Automatically find friends within your close network with shared interests.
Skip the awkward conversations and get straight into deep conversations about shared passions that go deeper than "how was your day?"
- Connect — Sync your Instagram account
- Analyze — AI analyzes your posts, captions, and bio to map your interests
- Discover — See a visual graph of your interests and find people who share them
- Connect — Get AI-generated icebreakers for shared interests
- Next.js + React + TypeScript
- Tailwind CSS + shadcn/ui
- D3.js force-directed graph visualization
- Bun runtime
- FastAPI (Python)
- Neo4j graph database
- Atlas Pipeline scraper-standalone (Instagram HTTP scraping)
- UV package manager
| Service | Role |
|---|---|
| Reka | Image analysis, icebreaker generation |
| Pioneer | GLiNER2 zero-shot NER — entity extraction |
| Yutori | Research & scouting tasks for interest enrichment |
| Modulate | Speech-to-text for voice ingestion |
- Node.js 18+ / Bun
- Python 3.11+
- Docker (for Neo4j)
# Clone
git clone https://github.com/shlawgathon/friendly.git
cd friendly
# Backend
cd backend
cp .env.example .env # Add your API keys
uv sync
source .venv/bin/activate
uvicorn app.main:app --port 8000 --reload
# Frontend (new terminal)
cd frontend
bun install
bun run dev# Local (Docker)
docker compose up neo4j -d
# Or use Neo4j Aura (cloud) — update NEO4J_URI in .env# Neo4j
NEO4J_URI=bolt://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=your_password
# AI Services
REKA_API_KEY=your_key
PIONEER_API_KEY=your_key
YUTORI_API_KEY=your_key
MODULATE_API_KEY=your_key
# Scraper Standalone
SCRAPER_URL=http://localhost:8090
SCRAPER_API_KEY=your_key| Setting | Default | Description |
|---|---|---|
MAX_POSTS_PER_INGEST |
10 | Posts scraped per ingestion request |
MAX_POSTS_HARD_LIMIT |
25 | Absolute cap on posts per scrape |
TOP_INTERESTS_FOR_YUTORI |
3 | Top interests sent to Yutori for enrichment |
MAX_PARALLEL_REKA_CALLS |
2 | Concurrent Reka image analysis calls |
API_TIMEOUT_SECONDS |
20 | Timeout for external API calls (Pioneer, Reka) |
MAX_RETRIES |
3 | Retry count for failed API calls |
INGEST_COOLDOWN_MINUTES |
5 | Cooldown between re-ingesting the same user |
SCRAPER_URL |
http://localhost:8090 |
scraper-standalone service URL |
The Pioneer GLiNER2 NER model extracts the following entity types from combined caption/bio text:
hobby · location · brand · activity · sport · food · music · art
Threshold: 0.5 (configurable per call)
graph LR
Frontend["Frontend<br/>Next.js"] --> Backend["Backend<br/>FastAPI"]
Backend --> Neo4j["Neo4j<br/>Graph DB"]
Backend --> Scraper["scraper-standalone<br/>:8090"]
Backend --> Reka
Backend --> Pioneer["Pioneer<br/>GLiNER2"]
Backend --> Yutori
Backend --> Modulate
flowchart TD
A["POST /api/ingest/instagram"] --> B["Scrape via scraper-standalone"]
B --> C["Profile + Posts + Reels"]
C --> D["Reka Vision — analyze images"]
C --> E["Collect captions + bio text"]
D --> F["Combined text corpus"]
E --> F
F --> G["Pioneer GLiNER2 — extract entities"]
G --> I["Write entities to Neo4j"]
I --> J["Yutori — research & scouting tasks"]
J --> K["Job complete ✓"]
MIT
