feat(providers): add siliconflow (GLOBAL) to built-in providers

[wu21-web]

Description

SiliconFlow is rapidly gaining traction as a major AI infrastructure and Model-as-a-Service (MaaS) provider, boasting over 10 million users. This PR adds the SiliconFlow (international) to built-in providers.

Type of Change

• New feature (non-breaking change that adds functionality)

How Has This Been Tested?

• make test passes locally
• Manual testing (describe below)
  go test and go build

Checklist

• My code follows the project's coding style (go fmt, go vet)
• I have performed a self-review of my code
• I have added tests that prove my fix is effective or my feature works
• New and existing unit tests pass locally with my changes
• I have updated the documentation accordingly (if applicable)
• I have signed the CLA

Related Issues

none

[github-actions]

🔍 OpenCodeReview found 1 issue(s) in this PR.

• ✅ Successfully posted inline: 1 comment(s)

[lizhengfeng101]

Thanks for adding the SiliconFlow preset — the wiring itself looks straightforward.

One concern on the Models list: several entries look like previous-generation models rather than what a new user would want as a first choice.

• deepseek-ai/DeepSeek-V3 and deepseek-ai/DeepSeek-R1 are superseded by the V3.2/V4 line that is already in the same list.
• Qwen/Qwen3-32B and Qwen/Qwen3-14B are both older and quite small for this workload — review runs depend heavily on reliable tool calling and long-context handling, and small models tend to degrade there.
• baidu/ERNIE-4.5-300B-A47B also looks like an older generation.

Since these presets are exactly what people pick from the provider TUI without thinking twice, the defaults carry a lot of weight. Could you re-verify the list against SiliconFlow's current model catalog and confirm the following?

1. Every model id still resolves on the global endpoint (https://api.siliconflow.com/v1) — some ids are only served on one region.
2. Each listed model supports the streaming + tool-calling behaviour the review loop needs.
3. The list reflects the current generation, and older/smaller entries are either dropped or intentionally kept (if intentional, a one-line rationale in the PR description would help).

A shorter list of models you have actually run a review with is more valuable here than a longer list that is partly stale. Thanks!

[wu21-web]

Thanks for adding the SiliconFlow preset — the wiring itself looks straightforward.

One concern on the Models list: several entries look like previous-generation models rather than what a new user would want as a first choice.

• deepseek-ai/DeepSeek-V3 and deepseek-ai/DeepSeek-R1 are superseded by the V3.2/V4 line that is already in the same list.
• Qwen/Qwen3-32B and Qwen/Qwen3-14B are both older and quite small for this workload — review runs depend heavily on reliable tool calling and long-context handling, and small models tend to degrade there.
• baidu/ERNIE-4.5-300B-A47B also looks like an older generation.

Since these presets are exactly what people pick from the provider TUI without thinking twice, the defaults carry a lot of weight. Could you re-verify the list against SiliconFlow's current model catalog and confirm the following?

1. Every model id still resolves on the global endpoint (https://api.siliconflow.com/v1) — some ids are only served on one region.
2. Each listed model supports the streaming + tool-calling behaviour the review loop needs.
3. The list reflects the current generation, and older/smaller entries are either dropped or intentionally kept (if intentional, a one-line rationale in the PR description would help).

A shorter list of models you have actually run a review with is more valuable here than a longer list that is partly stale. Thanks!

True, I used outdated sources. Will update soon.

[wu21-web]

		Models: []string{
			"deepseek-ai/DeepSeek-V4-Pro",
			"deepseek-ai/DeepSeek-V4-Flash",
			"meituan-longcat/LongCat-2.0",
			"Qwen/Qwen3.6-27B",
			"moonshotai/Kimi-K2.7-Code",
			"zai-org/GLM-5.2",
		},

What about this one? Referenced https://siliconflow.cn/models

[stay-foolish-forever]

Maybe you can remove LongCat-2.0 from the list :-)

[lizhengfeng101]

test failed

[wu21-web]

test failed

I enabled editing by maintainers.
It will be great if you can help me with that, I can't get access to my PC until tomorrow noon.
I did the merge on my phone and couldn't use tab instead of spaces.

[lizhengfeng101]

Done — just pushed a fix for the indentation (spaces → tabs). Should be green now. No worries, we've all been there with phone edits 😄

[lizhengfeng101]

LGTM