alex 9d5681b1ef feat(ai): implement agent-redesign plan with enhanced AI travel features ...

Phase 1 - Configuration Infrastructure (WS1):
- Add instance-level AI env vars (VOYAGE_AI_PROVIDER, VOYAGE_AI_MODEL, VOYAGE_AI_API_KEY)
- Implement fallback chain: user key → instance key → error
- Add UserAISettings model for per-user provider/model preferences
- Enhance provider catalog with instance_configured and user_configured flags
- Optimize provider catalog to avoid N+1 queries

Phase 1 - User Preference Learning (WS2):
- Add Travel Preferences tab to Settings page
- Improve preference formatting in system prompt with emoji headers
- Add multi-user preference aggregation for shared collections

Phase 2 - Day-Level Suggestions Modal (WS3):
- Create ItinerarySuggestionModal with 3-step flow (category → filters → results)
- Add AI suggestions button to itinerary Add dropdown
- Support restaurant, activity, event, and lodging categories
- Backend endpoint POST /api/chat/suggestions/day/ with context-aware prompts

Phase 3 - Collection-Level Chat Improvements (WS4):
- Inject collection context (destination, dates) into chat system prompt
- Add quick action buttons for common queries
- Add 'Add to itinerary' button on search_places results
- Update chat UI with travel-themed branding and improved tool result cards

Phase 3 - Web Search Capability (WS5):
- Add web_search agent tool using DuckDuckGo
- Support location_context parameter for biased results
- Handle rate limiting gracefully

Phase 4 - Extensibility Architecture (WS6):
- Implement decorator-based @agent_tool registry
- Convert existing tools to use decorators
- Add GET /api/chat/capabilities/ endpoint for tool discovery
- Refactor execute_tool() to use registry pattern

2026-03-08 23:53:14 +00:00

1.2 KiB

Raw Permalink Blame History

View Raw