AIRA-53: Analysis Service Core Implementation

[learn-design-develop]

👤 User Story

As the core processing layer, I need an analysis service to orchestrate the analysis workflow, integrate with LLM providers, and generate structured hypotheses.

🎯 Rationale

Core component for LLM-driven analysis pipeline.

✅ Acceptance Criteria

• Create aira/core/analysis_service.py with AnalysisService class
• Define prompt templates for incident and data context
• Call OpenAI provider with structured prompts
• Parse LLM responses into AnalysisResult model
• Cache results in Redis (analysis_cache_hits, TTL 1 hour)
• Confidence scoring based on data completeness and response quality
• Metrics: analysis_requests_total, analysis_duration_seconds, llm_api_calls_total, llm_response_time_seconds
• Error handling with circuit breaker, fallback to cache
• Unit & integration tests
• Honor LLM_TIMEOUT from configuration

📋 Metadata

• Status: MVP

• Category: Core Workflow

• Week: Week 5

• Complexity: Medium

• Critical Path: No

• Dependencies: AIRA-21

---

Original Ticket: #53
Phase 1 MVP Tracking Issue