Skip to main content
RAGLight keeps generation independent from retrieval. Every built-in adapter satisfies the same structural LLM protocol, so the pipeline only needs generate() and generate_streaming() methods.

Install a provider

Install only the extras used by your application. Multiple extras can be combined in one command.

Local providers

Ollama preloads the model by default. Set preload_model=False when startup should not send a warm-up request.

Hosted providers

Each hosted adapter also accepts an explicit API key where the upstream client supports one. Prefer environment variables or your secret manager over committed credentials.

Use an adapter directly

Adapters may be called without a RAG pipeline:
For retrieval-augmented generation, inject the same object into RAGPipeline or AgenticPipeline. Conversation state is explicit; use the history argument or create a session.

Custom endpoints and models

OpenAILLM accepts api_base and api_key, which makes it suitable for compatible endpoints. For a fully custom backend, implement the small LLM protocol instead of inheriting from a RAGLight base class. See Configuration and extensions.