DeepSeek Harness

Use uncensored models in DeepSeek Harness

Add Silk Compute to DeepSeek Harness with one config entry. Every model we serve then shows up in its model picker.

1. Set your API key

Create a key on the API keys page, then export it. Add the line to ~/.zshrc or ~/.bashrc so new terminals have it too.

Shell
export SILK_API_KEY="sk-..."

2. Start DeepSeek Harness

It needs Node.js. Start it from the same terminal, and it opens in your browser.

Shell
npx @deepseek-ai/dsh web

3. Add Silk Compute

Add this to ~/.dsh/profiles/web/cordis.patch.yml. If the file already has an llm-pi-ai entry, add silkcompute to its providers and keep the others. No restart needed.

cordis.patch.yml
- id: llm-pi-ai
  config:
    providers:
      silkcompute:
        displayName: Silk Compute
        apiKeyEnv: SILK_API_KEY
        api: openai-completions
        baseURL: https://api.silkcompute.com/v1
        defaultContextWindow: 1048576
        compat:
          supportsDeveloperRole: false
          maxTokensField: max_tokens
        models:
          - id: kimi-k3.1-pro-uncensored
            name: Kimi K3.1 Pro
            input: [text, image]
            reasoningEfforts: {off: none, low: low, high: high, max: max}
          - id: deepseek-v4.1-flash-1017-uncensored
            name: DeepSeek V4.1 Flash 1017
            input: [text, image]
            reasoningEfforts: {off: none, low: low, high: high, max: max}
          - id: deepseek-v4.1-flash-1017-fast-uncensored
            name: DeepSeek V4.1 Flash 1017 Fast
            input: [text, image]
            reasoningEfforts: {off: none, low: low, high: high, max: max}
          - id: glm-5.3-uncensored
            name: GLM-5.3
            input: [text]
            reasoningEfforts: {off: none, low: low, high: high, max: max}
          - id: glm-5.3-flash-uncensored
            name: GLM-5.3 Flash
            input: [text, image]
            reasoningEfforts: {off: none, low: low, high: high, max: max}
          - id: hy4-uncensored
            name: Hy4
            input: [text, image]
            reasoningEfforts: {off: none, low: low, high: high, max: max}

4. Pick a model

Choose a workspace, then pick Silk Compute and a model in the model picker. Set the effort in the Effort menu.

Troubleshooting

Run this in the same terminal. A list of models means the key works. A 401 means SILK_API_KEY is empty or wrong there.

Shell
curl -s https://api.silkcompute.com/models -H "Authorization: Bearer $SILK_API_KEY"

Notes

Written for DeepSeek Harness 0.2. It is in developer preview, so if a field is refused, check its Configure models guide.