Repository navigation
[Bug]: GENERATION_FAILED (fp: 7d688101) #414
Description
Description
Trying to use it with a local llama.cpp instance.
Steps to reproduce
I added my local llama.cpp as a custom provider, then went to start prompting for a design to be created.
Expected behavior
It should not have errored.
Actual behavior
Error invoking remote method 'codesign:v1:generate': CodesignError: preferences.json is invalid: unsupported field "proxyUrl"
open-codesign version
0.2.0
Platform
Windows
OS version
Windows 11 (10.0.26200)
Model provider (if relevant)
Other
Error code
GENERATION_FAILED
Diagnostics bundle
Bundle saved locally at D:\Downloads\open-codesign-diagnostics-2026-08-15T04-17-28.zip. Attach it to this issue after submitting.
Logs
[2026-08-15 12:16:20.659] [warn] (renderer:console) [i18n] missing translation key "settings.providers.custom.allowPrivateNetwork" in namespace "translation" for locale "en" {}
[2026-08-15 12:16:20.771] [warn] (renderer:console) [i18n] missing translation key "settings.providers.custom.allowPrivateNetwork" in namespace "translation" for locale "en" {}
[2026-08-15 12:16:22.709] [warn] (renderer:console) [i18n] missing translation key "settings.providers.custom.allowPrivateNetwork" in namespace "translation" for locale "en" {}
[2026-08-15 12:16:23.325] [warn] (renderer:console) [i18n] missing translation key "settings.providers.custom.allowPrivateNetwork" in namespace "translation" for locale "en" {}
[2026-08-15 12:16:24.523] [warn] (renderer:console) [i18n] missing translation key "settings.providers.custom.allowPrivateNetwork" in namespace "translation" for locale "en" {}
[2026-08-15 12:16:45.888] [info] (main:ipc) [generate] step=load_config { runId: 'mstv6hx2-fhgdta', generationId: 'mstv6hx2-fhgdta' }
[2026-08-15 12:16:45.889] [info] (main:ipc) [generate] step=load_config.ok {
runId: 'mstv6hx2-fhgdta',
generationId: 'mstv6hx2-fhgdta',
ms: 0,
hasApiKey: true,
baseUrl: '[url omitted]'
}
[2026-08-15 12:16:45.889] [info] (main:ipc) [generate] step=validate_provider {
runId: 'mstv6hx2-fhgdta',
generationId: 'mstv6hx2-fhgdta',
provider: 'custom-llama-cpp-2hfj',
modelId: 'gemma-4-26b'
}
[2026-08-15 12:16:45.889] [info] (main:ipc) [generate] step=validate_provider.ok {
runId: 'mstv6hx2-fhgdta',
generationId: 'mstv6hx2-fhgdta',
provider: 'custom-llama-cpp-2hfj'
}
[2026-08-15 12:16:45.891] [error] (renderer:console) [store] applyGenerateError {"generationId":"mstv6hx2-fhgdta","designId":"529fa649-daf1-43b3-b96b-da3485dc16b5","message":"Error invoking remote method 'codesign:v1:generate': CodesignError: preferences.json is invalid: unsupported field \"proxyUrl\""} {}
[2026-08-15 12:17:27.186] [warn] (main:ipc) provider.error.final {
runId: '8cc5c016-6fe2-4a31-a42e-71d97e5762b1',
upstream_provider: 'custom-llama-cpp-2hfj',
upstream_code: 'PROVIDER_ERROR',
upstream_message: 'Provider stopped before completion because the response hit the token limit',
retry_count: 0,
original_error_name: 'CodesignError'
}
[2026-08-15 12:17:27.187] [error] (main:ipc) [title] step=send_request.fail {
runId: '8cc5c016-6fe2-4a31-a42e-71d97e5762b1',
ms: 159377,
errorClass: 'CodesignError'
}
[2026-08-15 12:17:27.187] [error] (main:ipc) [title] generate-title.fail {
runId: '8cc5c016-6fe2-4a31-a42e-71d97e5762b1',
provider: 'custom-llama-cpp-2hfj',
modelId: 'qwen3.6-35b-a3b',
baseUrl: '[url omitted]',
message: 'Provider stopped before completion because the response hit the token limit',
code: 'PROVIDER_ERROR'
}
[2026-08-15 12:17:27.191] [warn] (renderer:store) [title] generateTitle failed, using prompt fallback {
designId: '529fa649-daf1-43b3-b96b-da3485dc16b5',
message: "Error invoking remote method 'codesign:v1:generate-title': CodesignError: Provider stopped before completion because the response hit the token limit"
}
Activity
github-actions commentedon Aug 15, 2026
Thanks for the detailed report. The key error is from validation of your local config:
CodesignError: preferences.json is invalid: unsupported field "proxyUrl"The error means the
preferences.jsonfile that stores app preferences contains aproxyUrlkey on your custom provider (custom-llama-cpp-2hfj), but the preferences schema in v0.2.0 does not accept that field. The generation IPC (codesign:v1:generate) validates preferences before it sends the request, so generation fails before llama.cpp is even called.Workaround for now:
preferences.json(the file referenced in the error).custom-llama-cpp-2hfjprovider entry and remove the"proxyUrl": ...line.For a local llama.cpp instance you normally don't need a proxy; keep
baseUrlset directly to your llama.cpp server (e.g.http://127.0.0.1:8080/v1) and leave the API key blank or set it to whatever your server expects.The missing
settings.providers.custom.allowPrivateNetworktranslation strings in the log look like a separate cosmetic i18n gap, but they may be related if that toggle is what writes theproxyUrlfield. I wasn't able to verify the exact writing path from the public checkout, so it's worth confirming whether the field appears after you toggle that option.Need more info (if it still fails after removing
proxyUrl):preferences.json(mask the API key and base URL).open-codesign-diagnostics-2026-08-15T04-17-28.zip.Separately, the later
Provider stopped before completion because the response hit the token limiterror is a different failure — that happens when llama.cpp stops a response mid-stream. If you still see that after the config issue is fixed, it usually means the server's max tokens/context is too small for the prompt; raisingn_predict(or context size) on the llama.cpp side should help.Thanks again for the report — happy to dig further once we know whether removing
proxyUrlresolves it.Open-CoDesign Bot