[FEATURE]: Strongly recommend supporting Anthropic and OpenAI API endpoints for the opencode Go model #30244

Activity

1. github-actions

   assigned

   StarpTech

   on Jun 1, 2026
2. github-actions commented on Jun 1, 2026

   

   github-actionsBot

   on Jun 1, 2026 – with GitHub Actions

   Contributor

   More actions

   This issue might be a duplicate of existing issues. Please check:

   • Anthropic-to-OpenAI schema translation #27047: Anthropic-to-OpenAI schema translation causing tool call failures with DeepSeek via the opencode.ai proxy (missing name field inside function wrapper)
   • DeepSeek thinking mode: reasoning_content not passed back for tool call turns, causing 400 errors #24722: DeepSeek tool call failures with 400 errors when reasoning_content is not passed back
   • Bug: OpenCode Go bridge layer incompatible with kimi-k2.6 tool calls #26331: OpenCode Go bridge layer incompatible with tool calls for models that use thinking/reasoning (same root cause pattern)

   These issues describe the same or closely related problems with the protocol translation layer and DeepSeek/thinking-enabled models. Please review them before proceeding — if your issue is already tracked there, consider adding context or a thumbs-up rather than opening a new report.
3. sontanon commented on Jun 7, 2026

   

   sontanon

   on Jun 7, 2026

   More actions

   OpenCode Go serves DeepSeek and Kimi only via /v1/chat/completions. Both providers also have Anthropic-compatible endpoints, which would fix the Claude Code tool-calling issues by eliminating the protocol translation layer entirely (see #24722, #26331)

   • DeepSeek: https://api.deepseek.com/anthropic Official DeepSeek Documentation
   • Kimi: https://api.moonshot.ai/anthropic Moonshot officially instructs Claude Code users to use this. Formal API reference is pending (docs: publish canonical API reference for /anthropic/v1/messages compatibility endpoint MoonshotAI/Kimi-K2#129), but the endpoint is functional and officially supported.

   This is a zero-code config change. The schema at packages/console/core/src/model.ts:33 already supports formatFilter for dual-format models. Qwen 3.6 Plus already works this way (both /v1/messages and /v1/chat/completions). The anthropicHelper and body converter (provider.ts:182-185) are transparent proxies when formats match.

   What's needed is to add two provider entries and formatFilter additions in the Go model data (ZEN_MODELS* secrets):

   + "deepseek-anthropic": {api: "https://api.deepseek.com/anthropic", format: "anthropic", ...}
   + "moonshot-anthropic": {api: "https://api.moonshot.ai/anthropic", format: "anthropic", ...}
   
     "deepseek-v4-pro": [
       {formatFilter: "oa-compat", providers: [{id: "deepseek", ...}]},
   +   {formatFilter: "anthropic", providers: [{id: "deepseek-anthropic", model: "deepseek-v4-pro"}]}
     ]
     // same pattern for deepseek-v4-flash, kimi-k2.5, kimi-k2.6, etc.
4. github-actions commented on Aug 7, 2026

   

   github-actionsBot

   on Aug 7, 2026 – with GitHub Actions

   Contributor

   More actions

   To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.
5. github-actions

   closed this as not plannedon Aug 7, 2026