tender-recommendation-api

This repository contains a Node.js application built with Express.js that demonstrates RESTful API routing with support for route parameters and query parameters. It includes examples of handling dynamic routes with single and multiple parameters, as well as parsing query strings for flexible client requests.

Features

• Bidder-based recommendations using cosine similarity of embeddings
• Dynamic filtering via query parameters: bidderId, country, state, min_value, currency, page, limit
• Paginated results, ranked by match_score
• Simple MongoDB + Mongoose backend with dummy data support for quick start and testing

API Usage

Endpoint

GET /api/v1/recommendations/tenders

Query Parameters

Parameter	Type	Description
bidderId	string	Unique identifier for the bidder (required)
country	string	ISO country code filter (optional)
state	string	State/province filter (optional)
min_value	number	Minimum tender value (optional)
currency	string	Currency filter using ISO code (optional)
page	number	Page number for pagination (default: 1)
limit	number	Items per page (default: 10, max: 100)

Sample Request

GET http://localhost:5000/api/v1/recommendations/tenders?bidderId=user_12345&country=IN&min_value=10000000&currency=INR&limit=15

How It Works (Controller)

1. Fetch the bidder by bidderId, return error if not found or incomplete.
2. Query and filter tenders in MongoDB based on provided parameters.
3. Compute cosine similarity between bidder's embedding and each tender’s embedding.
4. Sort tenders by relevance (match_score).
5. Apply pagination (page, limit) and respond with recommended tenders.

---

Testing & Development

• Inserted dummy bidders and tenders (with profile_embedding and vectorEmbedding arrays) into MongoDB.
• Tested endpoints using Postman, Thunder Client, or curl:

Tech Stack

• Node.js with Express for API server
• MongoDB + Mongoose for data modeling
• Manual cosine similarity computation