Skip to main content

gpt2giga documentation

gpt2giga is a FastAPI compatibility gateway that accepts requests in the OpenAI, Anthropic, and Gemini formats and forwards them to GigaChat. It is useful when a client, editor, agent framework, or SDK can talk to the OpenAI/Anthropic/Gemini API, but the real backend must be GigaChat.

Default local address:

http://localhost:8090

What the proxy covers​

CapabilityWhere to read
Quick start via Docker Compose or uvQuickstart
Supported OpenAI, Anthropic, Gemini, and LiteLLM routesAPI compatibility
Behavior of extra_headers, extra_query, extra_body, and optional fieldsClient parameters
GigaChat built-in tools and their mapping to OpenAI/Anthropic/GeminiBuilt-in tools
Environment variables, authentication, limits, metrics, observabilityConfiguration
Compose profiles, Traefik, nginx, Postgres, OpenSearch, PhoenixDeployment
Runtime logs, traffic logs, admin API, debug translateOperations
Editor, agent, SDK, and reverse-proxy setupIntegrations

Current API surface​

Public routes are available at the root and under versioned prefixes:

  • /chat/completions, /v1/chat/completions, /v2/chat/completions
  • /responses, /v1/responses, /v2/responses
  • /embeddings, /v1/embeddings, /v2/embeddings
  • /messages, /v1/messages, /v2/messages
  • /v1beta/models/{model}:generateContent and compatible Gemini paths
  • /models, /model/info, /health, /ping

The backend selection rule is the same for OpenAI-, Anthropic-, and Gemini-compatible routes: /v1/... always sends chat-like requests to the GigaChat v1 contract, /v2/... sends them to the GigaChat v2 contract, and the root path without /v1 or /v2 uses GPT2GIGA_GIGACHAT_API_MODE=v1|v2.

OpenAI Files/Batches, Anthropic Message Batches, and Gemini Files/Batches are prepared in the code but intentionally not mounted until end-to-end execution is available in the upstream SDK/backend.

Fast path​

  1. Copy .env.example to .env.
  2. Fill in GIGACHAT_CREDENTIALS, GIGACHAT_SCOPE, GIGACHAT_MODEL.
  3. Run docker compose --env-file .env -f deploy/base.yaml --profile DEV up -d.
  4. Check curl http://localhost:8090/health.
  5. Point the SDK at http://localhost:8090/v1 or http://localhost:8090/v2 for an explicit backend contract, or at http://localhost:8090 if the root should follow GPT2GIGA_GIGACHAT_API_MODE.

For developers​