Skip to content

README.md

Latest commit

 

History

History
75 lines (49 loc) · 1.5 KB

README.md

File metadata and controls

75 lines (49 loc) · 1.5 KB

ambala

Contextual search engine built on graph + vector retrieval.

Overview

ambala is a hybrid retrieval system that combines semantic similarity with graph traversal to produce context-aware results.

Instead of treating search as a flat nearest-neighbor problem, Ambala evaluates relevance within a structured graph.

This project focuses on:

  • Combining vector similarity with graph traversal
  • Supporting context-constrained queries
  • Making retrieval explainable

Architecture

  • HelixDB → graph + vector storage
  • TanStack Start → API layer + frontend
  • tRPC → type-safe API
  • Gemini → text embeddings (via Vercel AI SDK)

Setup

1. Install HelixDB

curl -sSL "https://install.helix-db.com" | bash
helix install

2. Initialize HelixDB

cd ambala
helix init

This creates db/schema.hx and db/queries.hx which are already configured.

3. Configure Environment

Create .env in apps/web/:

CORS_ORIGIN=http://localhost:3001
GOOGLE_GENERATIVE_AI_API_KEY=your_google_api_key
HELIX_DB_URL=http://localhost:6969

Get a Google API key from https://aistudio.google.com/app/apikey

4. Start HelixDB

helix dev

This starts HelixDB at http://localhost:6969 and deploys the schema/queries.

5. Start the App

bun run dev

Usage

  1. Open http://localhost:3001
  2. Upload a CSV file with messages in format: sender,datetime,content
  3. Search messages semantically (e.g., "when did mahek say hello")

License

MIT