Fix sessions_spawn Google/Gemini model routing by Octane0411 · Pull Request #36154 · openclaw/openclaw

Octane0411 · 2026-03-05T13:04:18Z

Fixes #36134 by resolving sessions_spawn model routing for bare Gemini model names and surfacing unavailable model overrides.

Changes:

Update subagent model resolution to infer provider for unqualified model names in spawn flow (including Google Gemini patterns) and normalize configured/bare overrides before patch.
Keep bare configured/per-agent defaults consistent with spawn model inference so Gemini values are normalized to google/gemini-style provider refs.
Return modelApplied: false on sessions.patch model patch failures so unavailable overrides are not reported as successfully applied.

Tests:

pnpm test src/agents/openclaw-tools.subagents.sessions-spawn.model.test.ts
pnpm test src/agents/model-selection.test.ts

greptile-apps · 2026-03-05T13:08:48Z

Greptile Summary

This PR fixes Gemini/Google model routing in the sessions_spawn flow by introducing a normalizeSubagentSpawnModelSelection helper that maps bare model names (e.g. gemini-2.5-flash) to fully-qualified provider/model refs (e.g. google/gemini-2.5-flash), and ensures all three model sources (runtime override, per-agent config, global default) are normalized before being used in session patches. It also correctly sets modelApplied: false when a model patch fails.

Key changes:

New inferProviderFromModelHint helper that first checks configured models for a unique provider match, then falls back to pattern-based detection for Gemini model names (gemini / gemini-* → google).
New normalizeSubagentSpawnModelSelection helper normalizes bare model names and provider-prefixed refs to canonical provider/model form.
resolveSubagentSpawnModelSelection now normalizes all three model sources before priority selection, so bare Gemini names are correctly routed regardless of which source supplies the model.
subagent-spawn.ts: modelApplied: false is now included in error returns on model patch failure.
Tests updated to assert the fully-qualified provider/model format and modelApplied: false on failure, with a new test covering bare Gemini model routing.

Issues found:

The ?? "" fallback in resolvedModelOverride creates a subtle edge case: when params.modelOverride is whitespace-only, the empty string short-circuits the ?? fallback chain instead of continuing to configured/primary model fallbacks.
resolveAgentModelPrimaryValue is called redundantly twice when computing normalizedPrimaryModel.

Confidence Score: 2/5

Core Gemini routing fix and modelApplied behavior are sound, but a subtle fallback-chain regression and redundant function call should be addressed before merging.
The Gemini model routing normalization and modelApplied: false error surfacing are correct and well-tested. However, the edge case where whitespace-only model overrides produce empty strings that short-circuit the ?? fallback chain is a real functional issue. Additionally, resolveAgentModelPrimaryValue is called redundantly. These should be fixed to prevent potential bugs and improve code quality. The test coverage for the happy path is strong, but edge cases around empty/whitespace model inputs are not covered.
src/agents/model-selection.ts requires fixes to the fallback logic and the redundant function call before merging.

_{Last reviewed commit: c91556c}

greptile-apps · 2026-03-05T13:08:52Z

src/agents/model-selection.ts

+  const resolvedModelOverride = params.modelOverride
+    ? normalizeSubagentSpawnModelSelection({
+        cfg: params.cfg,
+        model: normalizeModelSelection(params.modelOverride) ?? "",
+        defaultProvider: runtimeDefault.provider,
+      })
+    : undefined;


Empty string short-circuits the ?? fallback chain. When params.modelOverride is whitespace-only (e.g. " "), normalizeModelSelection returns undefined, then ?? "" converts it to "". Since normalizeSubagentSpawnModelSelection returns "" for empty input, resolvedModelOverride becomes "" — a truthy-but-empty string that bypasses the ?? operator's fallback logic (which only skips null/undefined). This causes the function to return "" instead of falling through to normalizedConfiguredModel or normalizedPrimaryModel.

To fix, check the normalizeModelSelection result before passing to normalizeSubagentSpawnModelSelection:

Suggested change

const resolvedModelOverride = params.modelOverride

? normalizeSubagentSpawnModelSelection({

cfg: params.cfg,

model: normalizeModelSelection(params.modelOverride) ?? "",

defaultProvider: runtimeDefault.provider,

})

: undefined;

const resolvedModelOverride = (() => {

const m = normalizeModelSelection(params.modelOverride);

if (!m) return undefined;

return normalizeSubagentSpawnModelSelection({

cfg: params.cfg,

model: m,

defaultProvider: runtimeDefault.provider,

});

})();

Prompt To Fix With AI

This is a comment left during a code review. Path: src/agents/model-selection.ts Line: 417-423 Comment: Empty string short-circuits the `??` fallback chain. When `params.modelOverride` is whitespace-only (e.g. `" "`), `normalizeModelSelection` returns `undefined`, then `?? ""` converts it to `""`. Since `normalizeSubagentSpawnModelSelection` returns `""` for empty input, `resolvedModelOverride` becomes `""` — a truthy-but-empty string that bypasses the `??` operator's fallback logic (which only skips `null`/`undefined`). This causes the function to return `""` instead of falling through to `normalizedConfiguredModel` or `normalizedPrimaryModel`. To fix, check the `normalizeModelSelection` result before passing to `normalizeSubagentSpawnModelSelection`: ```suggestion const resolvedModelOverride = (() => { const m = normalizeModelSelection(params.modelOverride); if (!m) return undefined; return normalizeSubagentSpawnModelSelection({ cfg: params.cfg, model: m, defaultProvider: runtimeDefault.provider, }); })(); ``` How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-03-05T13:08:54Z

src/agents/model-selection.ts

+  const normalizedPrimaryModel = resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model)
+    ? normalizeSubagentSpawnModelSelection({
+        cfg: params.cfg,
+        model:
+          normalizeModelSelection(
+            resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model),
+          ) ?? "",
+        defaultProvider: runtimeDefault.provider,
+      })
+    : undefined;


resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model) is evaluated twice — once in the guard condition (line 435) and once inside the normalizeSubagentSpawnModelSelection call (lines 440-441). Extract it to a variable to avoid the redundant call and improve clarity:

Suggested change

const normalizedPrimaryModel = resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model)

? normalizeSubagentSpawnModelSelection({

cfg: params.cfg,

model:

normalizeModelSelection(

resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model),

) ?? "",

defaultProvider: runtimeDefault.provider,

})

: undefined;

const primaryModelValue = resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model);

const normalizedPrimaryModel = primaryModelValue

? normalizeSubagentSpawnModelSelection({

cfg: params.cfg,

model: normalizeModelSelection(primaryModelValue) ?? "",

defaultProvider: runtimeDefault.provider,

})

: undefined;

Prompt To Fix With AI

This is a comment left during a code review. Path: src/agents/model-selection.ts Line: 435-444 Comment: `resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model)` is evaluated twice — once in the guard condition (line 435) and once inside the `normalizeSubagentSpawnModelSelection` call (lines 440-441). Extract it to a variable to avoid the redundant call and improve clarity: ```suggestion const primaryModelValue = resolveAgentModelPrimaryValue(params.cfg.agents?.defaults?.model); const normalizedPrimaryModel = primaryModelValue ? normalizeSubagentSpawnModelSelection({ cfg: params.cfg, model: normalizeModelSelection(primaryModelValue) ?? "", defaultProvider: runtimeDefault.provider, }) : undefined; ``` How can I resolve this? If you propose a fix, please make it concise.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c91556c06d

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".

chatgpt-codex-connector · 2026-03-05T13:09:01Z

src/agents/model-selection.ts

+  const selectedProvider = normalizeProviderId(inferredProvider ?? params.defaultProvider);
+  const selectedModel = normalizeProviderModelId(selectedProvider, trimmed);
+  return `${selectedProvider}/${selectedModel}`;


Preserve alias model names in subagent spawn selection

Do not unconditionally prefix bare model strings with a provider here, because this rewrites aliases (for example fast) into provider-qualified refs (for example anthropic/fast) before sessions.patch. Alias resolution in resolveModelRefFromString only runs for values without /, so provider-prefixing causes valid alias-based sessions_spawn overrides/defaults to bypass alias lookup and fail allowlist checks as model not allowed.

Useful? React with 👍 / 👎.

Fix sessions_spawn Google/Gemini model routing

c91556c

openclaw-barnacle bot added agents Agent runtime and tooling size: S labels Mar 5, 2026

greptile-apps bot reviewed Mar 5, 2026

View reviewed changes

chatgpt-codex-connector bot reviewed Mar 5, 2026

View reviewed changes

Workz777 mentioned this pull request Mar 6, 2026

[Bug]: Unrecognised model IDs silently fall back to primary default — bypasses configured fallback chain and tool permissions #37813

Open

thewilloftheshadow closed this Mar 6, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix sessions_spawn Google/Gemini model routing#36154

Fix sessions_spawn Google/Gemini model routing#36154
Octane0411 wants to merge 1 commit intoopenclaw:mainfrom
OasAIStudio:codex/fix-36134-gemini-spawn

Octane0411 commented Mar 5, 2026

Uh oh!

greptile-apps bot commented Mar 5, 2026

Uh oh!

greptile-apps bot Mar 5, 2026

Uh oh!

greptile-apps bot Mar 5, 2026

Uh oh!

chatgpt-codex-connector bot left a comment

Uh oh!

chatgpt-codex-connector bot Mar 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Octane0411 commented Mar 5, 2026

Uh oh!

greptile-apps bot commented Mar 5, 2026

Greptile Summary

Confidence Score: 2/5

Uh oh!

greptile-apps bot Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Mar 5, 2026

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 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants