Attempting to use LM Studio with OpenCode - TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as URL.

I am using OpenCode on an M4 Mac on OS Sequoia. I have LM Studio setup with to serve "unsloth/Magistral-Small-2506-GGUF" as the model. When trying to /init, I get the error mentioned above:

TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as URL.

I believe LM Studio is expecting /v1/chat/completions or /v1/completions but I'm unsure how to work around this for OpenCode.

Below are the the .toml files I created for the opencode.json, model & provider.

#.opencode.json

{
"$schema": "https://opencode.ai/config.json",
"provider": {
"lmstudio": {
"npm": "@ai-sdk/openai-compatible",
"options": {
"baseURL": "http://localhost:1234/v1"
"apiKey": "12345"
},
"models": {
"magistral-small-2506": {}
}
}
}
}

#magistral-small-2506.toml -

name = "magistral-small-2506"
attachment = true
reasoning = true
temperature = true

[cost]
input = 3.00
output = 15.00
cache_read = 3.75
cache_write = 0.30

[limit]
context = 200_000
output = 50_000

#provider.toml:

name = "lmstudio"
env = ["12345"]
url = ["http://localhost:1234/v1"]
npm = "@ai-sdk/openai-compatible"

hey we just reworked how config works - can you refer to readme and try again? should be simpler and all in one file now

hey we just reworked how config works - can you refer to readme and try again? should be simpler and all in one file now

Seems to work now. Here is what my config.json looks like. Thank you for your help:

{
"$schema": "https://opencode.ai/config.json",
"provider": {
"lmstudio": {
"npm": "@ai-sdk/openai-compatible",
"options": {
"baseURL": "http://localhost:1234/v1"
},
"models": {
"lmstudio/unsloth/magistral-small-2506": {}
}
}
},
"model": "lmstudio/unsloth/magistral-small-2506"
}

Nice! It would be neat to be able to add a full list of all of the locally downloaded models from lm studio into the config automagically.

Nice! It would be neat to be able to add a full list of all of the locally downloaded models from lm studio into the config automagically.

Agreed! If it could just populate the models list with /v1/models, that would make things so much easier. And also have intelligent model switching where you can allocate certain models for certain tasks. Line one for web research, one for Architect, one for coding, and so on.

Pasting beautified json just in case someone needs it for visual parsing

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "lmstudio": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://localhost:1234/v1"
      },
      "models": {
        "lmstudio/unsloth/magistral-small-2506": {}
      }
    }
  },
  "model": "lmstudio/unsloth/magistral-small-2506"
}

Im getting the same error trying to connect to LiteLLM v 0.174 released 4 hours ago

{
  "provider": {
    "liteLLM": {
      "name": "liteLLM",
      "npm": "@ai-sdk/openai-compatible",
      "models": {
        "chatgpt-4o-latest": {
          "name": "ChatGPT 4o Latest"
        },
        "claude-3-5-haiku-latest": {
          "name": "Claude 3.5 Haiku Latest"
        },
        "claude-3-5-sonnet-latest": {
          "name": "Claude 3.5 Sonnet Latest"
        },
        "claude-3-7-sonnet-latest": {
          "name": "Claude 3.7 Sonnet Latest"
        },
        "claude-haiku-4-5": {
          "name": "Claude Haiku 4.5"
        },
        "claude-sonnet-4-5": {
          "name": "Claude Sonnet 4.5"
        },
        "gemini/gemini-2.0-flash": {
          "name": "Gemini 2.0 Flash"
        },
        "gemini/gemini-2.0-pro-exp-": {
          "name": "Gemini 2.0 Pro Exp"
        },
        "gemini/gemini-2.5-flash": {
          "name": "Gemini 2.5 Flash"
        },
        "gemini/gemini-2.5-pro": {
          "name": "Gemini 2.5 Pro"
        },
        "gemini/gemma-3-27b-it": {
          "name": "Gemma 3 27b IT"
        },
        "gpt-3.5-turbo": {
          "name": "GPT 3.5 Turbo"
        },
        "gpt-4.1": {
          "name": "GPT 4.1"
        },
        "gpt-4.1-mini": {
          "name": "GPT 4.1 Mini"
        },
        "gpt-5-codex": {
          "name": "GPT 5 Codex"
        },
        "gpt-5.1": {
          "name": "GPT 5.1"
        },
        "gpt-5.2": {
          "name": "GPT 5.2"
        },
        "gpt-image-1": {
          "name": "GPT Image 1"
        },
        "o4-mini": {
          "name": "o4 Mini"
        },
        "o4-mini-deep-research": {
          "name": "o4 Mini Deep Research"
        },
      "options": {
        "baseURL": "http://192.168.1.12:4005/v1",
        "api-key": "sk-xxxxxxxxxxxxxxxxxx"
      }
      }
    }
  },
  "$schema": "https://opencode.ai/config.json"
}

note litellm is in a docker its internal port is 4000 but outside docker its 4005, docker does the mapping. this is not the faliure.

or should I open a fresh ticket?

feat(workflows): export a workflow as a runnable .claude/workflows/*.…

Navigation Menu

Repository navigation

Attempting to use LM Studio with OpenCode - TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as URL. #195

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Attempting to use LM Studio with OpenCode - TypeError [ERR_INVALID_URL]: "undefined/chat/completions" cannot be parsed as URL. #195

Description

Activity

thdxr commented on Jun 19, 2025

teslanaut commented on Jun 19, 2025

testacode commented on Jun 20, 2025

teslanaut commented on Jun 20, 2025

DimmKirr commented on Aug 13, 2025

TRusselo commented on Dec 20, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions