Quickstart

Install

uv add lexigram-ai
# Or with pip
pip install lexigram-ai

lexigram-ai depends on lexigram, lexigram-contracts, lexigram-ai-llm, lexigram-ai-rag, lexigram-ai-feedback, and lexigram-ai-observability.

Minimal wiring

import asyncio
from lexigram import Application
from lexigram.ai.module import AIModule
from lexigram.ai.llm import ClientConfig


async def main() -> None:
    config = AIConfig(
        llm=ClientConfig(
            provider="openai",
            model="gpt-4o",
            api_key="sk-...",
        ),
    )

    async with Application.boot(
        name="ai-demo",
        modules=[AIModule.configure(config)],
    ) as app:
        # LLMClientProtocol is now injectable
        from lexigram.contracts.ai import LLMClientProtocol

        llm = await app.container.resolve(LLMClientProtocol)
        result = await llm.complete([{"role": "user", "content": "Say hello!"}])

        if result.is_ok():
            print(result.unwrap().content)
        else:
            print(f"Error: {result.unwrap_err()}")


asyncio.run(main())

What you get

AIModule.configure(config) registers the AIProvider which wires:

LLM — multi-provider client (OpenAI, Anthropic, Ollama, Groq, Mistral, …)
Vector — vector store backends (optional, requires lexigram-vector)
RAG — retrieval-augmented generation pipelines (optional)
Observability — AI tracing, metrics, and health monitoring
Governance — audit logging and policy enforcement (optional)

Next steps

Guide — mental model, core concepts, end-to-end workflows
Architecture — provider composition and entry-point discovery
Configuration — all config keys and env-var overrides
Ecosystem — related AI packages