Test coverage: Ai::OpenAiService — summarization, vote extraction, and error handling

[AndreRobitaille]

Summary

Ai::OpenAiService has grown substantial test coverage since this issue was filed. Several method groups remain untested.

What's already covered

• Briefing pipeline — 5 tests in open_ai_service_briefing_test.rb (analyze_topic_briefing, render_topic_briefing, generate_briefing_interim, parse error handling)
• Topic extraction prompts — 2 tests in open_ai_service_extract_topics_test.rb (category name blocking, re_extract_item_topics)
• Description generation — 4 tests in open_ai_service_generate_description_test.rb (activity-based prompt, concept-based prompt, empty/nil response handling)
• Committee context — 3 tests in committee_context_test.rb (active committees, empty set, committee_type inclusion)
• Committee member extraction — 3 tests in extract_committee_members_test.rb (request/response, text truncation, JSON keyword in prompt)

Methods still needing tests

Summarization (meeting-level)

• summarize_packet_with_citations() — Two-pass summarization with JSON planning
• summarize_packet() — Packet analysis
• summarize_minutes() — Minutes recap
• analyze_meeting_content() — Meeting content analysis pass
• render_meeting_summary() — Meeting summary rendering pass

Topic analysis (per-meeting)

• analyze_topic_summary() — Governance-compliant topic analysis
• render_topic_summary() — Markdown rendering from analysis JSON

Vote extraction

• extract_votes() — Motion and vote extraction from minutes

Triage

• triage_topics() — Topic merging, approval, blocking recommendations

Catch-all refinement

• refine_catchall_topic() — Refines broad ordinance topics into specific concerns

Infrastructure

• Error handling and retries across all methods
• Provider selection (OpenAI vs Gemini)
• gemini_generate() path

Approach

• Stub API responses with WebMock or mock objects
• Test prompt construction (verify governance elements: citations, anti-speculation, confidence scoring)
• Test response parsing (JSON extraction from various formats)
• Test error handling (API timeouts, malformed responses)