perf: avoid reloading skill graph during recommendation scoring

[DevChinmaysamal17]

Summary

This PR improves the performance of the recommendation pipeline by avoiding repeated loading of the skill graph during project scoring.

Previously, score_single_project() loaded skill_graph.json for every project being evaluated, resulting in unnecessary filesystem I/O and repeated JSON parsing. This change loads the skill graph once in get_recommendations() and reuses the same instance throughout the recommendation process.

Related Issue

N/A

Type of Change

• Refactor — restructures code without changing behaviour

What Was Changed

File	Change made
src/utils/recommender.py	Load the skill graph once per recommendation request and pass it to score_single_project() instead of reloading it for every project.

How to Test This PR

1. Clone the repository and checkout this branch.
2. Create and activate a virtual environment.
3. Install dependencies:

   pip install -r requirements.txt

4. Run the test suite:

   python -m pytest

5. Verify that all tests pass and the recommendation functionality behaves as before.

Test Results

278 passed, 3 skipped in 5.17s

Notes for Reviewer

This change is intentionally limited to a performance optimization. It does not modify the recommendation algorithm or the API response. The only change is reusing a single loaded skill graph during a recommendation request to eliminate redundant file reads.

[vercel]

@DevChinmaysamal17 is attempting to deploy a commit to the komalsony234-1530's projects Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

Thank you for submitting your first pull request to DevPath.

Before review:

• Complete the PR template fully
• Ensure all tests pass
• Link your PR to an issue
• Keep changes scoped to the issue

A maintainer will review your contribution soon.

[komalharshita]

Looks good! This PR provides a clean performance optimization by loading the skill graph once per recommendation request instead of reloading it for every project evaluation. The change reduces unnecessary file I/O while preserving the existing recommendation logic and API behavior. The implementation is well-scoped, and the included test results indicate no regressions. I don't see any blocking issues.

Approved for merge. ✅