feat: enable agent mode support for LM Studio models

[veeceey]

Summary

Adds LM Studio to PROVIDER_TOOL_SUPPORT to enable Agent mode for LM Studio models.

Changes

• Added lmstudio entry to PROVIDER_TOOL_SUPPORT that references Ollama's tool support implementation
• Added comprehensive tests for LM Studio tool support in toolSupport.test.ts

Rationale

LM Studio uses an OpenAI-compatible API similar to Ollama and serves similar models with tool calling support. By referencing Ollama's implementation, LM Studio models can now use Agent mode with the same model compatibility logic.

Test Plan

Added unit tests that verify:

• Supported models (llama3.1, llama3.2, qwen2, mixtral, mistral) return true
• Explicitly unsupported models (vision, math, guard) return false
• Case insensitivity works correctly
• Empty model names return false

The implementation follows the same pattern as other providers in the codebase and reuses the battle-tested Ollama logic.

Closes #5182

---

Continue Tasks: ▶️ 2 queued — View all

---

Summary by cubic

Enables Agent mode for LM Studio models by delegating tool support to Ollama, with safe handling for hyphenated model IDs. Exclusion checks now block mistral-openorca and mistral-lite in both hyphenated and non-hyphenated forms. Addresses #5182.

• New Features
  • Added lmstudio to PROVIDER_TOOL_SUPPORT. Exclusions are checked via a helper against both raw and hyphen-stripped names, then we delegate to Ollama with the raw name and fall back to the normalized ID.
  • Added tests for hyphenated IDs (Meta-Llama-3.1-8B-Instruct-GGUF, Qwen2-7B-Instruct, Mixtral-8x7B-Instruct-v0.1), unsupported hyphenated names (Llama-Vision-Free, Math-Solver-7B, Guard-Model), mistral-openorca/mistral-lite variants, case insensitivity, and empty model names.

^{Written for commit cfff26f. Summary will update on new commits.}

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[cubic-dev-ai]

1 issue found across 2 files

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="core/llm/toolSupport.ts">

<violation number="1" location="core/llm/toolSupport.ts:192">
P2: LM Studio uses hyphenated model identifiers (e.g., “Meta-Llama-3.1-8B-Instruct-GGUF”), but lmstudio delegates to Ollama’s heuristic which only checks substrings like “llama3.1” without hyphens. This will fail to detect tool support for common LM Studio model IDs (e.g., “llama-3.1-8b-instruct”), disabling agent mode even when tool calling is supported.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[veeceey]

I have read the CLA Document and I hereby sign the CLA

[cubic-dev-ai]

1 issue found across 2 files (changes from recent commits).

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="core/llm/toolSupport.ts">

<violation number="1" location="core/llm/toolSupport.ts:195">
P2: LM Studio’s hyphen-stripping normalization bypasses Ollama’s explicit exclusion for "mistral-openorca" (and other hyphen-sensitive exclusions), potentially marking unsupported models as tool-capable.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

1 issue found across 1 file (changes from recent commits).

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="core/llm/toolSupport.ts">

<violation number="1" location="core/llm/toolSupport.ts:196">
P2: LM Studio hyphenated model IDs are rejected before normalization due to the new early return on the raw Ollama check, contradicting existing tests expecting these IDs to be supported.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

1 issue found across 2 files (changes from recent commits).

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="core/llm/toolSupport.ts">

<violation number="1" location="core/llm/toolSupport.ts:206">
P2: LM Studio exclusion check no longer tests normalized model IDs, so hyphenated models like `mistral-lite` bypass the `mistrallite` exclusion and get marked tool-capable via the Ollama heuristic.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}