.

Throttle response headers

Retry-AfterX-RateLimit-LimitX-RateLimit-RemainingX-RateLimit-Reset

Chat Runtime

/chat/completions with production control layers.

GET /api/v1/chat/health
POST /api/v1/chat/completions

Execution behaviors

  • Backpressure tracking and system overload rejection (503).
  • Distributed lock to prevent duplicate in-flight processing (409).
  • Dedup cache for non-stream responses.
  • Session/message persistence for user and assistant turns.
  • Runtime validation for request, attachments, and session access.

Voice Runtime

STT → LLM → TTS pipeline with streaming and interruption.

POST /api/v1/voice/chat/meta
POST /api/v1/voice/chat
POST /api/v1/voice/interrupt/{session_id}
WS /api/v1/voice/ws

Voice features

  • Fallback provider router (OpenAI + Azure).
  • Circuit breakers for STT, LLM, and TTS phases.
  • Audio conversion pipeline for non-wav input.
  • Streaming audio responses with metadata headers.
  • Cancellation context for session-level interruption.
  • Load shedding and quality degradation under high load.

Iris Multimodal

Model-routed image, video, OCR, and verification workflows.

Model catalog

iris-orbit

Auto route mode

iris-genesis

Image generation

iris-edit

Image editing (not implemented)

iris-lens

Image analysis (not implemented)

iris-video

Video generation

iris-ocr

Document OCR

iris-verify

Identity verification

iris-guard

Fraud detection (not implemented)

iris-face

Face verification (not implemented)

Console Knowledge

Workspace console knowledge APIs for ingestion and ask.

/api/v1/console/knowledge
GET /api/v1/console/knowledge
POST /api/v1/console/knowledge
POST /api/v1/console/knowledge/upload
GET /api/v1/console/knowledge/{item_id}
PATCH /api/v1/console/knowledge/{item_id}
DELETE /api/v1/console/knowledge/{item_id}
POST /api/v1/console/knowledge/ask

Knowledge constraints

  • Max file size: 20MB
  • Allowed types: PDF, DOC, DOCX, TXT, Markdown
  • Ingestion runs asynchronously in background
  • Statuses: pending → processing → processed | failed

Error Semantics

Common platform response classes.

400

Invalid request, missing data, or malformed input

401

Authentication missing or invalid

403

Scope not allowed / workspace role restriction

404

Resource not found

409

Request already processing (lock conflict)

413

Payload too large (e.g. audio/file limits)

415

Unsupported content type

422

Validation failure (scope/context/schema)

429

Rate limit exceeded

503

System overload / provider unavailable

Observability

Traceable execution across chat and voice lifecycles.

  • Trace IDs for request correlation.
  • Usage logging middleware pipeline.
  • OpenTelemetry initialization with OTLP export.
  • Response headers expose timing signals in voice flows.
X-Trace-IDX-Session-IDX-STT-Time-MSX-LLM-Time-MSX-TTS-Time-MS

Docs Navigation

Explore Documentation Sections

Jump directly to implementation guides and production references for each runtime surface.

Authentication

JWT/API key auth, scope headers, and validation model.

Chat API

Completions, streaming behavior, and runtime controls.

Voice API

Realtime voice pipeline, interrupt flow, and metadata headers.

Iris API

Image/video/OCR/verification endpoints and routing.

Knowledge API

Upload, indexing lifecycle, retrieval, and ask workflows.

Providers

Capability matrix, routing policy, and failover strategy.

Output Contracts

Unified response envelope across providers and modalities.

Architecture

Runtime layers, provider abstraction, and system guarantees.

Build with confidence on OpenQCore runtime.

Use the docs as your production map for chat, voice, multimodal generation, and knowledge-backed execution.

Open API ExplorerGo to Infrastructure