feat: 【wip】support ollama llm

[zzcr]

Pull Request (OpenTiny NEXT-SDKs)

PR Checklist

Please check if your PR fulfills the following requirements:

• The commit message follows our Commit Message Guidelines
• Tests for the changes have been added (for bug fixes / features)
• Docs have been added / updated (for bug fixes / features)

PR Type

What kind of change does this PR introduce?

• Bugfix
• Feature
• Code style update (formatting, local variables)
• Refactoring (no functional changes, no api changes)
• Build-related changes
• CI-related changes
• Documentation-related changes
• Other... Please describe:

What is the current behavior?

Issue Number: N/A

What is the new behavior?

Does this PR introduce a breaking change?

• Yes
• No

Other information

Summary by CodeRabbit

• New Features
  • Added support for qwen3‑vl:8b model with multimodal image uploads (supports image attachments up to 10MB).
  • Continues to include deepseek‑r1:1.5b model powered by Ollama, expanding available model options.

[coderabbitai]

Walkthrough

Adds an Ollama-based model configuration for qwen3-vl:8b and a new dependency ollama-ai-provider-v2 to the next-wxt package.

Changes

Cohort / File(s)	Summary
Model config update packages/next-wxt/entrypoints/sidepanel/model-config.ts	Inserted a new DEFAULT_MODEL_CONFIGS entry for model qwen3-vl:8b with baseURL: "http://localhost:11434/api", providerType: createOllama, and multimodal settings (supportImages: true, maxFileSize: 10, supportedMimeTypes: ['image/']).
Dependency added packages/next-wxt/package.json	Added dependency ollama-ai-provider-v2 version ~3.3.0.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐇 I nibble configs, neat and spry,

qwen3-vl:8b hops by,
Local Ollama hums a tune,
Images ready, under moon,
A rabbit cheer — deploy soon! ✨

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main change: adding support for Ollama LLM with multimodal capabilities (qwen3-vl:8b model configuration).
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Merge Conflict Detection	✅ Passed	✅ No merge conflicts detected when merging into dev

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch zzc/support-ollama-0214

---

No actionable comments were generated in the recent review. 🎉

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Fix all issues with AI agents

In `@packages/next-wxt/entrypoints/sidepanel/model-config.ts`:
- Line 50: DEFAULT_MODEL_CONFIGS currently contains an entry with baseURL:
'http://localhost:11434/api' which will break for users without Ollama; update
that model config (the entry in DEFAULT_MODEL_CONFIGS that references the
localhost baseURL) to either remove or blank the baseURL and mark it as
local-only by adding a property like localOnly: true or featureFlag:
'ollamaLocal', and update its label (e.g., label: 'deepseek-r1:1.5b (Local)') so
the UI can hide or display it behind a feature flag or show a clear
local/dev-only indicator; ensure any UI code that reads DEFAULT_MODEL_CONFIGS
respects the new localOnly/featureFlag property to avoid presenting the option
to general users.
- Line 53: The icon properties for the Deepseek entries are missing markRaw
wrapping which causes Vue to make the icon components reactive; update the icon
assignments (e.g., the line setting "icon: IconModelDeepseek as unknown as
Component" and the DeepSeek-R1 entry) to wrap the icon with Vue's markRaw (so
use markRaw(IconModelDeepseek) and markRaw for the DeepSeek-R1 icon) to prevent
deep reactivity and avoid dev warnings; locate these in the model-config entries
for the Deepseek models and replace the bare icon references with
markRaw-wrapped references.

🧹 Nitpick comments (1)

packages/next-wxt/package.json (1)

23-23: Inconsistent version management — consider using catalog: like other dependencies.

All other entries in dependencies use "catalog:" for centralized version management via the pnpm workspace catalog. Hardcoding "~3.3.0" here breaks that convention and may lead to version drift. Add ollama-ai-provider-v2 to the workspace catalog and reference it as "catalog:" here.

-    "ollama-ai-provider-v2": "~3.3.0",
+    "ollama-ai-provider-v2": "catalog:",