EdgeMind - Verifiable Onboard IR (initial skeleton)

This repository contains an initial Python skeleton for the EdgeMind IR (Intermediate Representation) and a small set of Pydantic models representing the canonical IR types used by the EnergiBridge and EdgeMind runtime.

What is included in this change:

• pyproject.toml with packaging metadata and dependency on pydantic
• specs/ir.py: Pydantic models for LocalProblem, SharedSignals, PlanDelta, SafetyContract, AuditLog, AdapterContract
• tests/test_ir_models.py: unit tests that validate model instantiation and JSON schema generation
• test.sh: convenience script to run tests and build

How to run tests and build (also invoked by the CI harness):

bash test.sh

Notes:

• This is a focused, well-scoped first chunk of the overall EdgeMind idea: it provides canonical, typed IR models and tests that the models behave and can produce JSON Schema. Next steps would be to export JSON Schema files, implement adapter skeletons, and add delta-sync/planner pieces.