Tapestry

Simulate your life and generate branching future possibilities. Tapestry pairs an AI interview with a 3D life-graph so you can explore next steps before you take them.

---

Overview

• Interactive life simulator: speak or type with an agent, then see possible futures visualized as a branching graph.
• Full-stack: Next.js 15/React 19 frontend + FastAPI backend with streaming agents.
• Data-backed: PostgreSQL via Drizzle on the frontend and psycopg2 on the backend keeps nodes and links consistent.

---

Principles

1. Future-first – Every action should reveal multiple plausible futures, not a single line.
2. Context-rich – Personal background captured in the interview steers all generations.
3. Human-in-the-loop – You can prune, extend, or rerun paths instead of accepting a fixed plan.

---

Architecture

Frontend (Next.js)

• Next.js 15 + React 19 with Tailwind CSS 4.
• NextAuth Google OAuth for sign-in; guards redirect to interview or life view.
• 3D visualization with react-force-graph-3d/Three.js for nodes, links, and “Now” anchoring.
• Optional speech-to-text for the interview; SSE client for live agent responses.

Backend (FastAPI + ADK agents)

• FastAPI service (backend/main.py) streaming via SSE and receiving messages via HTTP.
• Agent stack (ADK) orchestrates interviewer, node maker, and reviewer to grow paths.
• Relational storage with PostgreSQL (DATABASE_URL) using psycopg2; shared with frontend.
• CORS open to http://localhost:3000 for local dev.

Data & Graph

• Nodes/links persisted in Postgres; “Now” is fixed, future nodes branch with time offsets.
• Drizzle schema and migration helpers on the frontend (pnpm db:generate|push|migrate).
• Life graph UI lets you click to expand futures, delete, or inspect nodes.

Flow

1. Sign in with Google.
2. Complete the AI-led interview (text or voice) to capture background/goals.
3. Backend agents generate branching nodes; frontend renders and lets you extend or prune paths.

---

Setup

Requirements

• Node.js 20+
• pnpm 10.14+
• Python 3.10+ (backend) and virtualenv
• PostgreSQL (local or Docker)
• Google API key for the agent (Gemini) and Google OAuth credentials for NextAuth

1) Frontend env (.env in project root)

AUTH_SECRET=your_session_secret
GOOGLE_CLIENT_ID=your_google_oauth_client_id
GOOGLE_CLIENT_SECRET=your_google_oauth_client_secret
DATABASE_URL=postgresql://USER:PASSWORD@localhost:5432/tapestry
NEXTAUTH_URL=http://localhost:3000
BACKEND_URL=http://localhost:8000

2) Backend env (.env in project root, used by backend/main.py)

GOOGLE_API_KEY=your_gemini_api_key
DATABASE_URL=postgresql://USER:PASSWORD@localhost:5432/tapestry

3) Install deps

pnpm install
python3 -m venv backend/venv
source backend/venv/bin/activate
pip install -r backend/requirements.txt

4) Start Postgres

./start-database.sh           # uses DATABASE_URL
# or: docker-compose up -d    # uses docker-compose.yml

5) Run backend

source backend/venv/bin/activate
uvicorn backend.main:app --host 0.0.0.0 --port 8000 --reload

6) Run frontend

pnpm dev

Visit http://localhost:3000.

One-shot (docker/runtime helper)

./start-services.sh

---

Useful scripts

• pnpm db:generate | db:push | db:migrate – manage Drizzle migrations.
• pnpm lint, pnpm check, pnpm format:write – quality passes.
• curl -H "Content-Type: application/json" http://localhost:8000/adk/events/123 – quick SSE sanity check.

---

What you get

• AI-assisted interview that adapts to your background.
• Branching life graph with editable paths to explore multiple futures.
• Local-first, container-friendly stack for fast experimentation and deployment.