chore(ci): add OpenCode PR review workflow with deepseek

[spike-rabbit]

What kind of change does this PR introduce?

• CI related changes

What is the new behavior?

Adds automated code reviews based on opencode + the Siemens DeepSeek instance, modeled after #714 (which uses Qwen).

• Introduces opencode.json configuring the Siemens OpenAI-compatible provider and the deepseek-v4-flash model.
• Adds a GitHub Actions workflow that runs OpenCode on pull_request events and posts line-specific review comments.

Does this PR introduce a breaking change?

• No

[gemini-code-assist]

Code Review

This pull request adds a new configuration file opencode.json to define the Siemens LLM provider and configure the deepseek-v4-flash model. The feedback highlights two main issues: an excessively high output token limit of 1,048,576 that should be reduced to a standard limit like 8,192, and a likely typo in the model name deepseek-v4-flash since DeepSeek V4 has not been released.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

The output token limit is set to 1048576 (1M tokens), which is extremely high and likely unsupported as an output limit by any current LLM (typically output limits are 4,096 or 8,192 tokens). This might be a copy-paste error from the context limit. Setting this too high can lead to API errors or unexpected behavior.

Suggested change

	"output": 1048576
	"output": 8192

[gemini-code-assist]

The model name deepseek-v4-flash appears to be incorrect or a typo, as DeepSeek has not released a V4 model (the current latest versions are V3 and R1). Please verify the correct model identifier from the Siemens LLM API documentation.

[github-actions]

Pull Request Review

I've reviewed the two files added in this PR. Here are my findings:

---

File: opencode.json (line 15)

The output token limit is set to 1,048,576, which matches the context limit. Output limits are typically much lower (e.g., 4,096–128,000 tokens) even for models with large context windows. Setting this too high is unlikely to cause an error — the API will cap it — but it misrepresents the model's actual output capacity and could confuse future maintainers. I'd suggest reducing it to a value aligned with the model's documented max output (e.g., 8,192 or 128,000 depending on the model).

---

File: .github/workflows/opencode-review.yml (line 13)

The issues: read permission is granted but this workflow only interacts with pull requests, not issues. Following the principle of least privilege, this permission can be removed to keep the token scope minimal. Only pull-requests: write is needed for posting review comments.

opencode session  |  github run