fix(providers): fall back to configured models when models endpoint fetch fails by octogonz · Pull Request #7530 · block/goose

octogonz · 2026-02-26T07:34:41Z

Summary

Custom providers with a predefined models list could fail in Desktop when Goose attempted to fetch models from the provider API and the endpoint was not implemented. This change falls back to the configured models list if the API request fails.

Changes

Store optional custom_models from custom provider config.
Attempt API fetch first.
On error, fall back to configured models (with debug log).
Preserve existing behavior when no configured models are provided.

Why

Supports OpenAI-/Anthropic-compatible proxies and gateways that do not expose a models endpoint.

Fixes #7427

CC @tltsaia

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d23ffce1f4

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

crates/goose/src/providers/openai.rs

…mented errors Auth failures (401/403), rate limits (429), and server errors (5xx) now propagate correctly instead of being hidden by fallback. Only 404 and connection failures trigger fallback to preserve diagnostics. Fixes P1 feedback from PR block#7530

Custom providers with models defined in JSON config would fail when switching models in the Desktop UI because the code always tried to fetch from the /v1/models API endpoint, which many proxies don't implement. Now, for custom providers that define a models list: - Try to fetch models from /v1/models API first - On any error, gracefully fallback to predefined models from config - Log a debug message when falling back Backwards compatible: providers without custom models must still succeed with the API call (unchanged behavior). Fixes GitHub issue block#7427. Signed-off-by: Pete Gonzalez <[email protected]>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 664cc3f2cc

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

crates/goose/src/providers/errors.rs

…mented errors Auth failures (401/403), rate limits (429), and server errors (5xx) now propagate correctly instead of being hidden by fallback. Only 404 and connection failures trigger fallback to preserve diagnostics. Fixes P1 feedback from PR block#7530 Signed-off-by: Pete Gonzalez <[email protected]>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4e38b52c7f

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

crates/goose/src/providers/errors.rs

Connection failures should not trigger fallback to custom models - they indicate the provider host is down, not that the models endpoint is missing. This preserves diagnostics for outages. Addresses second P1 feedback from PR block#7530. Signed-off-by: Pete Gonzalez <[email protected]>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a3e88965ec

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

crates/goose/src/providers/errors.rs

Using contains("404") could incorrectly match port numbers like localhost:4040 in connection error messages. The 404 status code is always formatted as "(404)" in error messages from map_http_error_to_provider_error. Addresses third P1 feedback from PR block#7530. Signed-off-by: Pete Gonzalez <[email protected]>

octogonz · 2026-02-26T22:48:17Z

@katzdave issue #7427 was closed by your PR #7484, however it didn't seem to address the problem where the desktop app displays an error message when trying to connect to a proxy service that does not implement the /v1/models endpoint.

I think we still need my PR #7530 to fix that part. Thanks!

DOsinga · 2026-03-10T14:32:51Z

thanks for diving onto this and sorry for the late reply

two things:

I think we are also using the models endpoint to check if the provider works at all during config. if it errors out anyway, maybe there is no difference, but something to think about
should we maybe always return the custom models and just override what the provider gives you? I imagine a scenario where a company wants to say, just use any of these three models and they can just create a custom provider for that

Signed-off-by: Douwe Osinga <[email protected]>

DOsinga

The fix is correct and tightly scoped — the fallback only activates when custom_models is explicitly configured, so normal OpenAI/Anthropic flows are completely unaffected. Codex comments were all addressed well. I cleaned up a few inline comments that restated what the code already said clearly, and rebased onto main.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 073cc65212

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-20T18:00:11Z

crates/goose/src/providers/errors.rs

+
+    pub fn is_endpoint_not_implemented(&self) -> bool {
+        match self {
+            ProviderError::RequestFailed(msg) => msg.contains("(404)"),


Match endpoint-not-implemented on status, not message text

ProviderError::is_endpoint_not_implemented() currently checks RequestFailed text with msg.contains("(404)"), which can misclassify non-404 failures whenever the upstream error payload happens to include that substring (for example, a 400/405 response message mentioning 404). In that case fetch_supported_models() in the OpenAI/Anthropic providers will incorrectly return custom_models and hide the real API error. Because RequestFailed embeds arbitrary server text, this detection should use structured status information instead of substring matching.

Useful? React with 👍 / 👎.

…ring matching Signed-off-by: Douwe Osinga <[email protected]>

DOsinga

Thanks for the fix and for addressing all the codex comments thoroughly — the fallback logic is tightly scoped and correct.

