Project
vgrep
Description
The SearchEngine::new() function in src/core/search.rs lines 29-34 accepts a use_reranker parameter but completely ignores it. Users who download the reranker model (~400MB) and enable use_reranker = true in config get no benefit - the reranker is never used, but there's no warning.
Error Message
No error - silent feature non-implementation.
Debug Logs
System Information
- Bounty Version: 0.1.0
- OS: Ubuntu 24.04 LTS
- Rust: 1.75+
Screenshots
No response
Steps to Reproduce
- Download both models:
vgrep models download
- Enable reranker:
vgrep config set use-reranker true
- Verify config:
vgrep config show - shows "Reranker: enabled"
- Perform searches and observe no quality improvement
- Check code to confirm reranker is never invoked
Expected Behavior
Either:
- Implement the reranker functionality as advertised, OR
- Remove reranker configuration options and document it as "coming soon", OR
- Show a warning when
use_reranker = true that the feature is not yet implemented
Actual Behavior
- Users can download a 400MB reranker model
- Users can enable reranker in config
- Status shows reranker as "enabled"
- Reranker is never actually used
- No warning or indication that the feature is non-functional
Additional Context
The comment mentions "requires a separate backend which conflicts with the embedding engine's backend". This suggests a technical limitation with llama-cpp-rs that should be documented and tracked as a known issue.
Users relying on reranking for better search quality are silently getting degraded results.
Project
vgrep
Description
The
SearchEngine::new()function insrc/core/search.rslines 29-34 accepts ause_rerankerparameter but completely ignores it. Users who download the reranker model (~400MB) and enableuse_reranker = truein config get no benefit - the reranker is never used, but there's no warning.Error Message
Debug Logs
System Information
Screenshots
No response
Steps to Reproduce
vgrep models downloadvgrep config set use-reranker truevgrep config show- shows "Reranker: enabled"Expected Behavior
Either:
use_reranker = truethat the feature is not yet implementedActual Behavior
Additional Context
The comment mentions "requires a separate backend which conflicts with the embedding engine's backend". This suggests a technical limitation with llama-cpp-rs that should be documented and tracked as a known issue.
Users relying on reranking for better search quality are silently getting degraded results.