Jimmy Sprint 1

[jmacd867]

Fixes #172, #173, #175

---

What was changed?

• inference/README.md — rewritten to document the complete ML workflow end-to-end (data collection → label generation → label validation → dataset preparation → model training → model testing), with purpose, inputs/outputs, and copy-pasteable commands for each step.
• inference/README_TRAINING.md — removed; its content was folded into the main inference/README.md.
• inference/requirements.txt — filled in missing packages (e.g. scikit-learn, joblib) with version ranges so the pipeline installs cleanly with one command.
• inference/.python-version, inference/pyproject.toml, inference/uv.lock — added to pin and lock the supported Python version and dependencies.
• inference/validate_labels.py — updated during requirements/workflow verification.
• inference/Models/bestModel.pkl, inference/Models/finalModel.pkl, inference/evaluation_metrics.json, inference/predictions_output.csv — retrained model artifacts and regenerated outputs.
• .gitignore — updated (adds .venv exclusion per Complete and Verify Python Requirements #173).
• docs/telemetry_schema.md — new schema document for pilot telemetry: every field the X-Plane plugin currently emits (name, type, units, source DataRef), the target-state schema (including fields not yet implemented), and explicit deviation formulas for altitude, heading, vertical speed, and airspeed.
• CLAUDE.md, docs/agents/issue-tracker.md, docs/agents/triage-labels.md, docs/agents/domain.md — added repo configuration for AI coding agent skills (GitHub as issue tracker, default triage labels, single-context domain docs).

---

Why was it changed?

• The inference/ folder had separate scripts for each ML pipeline stage, but no documented order or setup instructions, making it hard for a new contributor to run the pipeline (Verify and Document the complete ML Workflow #172).
• inference/requirements.txt was missing packages the pipeline actually imports, so a clean install failed and required undocumented manual steps (Complete and Verify Python Requirements #173).
• No agreed-upon schema existed for pilot telemetry ahead of the client's real data arriving. The client's performance metric depends on designated-vs-actual deviation for altitude, heading, vertical speed, and airspeed, but the plugin only emits actual values today — this needed to be documented as a contract and the gap flagged explicitly rather than assumed (Define Telemetry Data Schema & Performance Deviation Metrics #175).

---

How was it changed?

• Reviewed every script under inference/ and documented its purpose, execution order, and inputs/outputs directly in inference/README.md; verified the instructions in a fresh virtual environment.
• Reviewed all Python imports under inference/, added the missing third-party packages and version ranges to requirements.txt, and introduced pyproject.toml/uv.lock plus .python-version to lock the environment; confirmed validate_labels.py, prepare_data.py, train_model.py, and test_model.py all run from that clean install.
• Reviewed xplane_plugin/src/pilotdatasync-xp11.cpp, xplane_plugin/docs/key_datarefs.md, and inference/Data/data_logger.py to document the plugin's actual field names, types, and units in docs/telemetry_schema.md. Defined deviation = actual − designated for each of the four flight-dynamics metrics, including a wrapped angular-difference formula for heading to handle the 0°/360° boundary. Called out that no designated/target values and no pilot metadata (certification level, flight hours, rating) exist anywhere in the current pipeline.

---

Screenshots (if applicable):

• N/A — this PR is documentation, dependency/config, and ML pipeline verification; no UI or runtime output changes.