2026-06-09站内改写5 min readUpdated: 2026-06-09

Show HN: Open-source, local-first legal AI workspace for lawyers

AI Blueprint is an open-source, local-first legal AI workspace designed for lawyers and legal teams, offering document-grounded chat, drafting, contract review, preparation workflows, translation, and more, with strong privacy and audit controls.

SourceHacker News AIAuthor: rohasnagpal

Notifications You must be signed in to change notification settings

Fork 5

Star 12

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

100 Commits

.github

app

docs

migrations

public

routes

scripts

tests

.dockerignore

.env.production.example

.gitignore

CODE_OF_CONDUCT.md

CONTRIBUTING.md

Dockerfile

LICENSE

OPERATIONS.md

README.md

SECURITY.md

ai-blueprint-multiuser-plugin-plan.html

alembic.ini

database.py

docker-compose.example.yml

main.py

multi-agent-contract-review-plan.html

requirements-local.txt

requirements.txt

webtools.py

Repository files navigation

A local-first legal AI workspace for document-grounded chat, live voice assistance, legal drafting, contract review, preparation workflows, translation, email drafting, and matter document management.

AI Blueprint is built for lawyers, firms, legal teams, and legal operations teams that need AI to work with confidential documents, matter context, source grounding, review workflows, and auditability.

The goal is not to replace legal judgment. The goal is to give lawyers a structured workspace for reviewing documents, preparing disputes and negotiations, drafting work product, translating legal material, and organizing matter knowledge with stronger control over data, models, and retrieval scope.

Core Capabilities

Document-grounded legal chat and RAG

Ask questions across uploaded files, matter documents, contracts, emails, precedents, policies, statutes, regulations, and approved sources with scoped retrieval.

Legal drafting workspace

Generate legal notices, agreements, replies, board documents, clauses, client-facing drafts, and other legal work product from facts, parties, terms, instructions, jurisdiction, tone, and optional source documents. Drafts include printable HTML, plain text, assumptions, missing information, review warnings, source usage, progress events, and saved draft history.

Contract review workflows

Review indexed contracts directly from a workspace and matter. Use review depth, playbooks, selected source contracts, and review instructions to produce clause analysis, risks, redline suggestions, summaries, warnings, source references, and downloadable Markdown output.

Preparation workflows Prepare structured arbitration, litigation, mediation, and negotiation work product from indexed matter documents. Prep runs can produce case snapshots, issue maps, chronologies, evidence matrices, procedural tasks, witness or participant preparation, damages and remedies analysis, risks, gaps, trace data, warnings, and source basis.

Legal research support Use Chat, Documents mode, Draft, and Prep workflows to research legal questions, analyze uploaded authorities, produce research memos, and test arguments. Current primary navigation does not expose a separate Legal Research screen.

Email and client communication drafting

Poll IMAP inboxes, generate AI-assisted replies using personas and document context, review drafts, and send approved replies through SMTP.

Translation for legal and business documents

Translate pasted text or uploaded documents with legal, business, technical, literal, and plain-language modes, review warnings, translator notes, preserved terms, and downloadable HTML output.

Live voice assistance

Use OpenAI Realtime voice for spoken legal assistance, matter walkthroughs, document questions, and workflow guidance, including document search in Documents mode.

Workspaces, matters, users, and permissions

Organize documents, runs, members, roles, outputs, and access by workspace and matter.

Current navigation Work from New Chat, Add Document, View Documents, Prep, Workflows, Personas, and Settings. The Prep menu includes Arbitration Prep, Litigation Prep, Mediation Prep, and Negotiation Prep. The Workflows menu includes Contract Review, Draft, Email, and Translate.

Document management

Upload files, ingest URLs, connect local folders, sync folder sources, index documents, search document libraries, and manage deletion.

Personas and role-specific behavior

Configure reusable legal roles such as Contract Reviewer, Litigation Associate, Legal Researcher, Partner Reviewer, Plain-English Explainer, Evidence Analyst, and Client Email Drafter.

Auditability, jobs, and trust controls

Track runs, job progress, exports, escalations, audit events, provider/model use, encrypted secrets, review warnings, and human approval points.

Local-first and privacy-conscious deployment

Run locally or deploy for teams with explicit runtime paths, secure cookies, CORS controls, migrations, encrypted secrets, and protected upload/database storage.

Example Workflows

Draft a legal notice from facts, parties, jurisdiction, tone, and selected source documents.

Review a contract against a playbook and produce risks, fallback language, summaries, warnings, and source references.

Prepare for arbitration by mapping issues, chronology, evidence, witnesses, damages, and procedural risks.

Build a litigation preparation package from pleadings, exhibits, emails, and notes.

Prepare mediation or negotiation strategy with positions, interests, information gaps, talking points, concessions, and settlement options.

Draft a legal research memo from a precise question, jurisdiction, facts, and uploaded source materials.

Translate a legal document into another language with preserved terms and review warnings.

Generate and approve a document-grounded client email reply.

Compare clauses across versions of an agreement.

