/claim #273

Proposed Changes

Add Qdrant vector database support to the JavaScript SDK using the Qdrant REST API directly (no external qdrant packages), as requested in #273.

New Qdrant class (arakoodev/src/vector-db/src/lib/qdrant/qdrant.ts)

Wraps the Qdrant REST API with the same retry-based resilience pattern used by the existing Supabase class:

Key design decisions

• No external packages: Uses native fetch + Qdrant REST API directly, as requested
• Retry logic: Exponential backoff with 5 retries (same as Supabase class)
• API key auth: Supports api-key header for Qdrant Cloud
• Configurable: URL/API key via constructor or env vars (QDRANT_URL, QDRANT_API_KEY)
• TypeScript types: Full interfaces for points, search params, etc.

Files changed

• arakoodev/src/vector-db/src/lib/qdrant/qdrant.ts — Qdrant client class
• arakoodev/src/vector-db/src/index.ts — Export Qdrant alongside Supabase
• arakoodev/src/vector-db/src/tests/qdrant/ — 5 test files (createCollection, upsertPoints, searchPoints, getPoints, deletePoints)
• examples/chat-with-pdf-qdrant/ — Working example demonstrating all operations

Proof

Usage:

import { Qdrant } from "@arakoodev/edgechains.js/vector-db";

const qdrant = new Qdrant("http://localhost:6333", "optional-api-key");

// Create collection
await qdrant.createCollection({
    collectionName: "documents",
    vectorSize: 1536,
    distance: "Cosine",
});

// Upsert vectors
await qdrant.upsertPoints({
    collectionName: "documents",
    points: [{
        id: 1,
        vector: embeddings,
        payload: { content: "document text", source: "page_1" },
    }],
});

// Search
const results = await qdrant.searchPoints({
    collectionName: "documents",
    vector: queryEmbedding,
    limit: 5,
});

Checklist

• PR created against the correct branch (ts)
• Tests added that prove the feature works
• No external qdrant packages used — REST API only
• Follows existing codebase patterns (retry logic, class structure)
• Working example included