Sparrow is an API-first platform for enterprise document intelligence that extracts structured data from documents using machine learning, large language models, and vision models.
The platform addresses the challenge of converting unstructured documents like invoices, receipts, statements, and forms into clean, validated structured data. It combines multiple extraction approaches: Vision LLM pipelines for document parsing, text LLM pipelines for instruction processing and validation, and agent frameworks for orchestrating multi-step workflows. All processing runs on your own infrastructure without external API dependencies, and the same API surface works across different backends including MLX for Apple Silicon, vLLM for NVIDIA GPUs, Ollama, Hugging Face, and Mistral OCR.
Teams should adopt this tool if they need production-ready document intelligence integrated directly into their backend systems via REST APIs. It suits enterprises processing high volumes of structured documents where data quality and control over infrastructure matter. The platform offers both programmatic access through APIs and a web UI with drag-and-drop document upload, real-time processing, and JSON-based schema configuration. Organizations can choose between Vision LLM extraction for visual document understanding, text LLM instruction processing for validation and decision-making, or agent pipelines for complex workflows with visual monitoring through Prefect.
The project shows consistent development activity with regular commits addressing core functionality and infrastructure improvements. Work spans multiple areas including API enhancements, pipeline optimization, backend integration, and UI refinement. The codebase maintains focus on production readiness with attention to error handling, performance tuning, and documentation across different deployment scenarios.
