A lightweight, Palantir Foundry-inspired platform for constructing domain ontologies from raw data. It targets engineers and researchers needing to transform heterogeneous data into structured, queryable knowledge graphs. The platform offers two paths: a visual data integration pipeline (v2) and a simplified LLM-based extraction (v1), enabling the creation of entities, relations, logic rules, and executable actions.

How It Works

Two primary paths exist: Pipeline Mapping (v2) uses a visual canvas for data connection, transformation, and curated dataset creation, featuring an auto-mapping engine for inferring entities, properties, and relations from diverse data types. Simple LLM Extraction (v1) generates knowledge graphs directly from documents via prompts and models. A novel LLM-driven Quality Audit agent systematically validates ontology integrity using built-in inspection tools.

Quick Start & Requirements

Installation via Docker Compose (v2 full stack or v1 lightweight) or manual setup. Docker requires cloning, copying .env.example to .env, and running docker compose -f docker-compose.v2.yml up --build. Manual setup needs Python 3.11+ and Node.js 18+, followed by backend/frontend service startups. Default login: admin/admin123. Optional services (Neo4j, MinIO, ChromaDB, Redis) integrate or fall back to SQLite/local files.

Highlighted Details

• Visual pipeline builder with connectors and transform routes.
• Automated ontology mapping engine with cross-dataset link inference, including LLM-assisted matching.
• LLM-driven Quality Audit agent for ontology validation.
• Graceful degradation ensures core functionality when optional services are unavailable.
• Multi-language UI (English/Chinese) and JWT user management.

Maintenance & Community

The README does not specify community channels (e.g., Discord, Slack), notable contributors, sponsorships, or a public roadmap.

Licensing & Compatibility

Released under the MIT license, permitting broad usage, including commercial applications and integration into closed-source projects.

Limitations & Caveats

LLM extraction may cause Out-of-Memory (OOM) errors on low-memory systems; the system defaults to serial extraction (max_workers=1) to mitigate this. Users may need to process domains individually or reduce file counts. Optional services' absence triggers fallback mechanisms, potentially impacting performance or features.