OpenSight

Document intelligence and OSINT platform for local corpus analysis.

Ingest documents, extract entities, build a knowledge graph, detect communities, generate LLM summaries, and export investigation bundles — all running locally, no cloud required.

What it does

• Document ingestion — txt, json, markdown, structured or unstructured
• Entity extraction — people, orgs, locations, dates with relationship mapping
• Knowledge graph — force-directed visualization, typed edges, community detection
• Timeline engine — chronological event reconstruction across documents
• Narrative analysis — LLM-assisted summarization via local or API-connected models
• Investigation bundles — export/import .osight packages for sharing findings

Stack

FastAPI · SQLAlchemy · SQLite (FTS5) · Python 3.12 · Docker

Quick start

cp .env.example .env   # add your API key if using LLM features
bash start_services.sh

The API runs on :8000, semantic service on :8010.

Examples

The examples/ directory contains synthetic demo datasets:

• opensight_synthetic_legal_dataset/ — 400 structured legal documents
• opensight_demo_wow_json/ — 5,000 document JSON corpus
• opensight_demo_wow_txt/ — 5,000 document plain text corpus

Requirements

See requirements.txt. Use a virtualenv.

Structure

apps/api/          FastAPI backend + routes
modules/           entity extraction, graph engine, timeline, analysis
semantic_service.py  semantic search microservice
ui/                investigation dashboard
examples/          demo datasets
tests/             test suite

---

LANimals collective // badBANANA

---

Screenshots