[ENHANCEMENT] Roo Code should report when models are excluded for not supporting tools

[jamestmartin]

Problem (one or two sentences)

I just spent my entire evening trying to debug why three quarters of my models were showing up with ollama ls didn't appear in Roo Code's UI and, when entered manually, got the error The model ID (seed-oss-36b:iq3_xss) you provided is not available. Please choose a different model.

I tried reloading the window, removing and re-adding models, renaming models, deleting models, clearing Roo Code's data, rm -rf-ing Roo Code's data from its folder, uninstalling and reinstalling Roo Code, checking Roo Code's output, checking the developer tools log, checking if the models were showing up in Ollama's API output. I tried reading the other issues to see if there was some known issue.

I then tried to use Ollama's OpenAI-compatible endpoint instead. The models showed up there. OK, clearly a Roo Code bug. I gathered all of the data about my system, logs, etc., and went as far as to write a complete bug report. Right before I submitted it--well, I should probably test those models actually work, first.

And I did. They didn't work.

API Request Failed
OpenAI completion error: 400 registry.ollama.ai/library/deepseek-coder-v2-lite-i1:Q4_K_M does not support tools

Oh.

Pretty frustrating. Removing XML tool calls is probably the right move long-term, but acting like the models outright don't exist is a major usability issue.

Context (who is affected and when)

Anyone using Ollama who has models that don't support native tool calling in the correct format installed.

Models affected (e.g.):

glm-4.7-flash:UD-Q4_K_XL
seed-oss-36b:IQ3_XXS
unsloth-qwen3-vl-32b:UD-Q4_K_XL
deepseek-coder-v2-lite-i1:Q4_K_M
devstral-small-24b:Q4_K_XL
gemma3:27b-it-qat
hf.co/unsloth/Qwen3-VL-32B-Thinking-GGUF:UD-Q4_K_XL
hf.co/unsloth/Seed-OSS-36B-Instruct-GGUF:IQ3_XXS
hf.co/unsloth/GLM-4.7-Flash-GGUF:UD-Q4_K_XL
hf.co/mradermacher/DeepSeek-Coder-V2-Lite-Instruct-i1-GGUF:Q4_K_M
hf.co/unsloth/Devstral-Small-2-24B-Instruct-2512-GGUF:Q4_K_XL

Models not affected:

milkey/Seed-OSS-36B-Instruct:Q4_K_M
gpt-oss:20b
gpt-oss:120b
qwen3-next:80b

That models list isn't complete since I removed and duplicated a bunch of models while trying to figure out the issue.

Desired behavior (conceptual, not technical)

The UI should indicate that models that don't support native tool calling in the correct format are excluded from the list. When a user tries to enter one of these models manually, they should be notified that the model doesn't support tool calls in the correct format, not that it's "not available".

Constraints / preferences (optional)

A simple sentence, "Models that do not support native tool calls are excluded from this list" would be adequate. Alternatively, continue to list the models, but explicitly mark them as not supporting tool calls in the correct format (e.g. greyed out/strikethrough).

Request checklist

• I've searched existing Issues and Discussions for duplicates
• This describes a specific problem with clear context and impact

Roo Code Task Links (optional)

No response

Acceptance criteria (optional)

No response

Proposed approach (optional)

No response

Trade-offs / risks (optional)

No response

[roomote]

I've analyzed this issue and here's my proposed implementation plan:

Problem Analysis

The issue occurs because:

1. Ollama models are filtered in src/api/providers/fetchers/ollama.ts based on the capabilities array - models without "tools" capability are silently excluded (return null)
2. When users manually enter a model that was filtered, they get the generic error "The model ID ... you provided is not available" which doesn't explain WHY
3. There's no indication in the UI that models are being filtered based on tool support

Proposed Solution

1. Add informational messaging to the Ollama settings UI (webview-ui/src/components/settings/providers/Ollama.tsx)

Add a note near the model picker explaining that models without native tool calling support are excluded from the list. This addresses the user's suggestion: "A simple sentence, 'Models that do not support native tool calls are excluded from this list' would be adequate."

2. Add a more specific error message for Ollama models

When a user manually enters a model that doesn't appear in the list, provide a more helpful error message explaining that the model may not support native tool calling, rather than just saying it's "not available".

3. Translation updates (webview-ui/src/i18n/locales/en/settings.json + other locales)

Add new translation keys:

• settings:providers.ollama.toolsNote - Explanatory note for the UI
• settings:validation.modelNotToolCapable - Specific error message for Ollama models that may lack tool support

This approach:

• Keeps the implementation simple and focused
• Addresses both user scenarios (browsing models and manual entry)
• Provides clear, actionable information to users
• Follows the existing patterns in the codebase

Please mention @roomote in a comment saying that you approve this plan, and I'll proceed with the implementation.

[jamestmartin]

This issue is exacerbated by this Ollama issue, which is why so few of my models work: ollama/ollama#13768

[jamestmartin]

@roomote I approve this plan.

[roomote]

Thanks for the approval! I'll proceed with implementing the plan now. I'll open a PR to share progress.

[roomote]

Opened a draft PR with my attempt at addressing this: #10915

The PR adds:

1. An informational note below the Ollama model picker explaining that models without native tool calling support are excluded
2. A more specific error message when manually entering a model that may not support tools

Feedback welcome!