Masive refactoring and migration to multiagent

[caviri]

Note

Migrates to a modular multi‑agent architecture with EPFL final assessment, JSON‑LD conversion, Infoscience tools, unified API responses with stats, and a 365‑day caching strategy.

• Architecture:
  • Introduces agents/ (repository, user, organization, org-enrichment, EPFL assessment) with prompts, retries/fallbacks, and tool wiring.
  • Adds analysis/ layer (Repository/User/Organization) orchestrating steps and tracking token/timing stats.
• API:
  • Rewrites endpoints to use analysis classes; returns APIOutput with APIStats (tokens, duration, status).
  • Adds JSON‑LD endpoints; consistent error handling; logging improvements.
• Data Models & Conversion:
  • Restructures data_models/ (type discriminators, validators); adds JSON‑LD converter and mappings; new EPFL assessment and Infoscience models.
• Context & Integrations:
  • Implements context/infoscience.py client and PydanticAI tools; enhances repository context extraction and git author parsing.
• Caching:
  • Sets all TTLs to 365 days; adds direct load/store helpers; updates cache namespace/exports and .env.
• Validation & Utils:
  • Adds URL validation and token counting; centralizes special‑token sanitization; tightens field validators.
• Parsers & Tooling:
  • Moves parsers to parsers/; updates imports; adds serve-dev-debug; upgrades deps (pydantic‑ai ≥1.0.15, google‑genai ≥1.31.0).
• Cleanup:
  • Removes legacy core/genai_model.py; migrates to new modules; updates examples/tests.

^{Written by Cursor Bugbot for commit 7a2f964. This will update automatically on new commits. Configure here.}