Agents that hit a Hyperia defect (e.g. the TTS long-text crash) tend to misdiagnose it — an OpenCode agent concluded "TTS is genuinely down" when it was an internal synth crash. Give them a first-class way to report failures so they become triageable fixes instead of dead ends.
Shipped (v0.16.8):
report_bug MCP tool → POST /api/bug → appends to ~/.hyperia/logs/bugs.jsonl (reporter identity, title, details, failing tool, exact error, repro context, version). Open to anonymous callers (we want reports).
bug_log tool + GET /api/bug/log for triage.
- Registered in the telemetry_diag door.
- Error hint on the TTS failure path instructs the agent to
report_bug (and not to claim "TTS is down").
Follow-ups: surface new reports to the human live (toast/sticky/dashboard); add hints to more error sites (SoftWall, internal errors) generically; optional promote-to-GitHub-issue helper for triage.
Agents that hit a Hyperia defect (e.g. the TTS long-text crash) tend to misdiagnose it — an OpenCode agent concluded "TTS is genuinely down" when it was an internal synth crash. Give them a first-class way to report failures so they become triageable fixes instead of dead ends.
Shipped (v0.16.8):
report_bugMCP tool →POST /api/bug→ appends to~/.hyperia/logs/bugs.jsonl(reporter identity, title, details, failing tool, exact error, repro context, version). Open to anonymous callers (we want reports).bug_logtool +GET /api/bug/logfor triage.report_bug(and not to claim "TTS is down").Follow-ups: surface new reports to the human live (toast/sticky/dashboard); add hints to more error sites (SoftWall, internal errors) generically; optional promote-to-GitHub-issue helper for triage.