v1.1.0

MCP server that turns Claude into a full-stack personal trainer. 27 tools across nutrition, strength training, powerlifting, strongman, recovery, and mental performance coaching. All data stays local in SQLite.

Fixed

• pl_analyze_sbd_ratio raised ZeroDivisionError when squat + bench + deadlift totaled 0. Now returns an error message instead of crashing.
• training_get_session_history capped its SQL fetch at limit * 5 rows, assuming no more than 5 sets per session. A higher-volume session could silently push earlier sessions out of the result. Removed the cap.
• README claimed Python 3.10+; the package actually requires >=3.12 (matches the Docker image). Badge corrected.
• README's quick start pointed at pip install robo-trainer-mcp on PyPI, which was never published there. Quick start now documents installing from source or via the GHCR Docker image, which are the two channels that actually work.
• pyproject.toml still said version 0.1.0 despite v1.0.0 already being tagged and released. Synced to 1.1.0.

Added

• training_log_bodyweight / training_get_bodyweight_trend: log and trend bodyweight over time. Wilks and DOTS scoring both take bodyweight as an input, but there was previously no way to log or trend it.
• training_check_overreaching: flags overreaching risk for an exercise by comparing weekly training volume against weekly average RPE, and calling out rising-volume-plus-rising-RPE with no deload.

Tests

16 passing (was 12 before this release; 4 new tests cover the bug fixes and new features).

Full Changelog: v1.0.0...v1.1.0