A sample implementation of a managed AI search service using RAG pipelines with embeddings, reranking, and pgvector - powered by Heroku Managed Inference.

• Index any documentation - Crawl and index URLs automatically
• AI-powered chat - Ask questions, get answers with citations
• Multiple pipelines - Create separate knowledge bases for different docs
• Embeddable widget - Add to your site with a simple script tag
• Real-time streaming - See responses as they're generated

Connect to multiple data sources including URLs, with support for Notion, Google Drive, Salesforce, Slack, and more coming soon.

Watch your documents flow through the full RAG pipeline: Fetch, Parse, Chunk, Embed, and Store in pgvector.

Get an embeddable widget for your indexed documentation. Copy the iframe code or widget URL to add AI-powered search to any website.

Ask questions and get grounded answers with source citations. Toggle Cohere Rerank to improve retrieval precision.

1. Click the Deploy to Heroku button above
2. Create the required AI model add-ons:

heroku addons:create heroku-inference:cohere-embed-multilingual
heroku addons:create heroku-inference:cohere-rerank-3-5
heroku addons:create heroku-inference:claude-3-5-sonnet

3. Set environment variables from the add-ons:

# Get the config vars from each addon and set them
heroku config:set EMBEDDING_URL=<from cohere-embed addon>
heroku config:set EMBEDDING_KEY=<from cohere-embed addon>
heroku config:set RERANKING_URL=<from cohere-rerank addon>
heroku config:set RERANKING_KEY=<from cohere-rerank addon>
heroku config:set INFERENCE_URL=<from claude addon>
heroku config:set INFERENCE_KEY=<from claude addon>

# Clone the repo
git clone https://github.com/dsouza-anush/heroku-rag-chatbot
cd heroku-rag-chatbot/frontend

# Install dependencies
npm install

# Set environment variables
cp .env.example .env.local
# Edit .env.local with your credentials

# Run development server
npm run dev

Open http://localhost:3000

┌─────────────────────────────────────────────────────────────────┐
│                         MIA CHATBOT                              │
└─────────────────────────────────────────────────────────────────┘

 INDEXING PIPELINE
 ─────────────────
 URL → Crawl → Chunk → Cohere Embed → pgvector

 QUERY PIPELINE
 ──────────────
 Question → Embed → Vector Search → Rerank → Claude → Answer

POST /api/pipelines
{ "name": "My Docs", "description": "Documentation for my project" }

POST /api/pipelines/:id/index
{ "url": "https://docs.example.com", "max_pages": 20 }

POST /api/pipelines/:id/chat
{ "message": "How do I get started?" }

GET /api/pipelines/:id/status

• Frontend: Next.js 16, React 19, Tailwind CSS, shadcn/ui
• Database: PostgreSQL with pgvector extension
• AI: Heroku Managed Inference (Cohere, Claude)

MIT