feat: add Synthetic LLM provider support

[meneizs]

Summary

Add support for Synthetic as a first-class LLM provider in swarm-tools.

New Models Added

Coordinator Models:

• synthetic/hf:moonshotai/Kimi-K2.5 - Strong reasoning model
• synthetic/hf:deepseek-ai/DeepSeek-V3.2 - Advanced reasoning

Worker Models:

• synthetic/hf:moonshotai/Kimi-K2.5 - Strong reasoning
• synthetic/hf:MiniMaxAI/MiniMax-M2.1 - Fast and cost-effective
• synthetic/hf:Qwen/Qwen3-Coder-480B-A35B-Instruct - Powerful coding model

Lite Models:

• synthetic/hf:MiniMaxAI/MiniMax-M2.1 - Fast and cost-effective
• synthetic/hf:Qwen/Qwen3-Coder-480B-A35B-Instruct - Powerful coding model

Why

• Synthetic provides OpenAI-compatible API at https://api.synthetic.new/openai/v1
• Enables users already configured with Synthetic in OpenCode to use swarm-tools
• Cost-effective alternative for multi-agent workflows

Testing

• All tests passing (16/16)
• Manual testing confirms models appear in swarm setup flow
• Includes test coverage for new models

Refs: #148, #150

Summary by CodeRabbit

• New Features

  • Synthetic is now available as a first-class LLM provider.
  • New model options added: Kimi K2.5, DeepSeek V3.2, MiniMax M2.1, and Qwen3 Coder 480B.
  • Models organized by task type: Coordinator (complex), Worker (parallel), Lite (docs/tests/simple edits).
  • Default model selections and setup prompts updated to include Synthetic options.

• Documentation

  • Added usage guidance, model enumeration, requirements, and integration notes.

• Tests

  • Added tests verifying Synthetic model selection.

[vercel]

Someone is attempting to deploy a commit to the joelhooks' projects Team on Vercel.

A member of the Team first needs to authorize it.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds Synthetic as a first-class LLM provider to swarm-tools, registering new Synthetic models (Kimi K2.5, DeepSeek V3.2, MiniMax M2.1, Qwen3 Coder 480B) across coordinator/worker/lite selections and adding tests to verify explicit synthetic model selection.

Changes

Cohort / File(s)	Summary
Changeset Documentation \.changeset/add-synthetic-provider.md	New changeset documenting Synthetic as a first-class LLM provider; enumerates Coordinator/Worker/Lite models, usage, requirements, and references.
Model Configuration / CLI packages/opencode-swarm-plugin/bin/swarm.ts	Added Synthetic model options to selection lists and propagated choices through interactive/default/reinstall paths (Coordinator: Kimi K2.5, DeepSeek V3.2; Worker: Kimi K2.5, MiniMax M2.1, Qwen3 Coder 480B; Lite defaults updated).
Test Coverage packages/opencode-swarm-plugin/src/model-selection.test.ts	Added tests verifying that explicit model fields in subtasks return exact synthetic model identifiers (Kimi K2.5, DeepSeek V3.2, MiniMax M2.1).

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Poem

🐰 I hopped into code with a joyful nibble,

added Synthetic pals to every nibble,
Kimi, DeepSeek, MiniMax join the run,
Qwen3 Coder scribbles lines of fun,
Tests now sparkle — a tiny hop, well done! 🎉

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'feat: add Synthetic LLM provider support' accurately summarizes the main change: adding Synthetic as a first-class LLM provider with multiple models registered for different roles.

_{✏️ 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

---

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: 1

🤖 Fix all issues with AI agents

In `@packages/opencode-swarm-plugin/bin/swarm.ts`:
- Around line 375-384: The WORKER_MODELS constant is missing the Qwen3-Coder
entry while the interactive worker selection includes
"synthetic/hf:Qwen/Qwen3-Coder-480B-A35B-Instruct"; add that same value and
label to the WORKER_MODELS array so both lists stay consistent (update the
WORKER_MODELS entry list to include value
"synthetic/hf:Qwen/Qwen3-Coder-480B-A35B-Instruct" and a matching label like
"Synthetic Qwen3-Coder"), ensuring any code that reads WORKER_MODELS (e.g.,
validation or listings) will recognize the model.

🧹 Nitpick comments (1)

.changeset/add-synthetic-provider.md (1)

1-39: Well-documented changeset with comprehensive details.

The changeset clearly describes what changed, why it matters, usage instructions, and requirements. The 🐝 emoji is appropriately included per coding guidelines.

One optional enhancement: The coding guidelines suggest including a quote from pdf-brain.pdf for context. If there's a relevant passage about multi-agent coordination or provider flexibility, consider adding it for thematic consistency.

,

[meneizs]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[Edo78]

Insted of having an hardcoded list of models shouldn't be way better to run opencode models to obtain the dynamic list of supported models and providers?