Documentation
Jangada AI 🛶 is a thin PT-BR layer over the LLM SDKs, with native
observability in 1 line in your .env. The goal is to swap
provider/model/api_key without changing the rest of your code, with {{ }}
templates, chained flows, structured output (Pydantic), vision, audio, async, and
error-based fallback. It works with Anthropic, OpenAI, Groq, Gemini, Mistral, OpenRouter
and the cloud gateways (AWS Bedrock, Azure OpenAI, Vertex AI).
Installation
pip install "jangada-ai[anthropic]"Where to start
Getting started
Install, configure your key, and make your first call.
Providers and keys
Anthropic, OpenAI, Groq, Gemini, and Mistral with the same API.
Structured output
Output validated with Pydantic on any provider.
Tools (function calling)
The model calls your functions, provider-agnostic.
RAG
Embeddings, vector store, and hybrid search.
Retry and fallback
Defenses against API failures.
Sections
- Getting started — getting-started, parameters, providers, capabilities.
- Providers — Anthropic, OpenAI, Groq, Gemini, Mistral.
- Capabilities — structured output, tools, vision, audio, documents, detection, streaming, RAG, MCP, observability.
- Reliability — errors, retry/fallback, cost, debug.
- Tutorials — step by step from zero to fallback, structured extraction, RAG, and an MCP agent.
- Advanced — extending (new provider) and flows.