AutoResearch: enriched store indexing missing spec fields (iteration, trend, variant ID)

[mrveiss]

Summary

Design spec section 2.1 requires the enriched per-experiment ChromaDB document to include:

• Iteration number within the session
• Prior results trend direction
• Prompt variant ID (when optimizer is active)

None of these fields appear in the current `_build_document` or `_build_metadata` in `store.py`. The enrichment added hyperparams, baseline comparison, and session context — but these three spec items were missed.

Also, `llm_service` in `LLMJudgeScorer` is typed as `Any` instead of the actual `LLMService` class — weak typing reduces IDE support.

Files

• `autobot-backend/services/autoresearch/store.py` (`_build_document`, `_build_metadata`)
• `autobot-backend/services/autoresearch/scorers.py` (line 162, `llm_service: Any`)

Origin

Discovered during code review of PRs #3203 and #3202

[github-actions]

PR #4210 (merged to Dev_new_gui) references this issue with a close keyword.

feat(autoresearch): enriched store indexing with missing spec fields

If this issue is fully resolved, close it manually. If work remains, no action is needed.