Skip to content

GEMINI.md

Latest commit

 

History

History
19 lines (15 loc) · 1.56 KB

GEMINI.md

File metadata and controls

19 lines (15 loc) · 1.56 KB
Raw

AI Guidance & Constraints

This file outlines the foundational mandates and engineering standards used to guide AI agents (like Gemini CLI) during the development of BugFlow.

Foundational Mandates

  • Pragmatic Minimalism: Prioritize simplicity over abstraction. Avoid unnecessary layers, files, or patterns.
  • Contextual Precedence: Instructions in this file and proj details.txt take absolute precedence over general defaults.
  • Technical Integrity: All changes must include validation logic. Bug fixes must be empirically reproduced before being patched.

Engineering Standards

  • Correctness: Status transitions must be strictly enforced (OPEN → IN_PROGRESS → RESOLVED → CLOSED).
  • Interface Safety: Use Pydantic for strict input validation. AI outputs must be parsed and validated against expected schemas with intelligent fallbacks.
  • Observability: Failures in external integrations (like LLMs) must be visible via logs and handled gracefully with user-facing fallbacks.
  • Change Resilience: Logic is encapsulated in services.py to ensure that adding features (like a new database or AI model) has minimal impact on the rest of the system.

AI Usage Rules

  • Temperature Control: LLM calls must use a low temperature (0.2) to ensure deterministic triaging.
  • Validation Layer: AI-generated content is treated as untrusted. It must be validated against AITriageResponse schemas.
  • Smart Fallback: When the AI fails or provides invalid data, the system must utilize a keyword-based heuristic layer to maintain functional correctness.