I've pushed two small follow-up commits to your branch:

Removed inline comments that restated what the code already said clearly, and collapsed the nested match/if into a cleaner guard pattern.
Addressed the copilot P2: replaced the msg.contains("(404)") substring check with a dedicated ProviderError::EndpointNotFound variant, emitted structurally from map_http_error_to_provider_error when the HTTP status is exactly 404. This means a 400/405 response whose body happens to contain (404) will no longer be mistaken for a missing endpoint. Added two test cases covering 404 with and without a payload that itself contains a 404 substring.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 39181a57b0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-03-20T19:10:25Z

crates/goose/src/providers/openai_compatible.rs

-        StatusCode::NOT_FOUND => {
-            ProviderError::RequestFailed(format!("Resource not found (404): {}", extract_message()))
-        }
+        StatusCode::NOT_FOUND => ProviderError::EndpointNotFound(extract_message()),


Keep inference 404s as RequestFailed for model hints

Mapping every 404 to EndpointNotFound here regresses the model-recommendation flow for failed inference calls. Agent::enhance_model_error in agents/reply_parts.rs only augments ProviderError::RequestFailed with “available models” suggestions, so a /chat/completions response like “model ... not found” now skips that path and surfaces a less actionable generic error. This is a user-visible regression from the previous RequestFailed mapping for 404 responses.

Useful? React with 👍 / 👎.

…stFailed for inference 404s The shared map_http_error_to_provider_error was changed to emit EndpointNotFound for all 404 responses, but enhance_model_error in reply_parts.rs only augments RequestFailed errors with available-model suggestions. A /chat/completions 404 (e.g. 'model X not found') would skip that path and surface a less actionable error. Fix: emit EndpointNotFound only from fetch_models_from_api (where we actually want the custom-models fallback), and keep the shared 404 mapper returning RequestFailed so inference errors still get the model-recommendation enhancement. Signed-off-by: Douwe Osinga <[email protected]>

DOsinga

Addressed the remaining Codex P2 about EndpointNotFound regressing the model-recommendation flow.

The fix in 6c756b50: instead of emitting EndpointNotFound from the shared map_http_error_to_provider_error (which is used for all HTTP calls including inference), the 404→EndpointNotFound conversion now only happens inside fetch_models_from_api on both OpenAI and Anthropic providers. The shared mapper keeps returning RequestFailed for 404s, so enhance_model_error in reply_parts.rs continues to match inference 404s and append the helpful "Available models" suggestion.

# Conflicts: # crates/goose/src/providers/openai.rs

Signed-off-by: Douwe Osinga <[email protected]>

* main: (337 commits) fix: replace panics with user-friendly errors in CLI session builder (#7901) fix: read GOOSE_CONTEXT_LIMIT from config.yaml, not just env vars (#7900) fix: deliver truncation notice as separate content block (#7899) fix: use platform-appropriate commands in developer extension instructions (#7898) fix: replace any with proper SVG types in icon components (#7873) chore: remove debug console.log statements, stale comments, and dead code (#8142) feat: Gemini OAuth provider (#8129) chore(deps): bump picomatch from 2.3.1 to 2.3.2 in /documentation (#8123) feat: show installed skills in UI (#7910) fix(deps): gate keyring platform features behind target-specific deps (#8039) chore(deps): bump yaml from 2.8.2 to 2.8.3 in /evals/open-model-gym/suite (#8124) fix: strip message wrapper in CLI session title generation (#7996) fix(providers): fall back to configured models when models endpoint fetch fails (#7530) chore(deps): bump brace-expansion from 5.0.3 to 5.0.5 in /evals/open-model-gym/suite (#8139) fix: prevent Ollama provider from hanging on tool-calling requests (#7723) fix: VMware Tanzu Platform provider - bug fixes, streaming, UI improvements (#8126) feat: allow GOOSE_CLI_SHOW_THINKING to be set in config.yaml (#8097) fix: GitHub Copilot auth fails to open browser in Desktop app (#6957) (#8019) fix(ci): produce .tar.gz archives for Zed ACP registry compatibility (#8054) feat: add GOOSE_SHOW_FULL_OUTPUT config to disable tool output truncation (#7919) ... # Conflicts: # crates/goose/src/providers/formats/openai.rs

…etch fails (block#7530) Signed-off-by: Pete Gonzalez <[email protected]> Signed-off-by: Douwe Osinga <[email protected]> Co-authored-by: Pete Gonzalez <[email protected]> Co-authored-by: Douwe Osinga <[email protected]> Signed-off-by: Cameron Yick <[email protected]>

octogonz · 2026-04-02T00:18:30Z

Thanks very much! 🚀

* main: (337 commits) fix: replace panics with user-friendly errors in CLI session builder (#7901) fix: read GOOSE_CONTEXT_LIMIT from config.yaml, not just env vars (#7900) fix: deliver truncation notice as separate content block (#7899) fix: use platform-appropriate commands in developer extension instructions (#7898) fix: replace any with proper SVG types in icon components (#7873) chore: remove debug console.log statements, stale comments, and dead code (#8142) feat: Gemini OAuth provider (#8129) chore(deps): bump picomatch from 2.3.1 to 2.3.2 in /documentation (#8123) feat: show installed skills in UI (#7910) fix(deps): gate keyring platform features behind target-specific deps (#8039) chore(deps): bump yaml from 2.8.2 to 2.8.3 in /evals/open-model-gym/suite (#8124) fix: strip message wrapper in CLI session title generation (#7996) fix(providers): fall back to configured models when models endpoint fetch fails (#7530) chore(deps): bump brace-expansion from 5.0.3 to 5.0.5 in /evals/open-model-gym/suite (#8139) fix: prevent Ollama provider from hanging on tool-calling requests (#7723) fix: VMware Tanzu Platform provider - bug fixes, streaming, UI improvements (#8126) feat: allow GOOSE_CLI_SHOW_THINKING to be set in config.yaml (#8097) fix: GitHub Copilot auth fails to open browser in Desktop app (#6957) (#8019) fix(ci): produce .tar.gz archives for Zed ACP registry compatibility (#8054) feat: add GOOSE_SHOW_FULL_OUTPUT config to disable tool output truncation (#7919) ... # Conflicts: # crates/goose/src/providers/formats/openai.rs

octogonz mentioned this pull request Feb 26, 2026

Cannot use Azure OpenAI deployment endpoints (no /models endpoint) #7427

Closed

octogonz force-pushed the main branch from d035c1d to d23ffce Compare February 26, 2026 07:38

chatgpt-codex-connector bot reviewed Feb 26, 2026

View reviewed changes

crates/goose/src/providers/openai.rs Outdated Show resolved Hide resolved

chatgpt-codex-connector bot reviewed Feb 26, 2026

View reviewed changes

crates/goose/src/providers/errors.rs Outdated Show resolved Hide resolved

octogonz force-pushed the main branch from 664cc3f to 4e38b52 Compare February 26, 2026 22:00

chatgpt-codex-connector bot reviewed Feb 26, 2026

View reviewed changes

crates/goose/src/providers/errors.rs Outdated Show resolved Hide resolved

chatgpt-codex-connector bot reviewed Feb 26, 2026

View reviewed changes

crates/goose/src/providers/errors.rs Outdated Show resolved Hide resolved

DOsinga self-assigned this Mar 9, 2026

Douwe Osinga added 2 commits March 20, 2026 13:52

chore: remove restatement comments from fetch_supported_models fallback

5a9936f

Signed-off-by: Douwe Osinga <[email protected]>

Merge remote-tracking branch 'origin/main' into pr-7530

073cc65

DOsinga approved these changes Mar 20, 2026

View reviewed changes

chatgpt-codex-connector bot reviewed Mar 20, 2026

View reviewed changes

fix(providers): use EndpointNotFound variant for 404 instead of subst…

39181a5

…ring matching Signed-off-by: Douwe Osinga <[email protected]>

DOsinga approved these changes Mar 20, 2026

View reviewed changes

chatgpt-codex-connector bot reviewed Mar 20, 2026

View reviewed changes

DOsinga reviewed Mar 22, 2026

View reviewed changes

Douwe Osinga added 5 commits March 26, 2026 13:00

Merge remote-tracking branch 'origin/main' into pr-7530

8a61304

# Conflicts: # crates/goose/src/providers/openai.rs

chore: update generated OpenAPI types after merge with main

0bf7e87

Signed-off-by: Douwe Osinga <[email protected]>

Merge remote-tracking branch 'origin/main' into pr-7530

97b271a

Merge remote-tracking branch 'origin/main' into pr-7530

55be883

chore: regenerate OpenAPI types with current codegen version

22703f2

DOsinga added this pull request to the merge queue Mar 26, 2026

Merged via the queue into block:main with commit 27e197e Mar 26, 2026
22 checks passed

github-actions bot mentioned this pull request Apr 2, 2026

chore(release): release version 1.30.0 #8261

Open

Conversation

octogonz commented Feb 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Why

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

octogonz commented Feb 26, 2026

Uh oh!

DOsinga commented Mar 10, 2026

Uh oh!

DOsinga left a comment

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

DOsinga left a comment

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector bot Mar 20, 2026

Choose a reason for hiding this comment

Uh oh!

DOsinga left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

octogonz commented Apr 2, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

octogonz commented Feb 26, 2026 •

edited

Loading