Tavily RAG Dashboard

A Nuxt.js + TypeScript dashboard that accepts prompts, searches the web with Tavily, and returns Tavily's generated answer with source links.

Live Demo

Try the deployed dashboard at https://tavily-rag-dashboard-20260603.netlify.app/.

Suggested Prompts

Use the prompt chips to fetch real-time web data through Tavily:

• What's the Nebius stock today?
• Anthropic latest funding round
• What are the latest npm vulnerabilities in agent plugins?

Setup

1. Copy .env.example to .env and fill in the keys.
2. Install JavaScript dependencies:

npm install

3. Start the app:

npm run dev

Open http://localhost:3000.

For production builds:

npm run build
npm start

Netlify Deployment

This project includes netlify.toml with the Nuxt build settings:

• Build command: npm run build
• Publish directory: dist
• Node version: 22

Set the required environment variables in Netlify under Site configuration -> Environment variables before deploying. Do not commit real API keys.

Required Environment

• TAVILY_API_KEY: Tavily Search API key.
• PINECONE_API_KEY: Pinecone API key.
• PINECONE_HOST: Pinecone index host for an index with integrated embedding enabled.

Optional knobs include TAVILY_MAX_RESULTS, PINECONE_NAMESPACE, PINECONE_TEXT_FIELD, and PINECONE_API_VERSION.

Flow

1. The Nitro API route searches Tavily for fresh web context.
2. It stores every Tavily source in Pinecone, linked to the original user_query and a stable query_id.
3. It returns Tavily's generated answer when one is available.
4. If Tavily does not provide a direct answer, it returns a short fallback built from top source snippets.
5. The API returns the answer and Tavily sources to the UI.

Pinecone Storage

The app uses Pinecone's integrated embedding upsert endpoint:

POST /records/namespaces/{namespace}/upsert

Each Tavily source is stored as a text record. The text field defaults to text, but you can change it with PINECONE_TEXT_FIELD to match your index field_map.

Each record includes metadata linking it back to the user prompt:

• user_query
• query_id
• source_rank
• source_title
• source_url
• source_type
• queried_at