justvugg/judicex

Judicex is an open-source workspace for lawyers and legal teams that uses AI to help with legal drafting, document analysis, and case research, while keeping client data entirely on your own machine. The central problem it addresses is that most legal AI tools are closed cloud services, meaning client files are sent to a third-party vendor's servers. Judicex runs locally on a laptop, workstation, or private server inside a law firm, with all data stored in a local database file you control. It supports multiple AI providers, including fully offline local models, or cloud providers like OpenAI and Anthropic, and can also run in a no-AI mode for purely deterministic work. Its distinctive design principle is what it calls an "answer contract": the AI is only allowed to answer questions that are supported by evidence in your local document library. If the evidence is insufficient, it abstains rather than guessing. Every answer cites the specific document it came from. This prevents the hallucination problem that makes generic AI tools risky in legal contexts. In practice, you ingest official legal sources and private matter files into the system. The web interface then lets you chat with those documents, run structured workflow checks against a matter (such as identifying present, partial, or missing elements for a civil case type), draft documents in a split-view editor, extract facts and timelines from uploaded files, and export finished drafts as Word or PDF files. The tech stack is Python with a Flask web interface, SQLite for storage, and vanilla JavaScript on the frontend. It is in alpha and licensed Apache-2.0. The full README is longer than what was provided.