Skip to main content
DeepSeek is CowAgent’s default provider. A single deepseek_api_key enables text chat, image understanding, speech-to-text, text-to-speech and web search.
All capabilities below can be configured in one place via the “Model Management” page in the Web Console, with no need to manually edit the configuration file.

Text Chat

Model Selection

Thinking Mode

deepseek-flash and the V4 series (deepseek-v4-flash / deepseek-v4-pro) support an explicit “thinking mode”: before producing the final answer, the model emits a chain of thought (reasoning_content) to improve answer quality.

Toggle

Controlled by the global enable_thinking config, and can also be toggled from the Web Console’s configuration page:
  • true: the model thinks before answering across all channels. The Web Console displays the thinking process; IM channels (WeChat / WeCom / DingTalk / Feishu) do not show it but still get better answers.
  • false: thinking is disabled, responses are faster, and time-to-first-token is lower.

Reasoning Effort

Under thinking mode, reasoning_effort controls reasoning intensity:
reasoning_effort only takes effect when enable_thinking is true; it is ignored automatically when the model does not support thinking mode.

Behavior Notes

  • Sampling parameters: in thinking mode, temperature, top_p, presence_penalty, and frequency_penalty are ignored by the server (without errors). CowAgent automatically skips them.
  • Multi-turn tool calls: when the history contains tool calls, DeepSeek requires every assistant message to include reasoning_content. CowAgent handles this automatically, so toggling thinking mode across turns will not cause errors.
deepseek-flash (V4.1 Flash, natively multimodal) is used by default; switch to deepseek-v4-pro for complex tasks; enable enable_thinking when deep reasoning is needed.

Image Understanding

deepseek-flash accepts images natively; as the main model it handles images directly, with no extra configuration.

Speech-to-Text (ASR)

Reuses deepseek_api_key; no model needs to be set. 21 languages are detected automatically, and each clip can be up to 120 seconds. Web console recordings work out of the box; AMR / SILK / MP3 and other formats need the optional dependencies (requirements-optional.txt) and ffmpeg.

Text-to-Speech (TTS)

10 languages are detected automatically, and Markdown is stripped before synthesis. When to reply with voice is set in the Web Console under “Model Management → Speech Synthesis”. Reuses deepseek_api_key with no extra setup. See Web Search.