v0.29.0

What's Changed

• feat(providers): llama.cpp (llama-server) as a first-class local backend (#161) by @artokun in #167
• feat(providers): Custom OpenAI-compatible endpoint as a first-class backend (#162) by @artokun in #170
• feat(orchestrator): inline media bytes in show_media for headless clients by @artokun in #171
• feat(tools): graph query — filter/traverse/aggregate over big workflows (#169) by @artokun in #179
• docs(design): graph-health heuristics in validate_workflow / analyze_workflow (RFC) by @artokun in #175
• docs(design): calculate — safe batch math evaluator with variables + seeded RNG (RFC) by @artokun in #176
• docs(design): get_comfyui_settings / set_comfyui_setting — ComfyUI user settings store (RFC) by @artokun in #174
• docs(design): path-jailed custom-node file/search/patch/git tools (RFC) by @artokun in #173
• docs(design): panel_auto_layout — one-shot canvas auto-arrange (RFC) by @artokun in #177
• feat(orchestrator): on-demand phone pairing — token-gated LAN/tunnel listener by @artokun in #180
• docs(design): panel_connect auto_match + slot diagnostics; dsl_to_workflow wiring warnings (RFC) by @artokun in #178

Full Changelog: v0.28.0...v0.29.0