# scinr

> AI-powered document knowledge library for life sciences — scinr.newton document ingestion engine.

## Overview
scinr is a Python library and CLI tool (`newton`) designed to process complex life sciences documents (PDFs, Word documents, PowerPoint presentations, Excel spreadsheets, CSVs) into structured graph knowledge in Neo4j and MongoDB.

## Key Capabilities
- **5-Stage Document Ingestion Pipeline**: Preprocess -> Extract -> Ingest -> Annotate -> Entity Extraction.
- **Tabular Data Normalization**: Automatic detection and normalization of spreadsheet headers and tabular schemas.
- **Graph Knowledge Storage**: Direct mapping of extracted entities, concepts, and relationships into Neo4j subgraphs.
- **Pydantic Extraction Catalogs**: Custom extraction models registered dynamically.

## Documentation
- [Getting Started](getting-started.md): Installation, prerequisites, and quick start example.
- [Architecture](architecture.md): Overview of stages 0-4 and internal storage model.
- [Configuration](configuration.md): Environment settings, database connections, LLM provider setup.
- [CLI Reference](cli.md): `newton` CLI command line tools and options.
- [User Guides - Custom Models](user-guides/custom-models.md): Defining Pydantic extraction models.
- [User Guides - Tabular Pipeline](user-guides/tabular-pipeline.md): Processing Excel and CSV spreadsheets.
- [User Guides - Neo4j Graph](user-guides/neo4j-graph.md): Neo4j graph schema and triple storage.
- [API Reference](api/index.md): Complete Python library API documentation.

## Key Public Python APIs
- `scinr.newton.run_pipeline`: Execute the document ingestion pipeline.
- `scinr.newton.configure`: Configure environment, database URIs, and LLM parameters.
- `scinr.newton.get_config`: Retrieve active library configuration.
- `scinr.newton.exceptions.ScinrError`: Base exception family for scinr error handling.
- `scinr.newton.results.PipelineResult`: Document execution summary and metrics.