Sara = LLaMA 3.3 70B + Supabase + TinyFish + Scrapling + GitHub RAG + 4-Agent pipeline
| Repo | Power gained |
|---|---|
| D4Vinci/Scrapling | Adaptive scraping, Cloudflare bypass, stealth mode, spider crawls |
| tinyfish-io/tinyfish-cookbook | Any website → structured JSON, real browser agent |
| Top 20 AI/ML repos | Training data: LangChain, SAM, LLMs-from-scratch, etc. |
New feature: select 2–5 indexed repos → Sara generates a complete working service with architecture, starter code, and deployment guide. Pre-built templates included.
# 1. Create Supabase project → run SQL migration
# 2. Get keys: Groq (console.groq.com) + TinyFish (tinyfish.ai)
# 3. Deploy 5 edge functions:
npx supabase login && npx supabase link --project-ref YOUR_REF
npx supabase functions deploy chat multiagent associate github-sync bulk-sync
# 4. Set secrets in Supabase Dashboard:
# GROQ_API_KEY, TINYFISH_API_KEY, GITHUB_TOKEN (optional)
# 5. npm install && npm run dev → Settings → paste Supabase URL+Key- Solo mode: LLaMA 3.3 70B + RAG + live web
- 4-Agent mode: Researcher (live web) → Architect → Critic → Synthesizer
- Association Market: 6 pre-built templates + custom combinations
- Bulk sync: Index Top 20 AI repos in 1 click
- Scrapling: Python adaptive scraping in every code answer