Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

121 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Launcher

An AI-powered startup analysis platform. Input a business idea and get back market validation, a lean business model, competitor positioning, and a 10-slide investor pitch deck generated through a 4-step LLM reasoning pipeline backed by a RAG system.

Live: https://launcher-frontend.onrender.com


What it does

Launcher runs a structured analysis pipeline on any startup idea:

  1. Market categorization - identifies industry, market size, and target customers
  2. SWOT analysis - conditioned on step 1 output
  3. Competitor and positioning analysis - conditioned on steps 1 and 2
  4. Pitch narrative synthesis - investor hook, recommendations, value proposition

Each step feeds into the next. Past analyses are stored as vector embeddings in ChromaDB and retrieved via cosine similarity to provide context for new queries.


Tech stack

Layer Technology
Frontend React, Tailwind CSS
Backend Python, Flask, Gunicorn
LLM Llama 3.3 70B (via Groq)
RAG ChromaDB (persistent vector store, cosine similarity)
Auth JWT (PyJWT)
Database SQLAlchemy, PostgreSQL (SQLite for tests)
CI/CD GitHub Actions
Deployment Render

API endpoints

Endpoint Method Auth Description
/api/validate-idea POST yes Full 4-step pipeline - market validation
/api/generate-plan POST yes Business plan generation
/api/generate-business-model POST yes Lean Canvas generation
/api/generate-pitch POST yes 10-slide investor pitch deck
/api/analyze-market POST no Market analysis via Google Trends + Census
/api/rag-stats GET no ChromaDB vector store stats
/api/health GET no Health check

Architecture

Every analysis endpoint uses a 3-tier fallback:

Groq LLM pipeline -> hardcoded business mappings -> web scraper

If the LLM pipeline returns a valid category, that result is used. If the HF API is down or returns garbage, it falls back to curated mappings (55KB of structured business data). The scraper is last resort.

RAG context is injected into Step 1 of the pipeline. Completed analyses are embedded and stored in ChromaDB so future similar queries benefit from past results.


Running locally

Backend:

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py

Frontend:

cd frontend
npm install
npm run dev

Set GROQ_API_KEY and SECRET_KEY as environment variables before starting the backend.


Tests

cd backend
source venv/bin/activate
pytest tests/test_api.py -v

10 tests covering auth, protected routes, and all major endpoints. 8 pass in CI (2 require a live Groq API key and are expected to skip in the GitHub Actions environment).


Author

Vishak Baddur - https://github.com/VishakBaddur

Releases

Packages

Used by

Contributors

Languages