[herjarsa]

Problem

Three related issues cause failures in OpenCode when using the OmniRoute plugin:

1. "No credentials for provider: combo"

OpenCode auto-discovers providers by grouping models under each unique owned_by value from /v1/models. When OmniRoute returns models with owned_by: "combo", OpenCode creates a separate provider entry without credentials:

No credentials for provider: combo

2. "Unable to determine provider for model"

OpenCode parses model IDs like combo/master and extracts combo as the provider name, ignoring the plugin's providerID: "omniroute". This causes authentication failures and the error:

Unable to determine provider for model 'master'. Use a provider/model prefix (e.g. openai/master) or ensure the model is added as a combo entry.

3. AbortError on slow connections

The default 10s timeout for /v1/models, /api/combos, and other fetch endpoints causes AbortError on WSL and other slow network setups where responses can exceed 270KB and take >10s to transfer.

Root Cause

1. owned_by mismatch: The /v1/models endpoint returns models with different owned_by values (e.g., combo, opencode-zen, kilocode). OpenCode's auto-discovery creates separate providers for each, but only omniroute has credentials.

2. Combo prefix mismatch: buildComboKey generates model IDs like combo/master. OpenCode's AI SDK parses this and tries to route to a combo provider, which doesn't exist.

3. Timeout too aggressive: 10s timeout is insufficient for large responses on slow networks.

Solution

Fix 1: Normalize owned_by

Normalize all owned_by values to the plugin's providerId (e.g., omniroute) after fetching from /v1/models. This ensures all models stay under the same provider block with valid credentials.

Fix 2: Use omniroute/ prefix for combos

Change buildComboKey to generate omniroute/{slug} instead of combo/{slug}. This ensures OpenCode's AI SDK correctly routes combo models to the registered omniroute provider with valid credentials.

Fix 3: Increase timeouts

Increase default timeouts from 10s to 60s for main fetchers, and from 5s to 30s for enrichment/autoCombos.

Changes

• Added normalization loop in createOmniRouteConfigHook after fetching models
• Added normalization loop in createOmniRouteProviderHook for consistency
• Changed buildComboKey to use omniroute/ prefix instead of combo/
• Increased default timeouts from 10s to 60s for main fetchers
• Increased enrichment/autoCombos timeouts from 5s to 30s
• Updated all call sites to use the new timeout values

Testing

• All 228 existing tests pass
• Build succeeds with no errors
• Verified on WSL2 with 271KB /v1/models response

Impact

This is a backward-compatible fix that doesn't change the plugin's API or behavior for users who don't experience these issues. The normalization only affects the internal owned_by field used by OpenCode's auto-discovery mechanism.

[gemini-code-assist]

Code Review

This pull request adds normalization logic for the owned_by field of fetched models to map them to the plugin's providerId, preventing OpenCode from creating separate credential-less provider entries. The review feedback suggests removing redundant array length checks before iterating over rawModels to simplify the code, and points out the repository style guide requirement to include unit tests for production code changes.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

According to the repository style guide, tests must always be included when changing production code (such as files under src/). Please add unit tests to verify that the owned_by normalization logic correctly maps various owned_by values to the resolved providerId.

References

1. Always include tests when changing production code (src/, open-sse/, electron/, bin/). ^(link)

[gemini-code-assist]

The check if (rawModels.length > 0) is redundant because the for...of loop will naturally not execute if the array is empty. Removing this check simplifies the code and reduces nesting.

        for (const model of rawModels) {
          if (model && typeof model.owned_by === "string" && model.owned_by !== resolved.providerId) {
            model.owned_by = resolved.providerId;
          }
        }

[gemini-code-assist]

The check if (rawModels.length > 0) is redundant because the for...of loop will naturally not execute if the array is empty. Removing this check simplifies the code and reduces nesting.

      for (const model of rawModels) {
        if (model && typeof model.owned_by === "string" && model.owned_by !== resolved.providerId) {
          model.owned_by = resolved.providerId;
        }
      }

[diegosouzapw]

Thanks @herjarsa! The owned_by normalization is a genuinely good fix — OpenCode grouping models by owned_by and creating a credential-less combo provider is a real bug, and normalizing to providerId solves it cleanly. But a few things need addressing before merge:

1. The timeout bump is currently a no-op. You raised the default timeoutMs from 10_000 → 60_000 on the 5 fetcher functions, but the actual hook call-sites still pass 10_000 explicitly (e.g. createOmniRouteProviderHook ~fetcher(baseURL, apiKey, 10_000) and the config hook). So production traffic still uses the 10s timeout — the "slow network timeouts" part of the fix doesn't take effect. Either update the call-sites to 60_000 (or a configurable value) or remove the default change.

2. No tests (repo Hard Rule #18). buildComboKey (the combo/ → omniroute/ prefix change) and the owned_by normalization are both pure and easy to unit-test. Please add coverage — at minimum: a buildComboKey test asserting the omniroute/ prefix + collision disambiguation, and a normalization test asserting non-providerId owned_by values get rewritten.

3. It's CONFLICTING against release/v3.8.20 — needs a rebase.

The combo/ → omniroute/ rename is also a user-visible change (existing configs referencing combo/... model ids) — worth a note in the PR description. Fix those and I'll merge. Leaving open. 🙏

[kilo-code-bot]

Kilo Code Review could not run — your account is out of credits.

Add credits or switch to a free model to enable reviews on this change.

[herjarsa]

Superseded by #3588 — this branch was force-pushed with a clean commit history