- Enhance LlmClient to handle malformed URLs and HTML error responses
- Improve logging in report-generator-service worker
- Update frontend API routes and hooks for analysis
- Update various service configurations and persistence logic
- Backend: Add template_id to DataRequest and AnalysisResult metadata
- Frontend: Add TemplateSelector to Report page
- Frontend: Refactor ReportPage to use AnalysisTemplateSets for dynamic tabs
- Frontend: Strict mode for report rendering based on template_id
- Cleanup: Remove legacy analysis config hooks
- Enable dynamic config loading from DB to respect enabled status
- Implement Yahoo Finance Crumb authentication flow
- Add browser-like User-Agent and cookie store
- Map .SH symbols to .SS for Yahoo compatibility
- Add Google DNS to docker-compose to fix container resolution issues
- Fix incorrect parameter injection in MCP client query method
- Correct tool name from OVERVIEW to COMPANY_OVERVIEW
- Add symbol conversion support (e.g., .SH -> .SS)
- Implement fault tolerance for empty data responses to prevent panics
- Add workaround for non-standard JSON responses (single quotes) in GLOBAL_QUOTE
- Add debug_mcp utility for tool inspection
- Tushare: Added /test endpoint to verify API token validity by fetching a small dataset.
- AlphaVantage: Implemented custom HTTP transport to handle MCP server's 400 Bad Request response on SSE endpoint gracefully (degrading to POST-only mode).
- AlphaVantage: Added /test endpoint using `list_tools` to verify MCP connection.
- AlphaVantage: Updated configuration polling to support dynamic API URLs.
This commit introduces a comprehensive, template-based analysis orchestration system, refactoring the entire analysis generation workflow from the ground up.
Key Changes:
1. **Backend Architecture (`report-generator-service`):**
* Replaced the naive analysis workflow with a robust orchestrator based on a Directed Acyclic Graph (DAG) of module dependencies.
* Implemented a full topological sort (`petgraph`) to determine the correct execution order and detect circular dependencies.
2. **Data Models (`common-contracts`, `data-persistence-service`):**
* Introduced the concept of `AnalysisTemplateSets` to allow for multiple, independent, and configurable analysis workflows.
* Created a new `analysis_results` table to persist the output of each module for every analysis run, ensuring traceability.
* Implemented a file-free data seeding mechanism to populate default analysis templates on service startup.
3. **API Layer (`api-gateway`):**
* Added a new asynchronous endpoint (`POST /analysis-requests/{symbol}`) to trigger analysis workflows via NATS messages.
* Updated all configuration endpoints to support the new `AnalysisTemplateSets` model.
4. **Frontend UI (`/config`, `/query`):**
* Completely refactored the "Analysis Config" page into a two-level management UI for "Template Sets" and the "Modules" within them, supporting full CRUD operations.
* Updated the "Query" page to allow users to select which analysis template to use when generating a report.
This new architecture provides a powerful, flexible, and robust foundation for all future development of our intelligent analysis capabilities.
- Fix Dockerfile stub builds; compile full sources (no empty binaries)
- Add ca-certificates and curl in runtime images for TLS/healthchecks
- Enable RUST_LOG and RUST_BACKTRACE for all providers
- Add HTTP /health healthchecks in docker-compose for ports 8000-8004
- Standardize Tushare /health to structured HealthStatus JSON
- Enforce strict config validation (FINNHUB_API_KEY, TUSHARE_API_TOKEN)
- Map provider API keys via .env in docker-compose
- Log provider_services at API Gateway startup for diagnostics
Outcome: provider containers no longer exit silently; missing keys fail fast with explicit errors; health and logs are consistent across modules.
