[FEATURE]: Opt-in to share anonymized conversation data with model providers for improving open-source models #35303

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Problem

Many open-source models available through OpenCode Go (such as DeepSeek V4 Flash/Pro, Qwen3.7 Max/Plus, MiniMax M3, GLM-5.2, Kimi K2.7 Code, MiMo V2.5, etc.) could benefit from real-world coding conversation data to improve their performance. However, there is currently no opt-in mechanism for users who would like to voluntarily contribute their anonymized conversation data to help improve these models.

Proposed Solution

Add an optional opt-in setting that allows users to choose whether their anonymized conversation data can be shared with model providers for training/optimization purposes.

Key principles:

• Opt-in only — must be explicitly enabled by the user, disabled by default
• No sensitive data — API keys, credentials, and personal identifiable information must be stripped before sharing
• Clear transparency — users should know exactly which model provider would receive their data
• Per-provider consent — separate opt-in for each provider (e.g., DeepSeek, Qwen, MiniMax, etc.)

Example UI concept in settings:

☐ Allow sharing anonymized conversation data to help improve open-source models
  ☐ DeepSeek (V4 Flash, V4 Pro)
  ☐ Alibaba Qwen (Qwen3.7 Max, Qwen3.7 Plus)
  ☐ MiniMax (M3, M2.7)
  ☐ Zhipu AI (GLM-5.2, GLM-5.1)
  ☐ Moonshot (Kimi K2.7 Code, K2.6)
  ☐ Xiaomi (MiMo V2.5, V2.5 Pro)

Benefits

• Helps improve open-source models for the entire community
• Gives users a way to contribute if they want to
• Full transparency and control over data sharing
• Strengthens the open-source AI ecosystem

Privacy Considerations

The current privacy policy mentions that OpenCode may create "aggregated, de-identified or anonymized data" from Personal Data for improving the Services. This feature would extend that capability in a transparent, opt-in manner specifically for model training purposes, while ensuring:

• Users retain full control via opt-in
• Data is anonymized before sharing
• No API keys or credentials are ever included