Use a Partner Reviewer persona before sending work product.

Local and Network Versions

AI Blueprint can be used in two ways:

Local version: run AI Blueprint privately on one machine for solo lawyers, experiments, document review, drafting, research, translation, and local-first workflows.

Network/server version: run AI Blueprint for a firm, team, or legal department with shared workspaces, matter access, permissions, audit trails, reusable workflows, and centralized configuration.

The local version is the fastest way to start. The network version is the direction for multi-user legal practice.

Quick Start

Requirements:

Python 3.10 or newer

A browser

Provider API keys depending on the models, voice, and retrieval mode you choose

Run locally:

git clone https://github.com/rohasnagpal/AI-Blueprint.git cd AIBlueprint

python3 -m venv .venv source .venv/bin/activate

pip install -r requirements.txt python main.py

Open:

http://127.0.0.1:8000

On first use, complete the setup flow, create or select a workspace, and open Settings to add the model or provider keys you need.

For closed local deployments that need a temporary bootstrap account, set AI_BLUEPRINT_BOOTSTRAP_DEFAULT_ADMIN=true before starting the server, then sign in with the bootstrap credentials and immediately change them. Do not use the bootstrap account path for public deployments.

Local-Only Beta Notes

For a local-only beta, run AI Blueprint on the user's own machine and open it only through:

http://127.0.0.1:8000 http://localhost:8000

Do not bind the app to a public interface or share the local port on a LAN unless you are intentionally running a network/server deployment.

Recommended local-only settings:

AI_BLUEPRINT_ENV=development AI_BLUEPRINT_SECURE_COOKIES=false AI_BLUEPRINT_CORS_ORIGINS=http://127.0.0.1:8000,http://localhost:8000

HTTPS is not required for same-machine localhost use. HTTPS and AI_BLUEPRINT_SECURE_COOKIES=true are required when AI Blueprint is exposed through a real domain, reverse proxy, tunnel, LAN hostname, or public server.

Local Data and Backups

Local users are responsible for their own backups. Back up these runtime files together:

ai_blueprint.db ai_blueprint_v2.db uploads/ uploads_v2/ chroma_db/ .secret_key .secret_key_v2

The secret key files are required to decrypt stored credentials after restore. Keep backup archives outside the repository and treat them as confidential client data.

Provider Privacy

AI Blueprint runs locally, but configured external providers may still receive data. Depending on enabled settings, model, embedding, retrieval, email, or voice providers may receive prompts, document excerpts, embeddings, credentials, audio, outputs, or related metadata.

Review each provider's data usage and retention policy before using confidential client material. If no external provider is configured, local documents and local databases remain on the user's machine.

First Run

The fastest successful path is:

Create or select a workspace.

Create a matter for the client, dispute, transaction, research project, or internal file.

Add provider API keys in Settings.

Upload or ingest relevant documents.

Wait until documents are indexed.

Use Chat, Voice, Prep, Contract Review, Draft, Translate, or Email.

Review all outputs before using them in client, court, regulatory, or transaction work.

Built With

Python

FastAPI

Uvicorn

SQLAlchemy

Alembic

SQLite

OpenAI and Groq provider integrations

Optional local RAG dependencies with Chroma, LlamaIndex, and sentence-transformers

Vanilla HTML, CSS, and JavaScript frontend

Configuration

AI Blueprint supports configurable model providers, document retrieval settings, personas, email settings, workspaces, matters, upload limits, app branding, users, activity logs, and deployment controls from inside the app and through environment variables.

For legal use, configure the system around the matter:

which documents are available

which sources are approved

which persona, draft settings, prep workflow, or contract review settings should be used

which users can access the matter

which outputs need review before use

which model providers may receive prompts, document snippets, embeddings, and outputs

For production settings, see Deployment Guide.

Privacy Note

AI Blueprint is designed as a local-first legal AI workspace. Sensitive runtime data such as uploaded documents, chat history, draft history, translation history, local databases, vector indexes, logs, and API keys should not be committed to version control.

When using external model, embedding, retrieval, email, or voice providers, review where document text, prompts, embeddings, credentials, audio, and outputs are sent before using the system with confidential material.

AI Blueprint supports legal workflows but does not replace professional legal judgment. Review all outputs before using them in client, court, regulatory, or transaction work.

Documentation

User Help Pack

Deployment Guide

Public Launch Checklist

Operations Guide

Security Policy

Contributing

Public Deployment

Before exposing AI Blueprint to public users, review:

Deployment Guide

Public Launch Checklist

Production deployments should run database migrations explicitly before app startup:

.venv/bin/python scripts/migrate.py

Keep AI_BLUEPRINT_RUN_MIGRATIONS_ON_STARTUP=false in production so multiple workers cannot race while applying migrations.

Development

Install dependencies and run the app as shown in Quick Start. Before opening a pull request, run:

python -m compileall main.py database.py routes rag app migrations python -m unittest discover -s tests

See Contributing for contribution expectations.

Development Status

AI Blueprint is

[truncated for AI cost control]