← ~/projects
$ cat projects/lexa-legal-ai.md
Lexa — Legal Research AI That Abstains
A grounded legal AI that says "I don't know" instead of hallucinating.
Sole developerJuly 2026AI/ML
// why: General models invent case law — one fake citation can cost a career.
$ cat projects/lexa-legal-ai.md▌
TL;DR
- 3 agents: retrieve → draft with citations → independent verify
- Abstains honestly: score floor refuses low-evidence answers
- Offline by default: bundled corpus, live connectors optional
// what does it do?
Legal research assistant on a 3-agent pipeline: Researcher retrieves, Drafter writes with citations, independent Verifier checks every claim. Weak evidence → abstains. Streams live to UI.
// why this project?
General-purpose models happily invent case law, and they rarely push back on a weak premise. In legal research one fabricated citation can cost a career.
// what broke / learned
The verifier rejecting my own drafts hurt, which means it works.
// results
- A deterministic citation guard enforces that every claim maps to a retrieved source
- a score-floor abstention gate refuses low-evidence answers
- a bounded repair loop fixes draft issues before anything reaches the user. Ships offline out-of-the-box (mock model + local embeddings + a bundled US Constitutional corpus), with optional live connectors to Federal Register, CourtListener, and GovInfo. pytest suite plus live-source E2E tests on GitHub Actions CI.
3-agent research, draft, verify pipelineVerifier rejects claims with no retrieved sourceScore-floor abstention gate
Tech stack
PythonFastAPIPostgreSQLAlembicDockerpytestGitHub Actions