Skip to content

BACKEND_MATCHING_TESTING.md

Latest commit

 

History

History
81 lines (60 loc) · 2.49 KB

BACKEND_MATCHING_TESTING.md

File metadata and controls

81 lines (60 loc) · 2.49 KB

Semantic Job Matching - Testing Guide

This guide explains how to test the new Semantic Job Matching Agent.

Prerequisites

Ensure you have:

  1. Running PostgreSQL with pgvector extension (and Docker container running).
  2. GEMINI_API_KEY set in your .env file.
  3. At least one parsed CV and one analyzed Job Post in the database.

1. Automated Testing Script

We have provided a script to quickly verify the matching logic using existing data in your database.

Run the following command:

python scripts/test_job_matching.py

This script will:

  • Connect to your database.
  • Find a candidate with embeddings (ParsedCV).
  • Find a job with embeddings (JobPost).
  • Run matching in both directions (Job -> Candidate, Candidate -> Job).
  • Print the top matches with scores and explanations.

2. API Testing (via Swagger UI)

  1. Start the backend server: 
    uvicorn app.main:app --reload
  2. Open Swagger UI: http://localhost:8000/docs

Test Finding Jobs for a Candidate

  1. Locate GET /api/v1/documents/{document_id}/matching-jobs.
  2. Enter a valid document_id of a parsed CV.
  3. Execute.
  4. Verify the response contains a list of jobs with match_percentage, similarity_score, and recommendation.

Test Finding Candidates for a Job

  1. Locate GET /api/v1/jobs/{job_id}/matching-candidates.
  2. Enter a valid job_id of an analyzed Job Post.
  3. Execute.
  4. Verify the response contains a list of candidates.

3. Sample SQL Queries

You can also inspect the embeddings and basic similarity directly in SQL.

Check Job Embeddings:

SELECT job_post_id, substring(embedded_text for 50) as text_preview, created_at 
FROM job_embeddings 
LIMIT 5;

Find Similar Jobs manually (SQL):

-- Replace [VECTOR] with a real vector array from a document_embedding
SELECT jp.title, (1 - (je.embedding <=> '[VECTOR]')) as similarity
FROM job_posts jp
JOIN job_embeddings je ON jp.id = je.job_post_id
ORDER BY similarity DESC
LIMIT 5;

Troubleshooting

  • No matches found?

    • Ensure you have run /analyze on your Job Posts (creates job_embeddings).
    • Ensure you have run /parse and /analyze (or just text extraction?) on CVs. Note: Embeddings for CVs are created during parsing/upload?
    • Correction: Embeddings for CVs are usually created during parsing or upload. Check DocumentEmbedding table.

  • pgvector error?

    • Ensure the extension is enabled: CREATE EXTENSION IF NOT EXISTS vector;
    • Ensure your Docker container has the pgvector image.