feat: knowledge graph re-ranking for search results

[vitali87]

Task Description

Explore using the code knowledge graph to re-rank search results, improving relevance by incorporating structural relationships between code entities.

The idea is that when a user queries the codebase, the graph topology (call chains, imports, inheritance) can boost or demote results based on how closely related they are to the query context.

Acceptance Criteria

• Research feasibility of graph-based re-ranking
• Prototype a re-ranking step that uses graph proximity/relationships
• Compare retrieval quality with and without graph re-ranking
• Integrate into the existing query pipeline if results are positive

References

• Original discussion: Run this project as MCP Server #61 (comment by @qdrddr)
• Knowledge Graph Re-ranking notebook: https://github.com/Bavalpreet/MediumBlogs/blob/main/Knowledgegraph%20Reranking/Knowledge_graph_re_ranking.ipynb

Priority

• High
• Medium
• Low

Estimated Effort

• Small (< 2 hours)
• Medium (2-8 hours)
• Large (> 8 hours)