Legal Communications Template Engine

Generate common legal communications from templates with intelligent variable insertion. Supports 7 communication types with a web interface, API, and optional AI-powered text enhancement.

Communication types: Breach/Cure Notices, Privacy Incident Notifications, Contract Award/Rejection Letters, Vendor Termination Letters, Legal Hold Notices, Regulatory Inquiry Responses, Board/Executive Legal Updates.

---

Quick Start

Option 1: Run locally (recommended)

git clone https://github.com/lquantprojects/legal-comms-engine.git
cd legal-comms-engine
./start.sh          # macOS/Linux
start.bat           # Windows

This installs dependencies, starts the API and frontend, and opens your browser.

• Frontend: http://localhost:8501
• API Docs: http://localhost:8000/docs

Option 2: Docker

docker compose up

• Frontend: http://localhost:8501
• API Docs: http://localhost:8000/docs

Option 3: Manual setup

python3 -m venv .venv
source .venv/bin/activate       # macOS/Linux
pip install -e .
python scripts/create_templates.py

# Terminal 1: Start API
uvicorn api.main:app --reload --port 8000

# Terminal 2: Start frontend
streamlit run lce_lce_frontend/app.py --server.port 8501

---

Features

• 7 communication types with structured input forms and Word (.docx) output
• Template-driven: .docx templates with Jinja2 placeholders — easy to customize
• Communication log with search, filter, and status tracking (draft/sent/final)
• AI enhancement (optional): Bring your own Anthropic API key to refine tone and adapt for jurisdiction
• API-first: Full REST API with Swagger docs — usable from curl, scripts, or integrations
• Extensible: Add a new communication type in 3 steps (schema + registry entry + template file)

AI Enhancement (Optional)

To enable Claude-powered text enhancement:

1. Go to Settings in the sidebar
2. Enter your Anthropic API key (starts with sk-ant-)
3. When generating a communication, check "Enhance with Claude"
4. Choose a tone (formal, firm, conciliatory, neutral, urgent) and/or jurisdiction
5. Review the enhanced text before generating

Your API key is stored only in your browser session and is never saved to disk.

---

API Reference

The API is documented at http://localhost:8000/docs (Swagger UI).

Key endpoints:

Method	Endpoint	Description
GET	/templates	List available communication types
GET	/templates/{type}/schema	Get JSON Schema for a type's fields
POST	/communications/generate	Generate a communication
GET	/communications/{id}/download	Download generated .docx
GET	/communications	List/search/filter communication log
PATCH	/communications/{id}/status	Update status (draft/sent/final)
POST	/claude/enhance	Enhance text with Claude (requires API key header)

Example: Generate via curl

curl -X POST http://localhost:8000/communications/generate \
  -H "Content-Type: application/json" \
  -d '{
    "communication_type": "breach_cure_notice",
    "fields": {
      "counterparty": "Acme Corp",
      "breach_description": "Failure to deliver under Section 4.2",
      "cure_period": "30 days",
      "remediation_steps": ["Deliver outstanding items", "Provide status report"]
    },
    "author": "Jane Smith"
  }'

---

Project Structure

api/                  FastAPI backend
  routers/            API endpoints
  schemas/            Pydantic models (including communication type registry)
  services/           Business logic (generator, template engine, Claude service)
  models/             SQLAlchemy ORM models
lce_frontend/             Streamlit web interface
  pages/              Generate, Log, Settings pages
templates/            .docx template files with Jinja2 placeholders
data/sample/          Synthetic example data
tests/                pytest test suite

---

Adding a New Communication Type

1. Add a Pydantic schema class in api/schemas/communication_types.py
2. Add a registry entry in COMMUNICATION_TYPE_REGISTRY in the same file
3. Create a .docx template file in templates/ with matching placeholders

No router, frontend, or database changes required.

---

Tech Stack

• Backend: Python 3.11+, FastAPI, SQLAlchemy, Pydantic v2
• Document generation: docxtpl (Jinja2 templates in Word)
• Frontend: Streamlit
• Database: SQLite (designed for PostgreSQL migration)
• AI: Anthropic API (Claude) — optional, BYOK
• Deployment: Docker, Railway

License

MIT