enhancement: Make result fusion weights configurable via env vars

[QuantumByte-01]

Problem

In fuse_results (agents.py), vector and KS search results are merged with hardcoded weights:

"final_score": res.get("similarity", 0) * 0.6   # vector
"final_score": res.get("_score", 0) * 0.4        # KS

No explanation for the 0.6/0.4 split. These cannot be tuned without code changes.

Fix

Make them env-var configurable:

VECTOR_WEIGHT = float(os.getenv("FUSION_VECTOR_WEIGHT", "0.6"))
KS_WEIGHT = float(os.getenv("FUSION_KS_WEIGHT", "0.4"))

Add to .env.template with a comment explaining the tradeoff.

[Nabhye]

Hi, I’d like to work on this issue.

My understanding:
Currently, the fusion weights (0.6 and 0.4) are hardcoded in fuse_results, and the goal is to make them configurable using environment variables with sensible defaults.

Proposed approach:

• Replace hardcoded values with os.getenv-based variables
• Add defaults (0.6 / 0.4)
• Update .env.template with documentation

Please let me know if this approach looks good.

[QuantumByte-01]

Approach looks good! Assigning to you now.

[QuantumByte-01]

Approach looks good! Please proceed with the implementation:

• Replace hardcoded 0.6/0.4 with env vars (FUSION_VECTOR_WEIGHT, FUSION_KS_WEIGHT)
• Use sensible defaults
• Update .env.template with documentation

Note: PR #93 will implement RRF which replaces this weighting altogether, but this env-var approach is still valid for the current codebase.

[Nabhye]

Thank you for assigning this!

I’ll start working on implementing the environment variable configuration, sensible defaults, and updating the .env.template as suggested.

I’ll keep you updated with progress and open a PR soon.