Skip to content

template.md

Latest commit

 

History

History
95 lines (80 loc) · 5.15 KB

template.md

File metadata and controls

95 lines (80 loc) · 5.15 KB

Personal Lawyer Live Docket

Hackathon-ready Copilot experience for personal injury lawyers. The app ingests police-report data from Google Sheets, stages it in a flashy dashboard, and surfaces triage alerts, legal follow-ups, and AI-assisted summaries via CopilotKit + LlamaIndex.

Stack Overview

  • UI: Next.js (App Router), Tailwind, CopilotKit React UI, AG-UI interaction patterns
  • Agent: Python FastAPI service (agent/) running LlamaIndex with Composio MCP tooling
  • Data Source: Google Sheets (read via Composio) – lawyer teammate uploads CSV data to a single tab
  • Notifications & Preferences: In-memory profile store on the Python agent with UI controls for triage filters
  • Outbound Hooks: Resend (email) placeholder wired in UI, outbound voice button reserved for teammate integration

Feature Highlights

  • Google Sheet ingestion with column normalization (incident_id, full_name, incident_category, etc.)
  • Live incident feed that reveals queued reports every few seconds to mimic real-time intake
  • Summary metrics (injury count, property damage count, per-category totals)
  • Triage preference panel (categories, jurisdictions, injury/property toggles) stored server-side
  • Alert panel listing incidents matching the profile; dismissable in UI
  • Copilot chat (CopilotKit + LlamaIndex) grounded on shared dashboard state for natural language queries

Getting Started

Prerequisites

  • Node.js 20+
  • Python 3.10+
  • uv (for Python deps)
  • OpenAI API key (agent/.env)
  • Composio API key & Google Sheets auth config (agent/.env)

Install & Run

# Install (installs Node + Python deps)
pnpm install
# or npm install / yarn install / bun install

# Start both Next.js UI and Python agent
yarn dev
# or pnpm dev / npm run dev / bun run dev

# UI: http://localhost:3000
# Agent: http://localhost:9000

Configure secrets in agent/.env (copy from agent/.env.example) and optional frontend keys in .env.local.

Importing Google Sheet Data

  1. Prepare a Google Sheet tab with the following headers (case-insensitive): 
    incident_id, full_name, sex, home_address, phone_number,
incident_date, incident_time, location, incident_category,
resolution, injury_reported, property_damage,
fault_determination, incident_description
  2. Launch the app — the dashboard automatically connects to the preconfigured sheet (1Dam-5BADE3dYCib1uFdhSNJ8aGkUMJCOEwYCQifsfbk).
  3. The first six rows render instantly; the remaining rows stream into the live feed every few seconds to simulate real-time ingestion.
  4. Update the Google Sheet and press Save in the triage panel (or refresh) to re-sync with the new data.

Triage Preferences & Alerts

  • Update categories or jurisdictions to monitor from the Triage Preferences form. Preferences persist on the backend (/profile/triage).
  • Toggle Alert only when an injury is reported to focus on more severe incidents.
  • Disable property-damage-only alerts by unchecking Include property-damage-only incidents.
  • Matching incidents appear in the Triage Alerts panel; dismissing hides them client-side while the backend keeps historical matches for re-imports.

Copilot Chat

<CopilotChat /> (from CopilotKit) stays pinned on the right rail. It uses the shared DashboardState and LlamaIndex agent instructions defined in agent/agent/agent.py. Use it to:

  • Summarize newest incidents (the model reads cases + queuedCases)
  • Ask for follow-up tasks or email drafts (Resend integration stub)
  • Confirm triage configuration and next steps

Project Structure

.
├── agent/
│   ├── agent.py               # LlamaIndex agent + system prompt + shared state
│   ├── server.py              # FastAPI endpoints (/sheets/sync, /profile, ...)
│   ├── sheets_integration.py  # Google Sheets ingestion + triage evaluation
│   └── profile.py             # Backend profile store for triage prefs
├── src/
│   ├── app/page.tsx           # Next.js dashboard (live feed, metrics, chat)
│   ├── app/api/               # Next.js API routes proxying to Python agent
│   ├── components/dashboard/  # Dashboard UI building blocks
│   └── lib/dashboard/         # Shared types + API helpers
└── docs/                      # Source schema example + LLM doc links

Development Tips

  • Live feed cadence (intervalMs) and streaming toggle live in shared state (DashboardState.liveFeed). Adjust via useCoAgent or Python initial state.
  • To re-run triage after editing preferences, simply save the form—the UI calls /profile/triage then re-imports the sheet.
  • The Resend + voice call actions are currently console stubs. Wire actual APIs by extending onSendEmail / onTriggerVoiceCall in page.tsx.

Roadmap

  • Hook Resend SDK for real outbound notifications
  • Swap voice-call button to teammate’s integration endpoint
  • Persist profile + notifications to a durable store (Redis/Postgres) if required
  • Add richer charts (Recharts already installed) for category trends and response SLAs
  • Expand case workflow with tasks / notes for CRM-style functionality once MVP is locked