OpenCode: how to configure Inline Prompts?

[stonemaster]

I recently switch from Copilot to Opencode.ai (Go) to evaluate the quality of models. While the chat/agentic mode works perfectly fine, the inline feature fails with Only HTTP adapters are for inline interactions.

Has anybody managed to get the opencode.ai API working? I tried to set the API endpoint to https://opencode.ai/zen/go/v1/chat/completions but I am getting a 404 error there 🤔

[stonemaster]

Ok I got to work now. I was near but I think I missed that GO doesn't use a prefix (opencode-go) in the HTTP mode:

    opts = {
      adapters = {
        http = {
          opencode_http = function()
            return require("codecompanion.adapters").extend("openai_compatible", {
              env = {
                api_key = "OPENCODE_API_KEY",
                url = "https://opencode.ai/zen/go",
              },
              schema = {
                model = {
                  default = "qwen3.6-plus", -- No opencode-go prefix here
                },
              },
            })
          end,
        },
      },

      interactions = {
        inline = {
          adapter = "opencode_http",
        },
        chat = {
          adapter = "opencode",
          model = "opencode-go/qwen3.6-plus",
        },
      },
    },

A thing like that might make sense to put in the documentation so if the maintainers agree, I'll happily create a PR :-)

Thanks for a great plugin btw!

[stonemaster]

This integration works but it takes quite a while until the results are there. Unlike in copilot mode where the diff is shown almost instantly. So probably there is some piece missing, or the model just does just too much work 🤔