How to configure OpenCode when running deeptutor init?

[GF-Huang]

────────  Step 1/5 · Ports  ────────

Backend port [8001]:
Frontend port [3782]:

────────  Step 2/5 · LLM provider  ────────

Pick an LLM provider

 [1] OpenAI             https://api.openai.com/v1
 [2] Anthropic          https://api.anthropic.com/v1
 [3] DeepSeek           https://api.deepseek.com
 [4] DashScope          https://dashscope.aliyuncs.com/compatible-mode/v1
 [5] Zhipu AI           https://open.bigmodel.cn/api/paas/v4
 [6] Moonshot           https://api.moonshot.cn/v1
 [7] Gemini             https://generativelanguage.googleapis.com/v1beta/openai/
 [8] SiliconFlow        https://api.siliconflow.cn/v1
 [9] OpenRouter         https://openrouter.ai/api/v1
[10] GitHub Copilot     https://api.githubcopilot.com
[11] Ollama             http://localhost:11434/v1
 [a] Show all providers
 [c] Custom / Other

Choice [1]: a
Pick an LLM provider

 [1] Custom
 [2] Custom (Anthropic API)
 [3] Azure OpenAI
 [4] Eden AI                https://api.edenai.run/v3
 [5] AiHubMix               https://aihubmix.com/v1
 [6] Novita AI              https://api.novita.ai/openai
 [7] Atlas Cloud            https://api.atlascloud.ai/v1
 [8] Unifically             https://api.unifically.com/v1
 [9] Cheaper Inference      https://api.cheaperinference.com/v1
[10] API Route              https://global.api-route.com/v1
[11] Requesty               https://router.requesty.ai/v1
[12] FutureInfra            https://futureinfra.ai/v1/ai
[13] Opper                  https://api.opper.ai/v3/compat
[14] Y-API                  https://api.y-api.bestvirtualgoods.com/v1
[15] VolcEngine             https://ark.cn-beijing.volces.com/api/v3
[16] VolcEngine Coding Plan https://ark.cn-beijing.volces.com/api/coding/v3
[17] BytePlus               https://ark.ap-southeast.bytepluses.com/api/v3
[18] BytePlus Coding Plan   https://ark.ap-southeast.bytepluses.com/api/coding/v3
[19] MiniMax                https://api.minimax.io/v1
[20] MiniMax (Anthropic)    https://api.minimax.io/anthropic
[21] Mistral                https://api.mistral.ai/v1
[22] Step Fun               https://api.stepfun.com/v1
[23] Xiaomi MIMO            https://api.xiaomimimo.com/v1
[24] vLLM/Local             local
[25] LM Studio              http://localhost:1234/v1
[26] llama.cpp              http://localhost:8080/v1
[27] Lemonade               http://localhost:13305/api/v1
[28] OpenVINO Model Server  http://localhost:8000/v3
[29] NVIDIA NIM             https://integrate.api.nvidia.com/v1
[30] Groq                   https://api.groq.com/openai/v1
[31] Qianfan                https://qianfan.baidubce.com/v2
 [b] Back
 [c] Custom / Other

[usmanmateen]

OpenCode isn't in the provider list, but OpenCode Zen and OpenCode Go both expose an OpenAI-compatible API, so the Custom provider works:

1. At "Pick an LLM provider", choose a (Show all providers), then 1 Custom (the OpenAI-compatible binding).
2. Answer yes to "Edit Base URL?" and enter:
   • OpenCode Go: https://opencode.ai/zen/go/v1
   • OpenCode Zen: https://opencode.ai/zen/v1
3. Paste your API key from https://opencode.ai/auth.
4. The wizard calls {base_url}/models and shows OpenCode's model list. Pick a model by its bare id as listed (e.g. kimi-k3), not the opencode-go/... form used in OpenCode's own config.
5. Accept the connectivity check at the end.

One catch: the Custom binding sends requests to /chat/completions, but the /models list also includes models that OpenCode's docs list under other endpoints: /messages (Claude, plus MiniMax and some Qwen models on Go), /responses (GPT, Grok), and per-model paths for Gemini on Zen. Pick a model from the /chat/completions group in OpenCode's docs (Go, Zen), e.g. the GLM, Kimi or DeepSeek models.

If you meant the opencode CLI itself (the coding agent) rather than its models: that isn't configured in deeptutor init. DeepTutor detects the opencode command on your PATH, and you connect it under My Agents in the web UI (added in v1.5.3), then consult it from a chat turn.