deepseek-protocol-doctor
Run the following command in DeepSeek Harness:
dsh plugin install Whning0513/deepseek-protocol-doctor
Paste the following prompt into your AI chat to install this plugin:
Install the protocol checker plugin in DeepSeek Harness by running dsh plugin install Whning0513/deepseek-protocol-doctor; full source is available at https://github.com/Whning0513/deepseek-protocol-doctor
About this plugin
Integrating DeepSeek tool calling often surfaces bugs unrelated to the model: a tool result is sent back yet the request still 400s, a streamed parameter looks fine in real time but is not valid JSON once assembled, or the same history works with thinking off and fails with it on. deepseek-protocol-doctor is a static checker built to catch exactly these protocol-layer mistakes.
It takes your existing request JSON or a recorded SSE/JSONL stream and flags issues such as missing or mismatched tool_call_id, reasoning_content silently dropped by the client, function.arguments parsed before the stream is complete, interleaved multi-tool-call deltas appended in arrival order, strict schemas missing required or additionalProperties, and subtle max_tokens or thinking-routing misconfigurations. Every finding carries a stable code, and output is available as text, JSON, or SARIF for direct CI integration.
Designed for developers wiring up DeepSeek tool calls, streaming inference, or custom agent pipelines. It does not benchmark or judge answer quality; it simply isolates the protocol-layer bugs that masquerade as model flakiness.
Use Cases
- Diagnose 400 errors and parameter assembly issues in request JSON or SSE recordings
- Catch interleaved multi-tool-call deltas that corrupt streamed function.arguments
- Gate CI on protocol-level static validation of DeepSeek requests
Best For
- API developers integrating DeepSeek tool calling
- Engineers debugging streaming inference protocol issues in agent pipelines
- Backend teams enforcing consistent DeepSeek request standards in CI
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.