[Bug]: GENERATION_FAILED (fp: f97ed9db)

[davevr]

Description

Just entered a prompt and tried to start a project. Got the error immediately.

Steps to reproduce

1. Click Start New Design
2. Type a simple prompt

Expected behavior

Prompt is executed

Actual behavior

Error invoking remote method 'codesign:v1:generate': CodesignError: 400 invalid request: unsupported role ROLE_UNSPECIFIED — user notes: Entered a prompt - failed.

open-codesign version

0.1.4

Platform

macOS

OS version

25.4.0

Model provider (if relevant)

Other

Error code

GENERATION_FAILED

Diagnostics bundle

Bundle saved locally at ~/Downloads/open-codesign-diagnostics-2026-05-08T17-02-08.zip. Attach it to this issue after submitting.

Logs

tools: 0
}
[2026-05-09 01:01:18.179] [info]   (main:ipc) agent.end { runId: 'mox5wc2b-p6ipuy', generationId: 'mox5wc2b-p6ipuy' }
[2026-05-09 01:01:18.179] [error]  (main:ipc) [generate] step=send_request.fail {
  runId: 'mox5wc2b-p6ipuy',
  generationId: 'mox5wc2b-p6ipuy',
  provider: 'custom-kimi-ry0v',
  modelId: 'kimi-k2.6',
  ms: 1443,
  stopReason: 'error'
}
[2026-05-09 01:01:18.180] [error]  (main:ipc) generate.fail {
  runId: 'mox5wc2b-p6ipuy',
  generationId: 'mox5wc2b-p6ipuy',
  ms: 1448,
  provider: 'custom-kimi-ry0v',
  modelId: 'kimi-k2.6',
  baseUrl: '[url omitted]',
  status: 400,
  message: '400 invalid request: unsupported role ROLE_UNSPECIFIED',
  code: 'PROVIDER_ERROR'
}
[2026-05-09 01:01:18.185] [error]  (renderer:console) [store] applyGenerateError {"generationId":"mox5wc2b-p6ipuy","designId":"fa3555d9-9cbc-4ebf-be0e-36e9acacf59d","message":"Error invoking remote method 'codesign:v1:generate': CodesignError: 400 invalid request: unsupported role ROLE_UNSPECIFIED"} {}
[2026-05-09 01:01:18.187] [info]   (chat-messages-ipc) chat.append {
  designId: 'fa3555d9-9cbc-4ebf-be0e-36e9acacf59d',
  seq: 1,
  kind: 'error'
}
[2026-05-09 01:01:23.486] [info]   (main:ipc) [title] step=send_request.ok { runId: '0db5e82f-fc58-4dc4-8039-4d9910b454cc', ms: 6778 }
[2026-05-09 01:01:23.488] [error]  (main:ipc) [title] step=send_request.fail {
  runId: '0db5e82f-fc58-4dc4-8039-4d9910b454cc',
  ms: 6780,
  errorClass: 'CodesignError'
}
[2026-05-09 01:01:23.489] [error]  (main:ipc) [title] generate-title.fail {
  runId: '0db5e82f-fc58-4dc4-8039-4d9910b454cc',
  provider: 'custom-kimi-ry0v',
  modelId: 'kimi-k2.6',
  baseUrl: '[url omitted]',
  message: 'Model returned empty title',
  code: 'PROVIDER_ERROR'
}
[2026-05-09 01:01:23.491] [warn]   (renderer:store) [title] generateTitle failed, using prompt fallback {
  designId: 'fa3555d9-9cbc-4ebf-be0e-36e9acacf59d',
  message: "Error invoking remote method 'codesign:v1:generate-title': CodesignError: Model returned empty title"
}
[2026-05-09 01:01:23.493] [info]   (snapshots-ipc) design.renamed {
  id: 'fa3555d9-9cbc-4ebf-be0e-36e9acacf59d',
  name: 'My son is a poet, song writer, and music…'
}

[github-actions]

Thanks for the detailed report, Dave! This error ROLE_UNSPECIFIED typically means the model API received a message with an empty or unsupported role field. Since you're using a custom provider (custom-kimi-ry0v) with kimi-k2.6, the issue is likely in how the message roles are mapped in your provider configuration or in the pi-ai adapter.

Possible causes:

1. The custom provider configuration might be missing a role mapping (e.g., system → developer or assistant → assistant).
2. The model endpoint may expect a different set of roles than what pi-ai sends by default.
3. A message in the conversation might have an empty role field due to a parsing issue.

Next steps to help us triage faster:

1. Attach the diagnostics bundle you saved (open-codesign-diagnostics-2026-05-08T17-02-08.zip) – it contains redacted config and recent errors that would show the exact request payload.
2. Share the custom provider configuration (redacted base URL and any role mappings you set). You can paste it here after removing your API key.
3. Does this happen with every prompt, or just some? If you try a very short prompt (e.g., "Hello"), does it still fail?
4. Have you tested other models (e.g., kimi-k2.5) on the same provider? That would help narrow down whether it's model-specific.

We'll look into the provider adapter code to see if there's a known role‑assignment issue with Kimi endpoints. In the meantime, double‑check that your base URL points to the correct chat completions endpoint and that the model ID matches exactly what the provider expects.

---

Open-CoDesign Bot

[davevr]

Model config:

[davevr]

open-codesign-diagnostics-2026-05-09T08-51-35.zip

[davevr]

BTW, this works if I use OpenAI instead of moonshot.