feat: include index_id in context grounding tool output for tracing

[cfauchere]

Summary

• Capture index_id from the SDK's unified_search result on ContextGroundingRetriever (both sync and async paths)
• Include index_id in the semantic search tool return value so the trace instrumentor in uipath-agents-python can write it to the contextGroundingTool span

Context

The contextGroundingTool trace span has an index_id field defined in uipath-agents-python (ContextGroundingToolSpanAttributes.index_id), and set_context_grounding_results already looks for content.get("index_id") in the tool output. But it was never populated because:

1. The SDK resolves index_name → index_id internally in unified_search but doesn't return it
2. The retriever discards it
3. The tool output only contains {"documents": [...]}

Without index_id in the trace, debugging Context Grounding issues (like empty search results) requires manual lookup — you can't go from the trace to the CG service logs.

Dependencies

Requires a corresponding change in uipath-python to expose index_id on UnifiedQueryResult. Until then, resolved_index_id will be None (the code uses getattr with a fallback to avoid breaking).

Test plan

• Verify resolved_index_id is populated after unified_search when SDK exposes index_id
• Verify index_id appears in the contextGroundingTool trace span attributes
• Verify no regression when SDK does not yet expose index_id (getattr fallback returns None)

🤖 Generated with Claude Code