openai-session-semantic-routing-checklist.md

CE_AI_SESSION Semantic Routing Checklist

Goal

Migrate ContextServiceClient.semanticSearch to use the OpenAI session provider when CE_AI_PROVIDER=openai_session while preserving existing behavior for augment.

Plan (implemented)

• Add provider-resolution and provider instance wiring in ContextServiceClient.
• Add provider-scoped cache-key namespace (provider:query:topK).
• Route semanticSearch based on active provider (augment -> DirectContext, openai_session -> searchAndAsk).
• Add JSON-array parsing for openai_session search responses.
• Add strict result-path sanitization and unsafe-path filtering.
• Preserve fallback path for malformed provider output (keywordFallbackSearch).
• Keep existing semantic_search and codebase_retrieval tool contracts intact.
• Add targeted unit tests for openai provider response parsing/fallback/cache-keying.
• Preserve offline-mode guard behavior for openai_session in searchAndAsk.
• Preserve offline-mode behavior for augment in initialization by pre-invoking ensureInitialized() when needed.
• Optional hardening: move provider-parsing retry and telemetry into shared provider layer.
• Optional cleanup: reduce noisy console output in test environments.

Evidence

• npm test -- --runInBand tests/serviceClient.test.ts
• npm test -- --runInBand tests/tools/search.test.ts tests/tools/codebaseRetrieval.test.ts