Planora AI helps you plan an event. Fill in the event type, city, date, guest count, budget, and preferences and the app returns a starter plan with venue ideas, a budget breakdown, vendor suggestions, a checklist, and a timeline.
The backend uses LangGraph to run specialist agents (venue, budget, logistics) with a budget validator. Venues are grounded in Google Places results, with a Supabase pgvector cache to speed up repeat searches. A React frontend presents the plan with copy and PDF export.
- Event form with example fill, two-column layout, and loading progress
- Multi-agent generation: venue research → venue summary, budget, vendors/checklist/timeline
- Real venue search via Google Places, with pgvector RAG cache (Gemini embeddings)
- Budget validator with automatic retry when totals are off
- Plan results panel: budget bar, interactive checklist, timeline, copy, PDF download
- Groq LLM integration (Llama 3.3 70B)
- Mock fallback when AI is unavailable or when
USE_MOCK_PLAN=true
| Layer | Stack |
|---|---|
| Frontend | React, TypeScript, Vite, jsPDF |
| Backend | Python, FastAPI, Pydantic |
| AI | LangGraph (StateGraph), Groq |
| Data | Google Places API, Supabase (pgvector), Gemini embeddings |
| Validation | Budget sum check with retry loop in the orchestrator |
User fills form (React)
↓
POST /plans/generate (FastAPI)
↓
plan_service → LangGraph orchestrator
↓
venue_research → RAG cache (Supabase) → Google Places fallback
↓
venue_agent → event summary + venue ideas (from real results when available)
↓
budget_agent → budget breakdown
↓
details_agent → vendors, checklist, timeline
↓
validator → check budget total, retry budget if needed
↓
GeneratedPlan JSON → PlanResults UI (copy / PDF)
Venue RAG flow: lookup_venues searches the pgvector cache first. If enough matches exist for the city, those are returned. Otherwise Google Places is queried and results are embedded and saved for next time.
Planora/
frontend/src/
App.tsx Form, hero, loading state
components/PlanResults.tsx Results panel (budget bar, checklist, timeline)
utils/planExport.ts Markdown copy + PDF export
api/plans.ts API client
backend/
main.py FastAPI entry point
routers/plans.py POST /plans/generate
schemas/event.py Pydantic models
services/plan_service.py Orchestrator entry + mock fallback
tools/venue_lookup.py Google Places + cache lookup
memory/vectordb.py pgvector RAG (save + search venues)
agents/
orchestrator.py LangGraph pipeline
venue_research.py Venue lookup node
venue_agent.py Summary + venue selection
budget_agent.py Budget breakdown
details_agent.py Vendors, checklist, timeline
validator.py Budget total check
llm.py Shared Groq helper
Prerequisites: Python 3.10+, Node.js 18+, Groq API key
Optional for real venues and RAG cache: Google Places API key, Supabase project, Gemini API key (see environment variables below).
Backend
cd backend
python -m venv .venv
source .venv/bin/activate # Mac and Linux
pip install -r requirements.txt
cp .env.example .envOn Windows, activate with .venv\Scripts\activate instead of source.
Add keys to backend/.env, then:
uvicorn main:app --reload --port 8000API: http://127.0.0.1:8000
Health: http://127.0.0.1:8000/health
Frontend (second terminal)
cd frontend
npm install
npm run dev| Variable | Required | Description |
|---|---|---|
GROQ_API_KEY |
Yes | Groq API key for LLM calls |
GOOGLE_PLACES_API_KEY |
For real venues | Google Places text search |
SUPABASE_URL |
For RAG cache | Supabase project URL |
SUPABASE_KEY |
For RAG cache | Supabase service role key |
GOOGLE_API_KEY |
For RAG cache | Gemini API key for embeddings (AI Studio) |
USE_MOCK_PLAN |
No | Set to true to skip AI and return a mock plan |
Without Places/Supabase keys, venue research returns empty and the venue agent falls back to LLM-only suggestions.
POST /plans/generate
Example request:
{
"event_type": "Birthday party",
"city": "Chicago",
"date": "2026-06-26",
"guest_count": 50,
"budget": 10000,
"vibe": "Elegant",
"food_preferences": "Both veg and non-veg",
"indoor_outdoor": "Indoor"
}Returns event_summary, venue_ideas, vendor_suggestions, budget_breakdown, checklist, and timeline.
- Built by Bhavneet Kaur
- Feel free to fork, clone, or contribute.