v0.0.7
What's Changed
- feat(imessage): plain-text replies (strip markdown) by @Spshulem in #42
- feat(mcp): add Airtable to the MCP catalog by @Spshulem in #43
- fix(budget): correct gpt-5.x pricing (nano was billed at flagship rates) by @Spshulem in #44
- fix(imessage): decode attributedBody correctly (Peri trigger was broken) by @Spshulem in #45
- fix(dashboard): keep selected MCP server across refresh (OAuth banner was hidden) by @Spshulem in #46
- feat(tasks): queue classification + task-list hygiene sweep by @Spshulem in #47
- fix(mcp): sanitize server name in tool names (a space broke ALL tool calls) by @Spshulem in #48
- feat(computer-use): real execution via a computer-use node by @Spshulem in #49
- feat(computer-use): downscale screenshots + scale click coords to logical points by @Spshulem in #50
- fix(computer-use): logical resolution via system_profiler (no TCC needed) by @Spshulem in #51
- feat(computer-use): feed screenshots to the model as images (it was clicking blind) by @Spshulem in #52
- fix(tools): cap tools advertised to the model (256 tools = OpenAI 500 on every call) by @Spshulem in #53
- feat(autopilot): gate on queued agent work (kills idle base-model spend) by @Spshulem in #54
- docs(skills): bundle operational skills (nodes, iMessage, MCP, cron, computer use) by @Spshulem in #55
- docs(skills): operational runbooks (deploy nodes, diagnose iMessage + tool errors) by @Spshulem in #56
- fix(mac): release build fails to compile on newer runner toolchain (breaks Sparkle updates) by @Spshulem in #57
Full Changelog: v0.0.6...v0.0.7