saidex 0.2.0

SAIDEX — Structured AI Data EXtraction: validated Pydantic models from LLM responses with automatic retry, fallback, and an agentic tool loop

SAIDEX — Structured AI Data Extraction

Extract validated Pydantic models from LLM responses — with automatic retry, field-level error feedback, an optional fallback model, and an agentic tool loop.

Works with any LangChain-compatible model: GPT-4o, Claude, Gemini, and local models via Ollama or vLLM. No tool-calling support required — use ExtractionMode.JSON for any chat model.

Installation

pip install saidex

With automatic OpenAI rate-limit handling:

pip install "saidex[openai]"

Quick start

import asyncio
from pydantic import BaseModel, Field
from langchain_openai import ChatOpenAI
from saidex import extract_from_text

class PersonInfo(BaseModel):
    name: str
    age: int
    occupation: str

async def main():
    llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

    person, stats = await extract_from_text(
        llm,
        PersonInfo,
        "Alice Müller, 34, works as a software engineer in Munich.",
    )

    print(person)               # name='Alice Müller' age=34 occupation='software engineer'
    print(stats.total_retries)  # 0 — first attempt was valid

asyncio.run(main())

Key features

Feature	Detail
Validated output	Pydantic validation with field-level error feedback fed back to the LLM for self-correction
Auto-retry	Configurable retries with exponential back-off for transient errors and rate limits
Fallback model	Pair a cheap primary with a powerful fallback — pay for the big model only when needed
No tool-calling required	ExtractionMode.JSON works with any chat model including local models via Ollama / vLLM
Agentic tool loop	Give the LLM your own tools (lookups, API calls) — it calls them freely, then delivers a validated final answer
Multimodal	Pass images via standard LangChain messages to any vision-capable model
LangChain-native	Plugs into any LangChain chat model; supports callbacks (Langfuse, LangSmith, …)

Links

• Full documentation
• Examples
• Changelog
• GitHub repository

License

Copyright 2026 Martin Lauff. Licensed under the Apache License, Version 2.0.