App is calling the wrong endpoint for the Ollama provider when using local models

Description

I discovered that the application is using the wrong endpoint when sending requests to the Ollama provider for local models.
Currently, OpenCode is calling:

/chat/completions

However, the correct endpoint for the Ollama chat API is:

/chat

This causes messages to not be processed correctly by the local Ollama models.
The implementation needs to be updated so that the provider uses the /chat endpoint as defined in the official Ollama documentation.

OpenCode version

v1.0.65

Steps to reproduce

services:
local-llm:
image: ollama/ollama
container_name: local-llm
ports:
- "11434:11434"
volumes:
- ollama_data:/root/.ollama
healthcheck:
test: ["CMD-SHELL", "ollama ls >/dev/null 2>&1 || exit 1"]
interval: 10s
timeout: 5s
retries: 60
start_period: 20s
restart: unless-stopped

ollama-puller:
image: alpine:latest
container_name: ollama-puller
depends_on:
local-llm:
condition: service_healthy
volumes:
- "./pull-models.sh:/pull-models.sh"
environment:
MODELS: "gemma3:4b embeddinggemma:latest"
entrypoint: ["/bin/sh", "/pull-models.sh"]
healthcheck:
test: ["CMD-SHELL", "test -f /tmp/models_ready"]
interval: 30s
timeout: 10s
retries: 600
start_period: 60s
restart: unless-stopped

ai-cli:
image: node:22-bookworm
container_name: ai-cli
tty: true
stdin_open: true
ports:
- "4096:4096"
volumes:
- "C:\Users\...\Documents\GitHub\...\...\environment:/environment"
environment:
OPENCODE_CONFIG: "/environment/custom-config.json"
depends_on:
local-llm:
condition: service_healthy
ollama-puller:
condition: service_healthy
working_dir: /environment/workspace
command: >
bash -lc '
npm install -g npm@latest &&
npm install -g opencode-ai@latest &&
opencode serve --hostname 0.0.0.0 --port 4096
'
healthcheck:
test: ["CMD-SHELL", "curl -f http://localhost:4096/app || exit 1"]
interval: 10s
timeout: 5s
retries: 5
start_period: 30s
restart: unless-stopped

volumes:
ollama_data:

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

@filaupe

@filaupe can you show me your opencode config? If you are using the openai compatible provider then it will use the openai compatible endpoint because /chat isn't compatible

@filaupe can you show me your opencode config? If you are using the openai compatible provider then it will use the openai compatible endpoint because /chat isn't compatible

Apparently that was it, but what would be the correct config? After I removed that line, it started returning an error.

|opencode server listening on http://0.0.0.0:4096⁠

507 | ...options,

508 | })

509 | s.sdk.set(key, loaded)

510 | return loaded as SDK

511 | })().catch((e) => {

512 | throw new InitError({ providerID: provider.id }, { cause: e })

ProviderInitError: ProviderInitError

data: {

providerID: "ollama",

},

  at <anonymous> (src/provider/provider.ts:512:13)

500 | timeout: false,

501 | })

502 | }

503 | }

504 | const fn = mod[Object.keys(mod).find((key) => key.startsWith("create"))!]

505 | const loaded = fn({

TypeError: fn2 is not a function. (In 'fn2({

      name: provider.id,

      ...options2

    })', 'fn2' is undefined)

  at <anonymous> (src/provider/provider.ts:505:22)

Maybe try replacing the npm package with ollama-ai-provider-v2

If u show ur config I can help you through it

Maybe try replacing the npm package with ollama-ai-provider-v2

If u show ur config I can help you through it

Ok:

{
"$schema": "https://opencode.ai/config.json",
"agent": {
"build": {
"mode": "primary",
"disable": false,
"model": "ollama/gemma3:4b",
"prompt": "{file:./prompts/build.txt}",
"description": "Builds code for best practices and potential issues"
},
"code-review": {
"mode": "subagent",
"disable": false,
"model": "ollama/gemma3:4b",
"prompt": "{file:./prompts/code-review.txt}",
"description": "Reviews code for best practices and potential issues"
}
},
"disabled_providers": ["opencode"],
"provider": {
"ollama": {
"name": "Ollama (local)",
"options": {
"baseURL": "http://local-llm:11434"
},
"models": {
"gemma3:4b": {
"name": "gemma3:4b"
}
}
}
},
"tools": {
"write": true,
"edit": true,
"read": true,
"bash": true,
"list": true,
"grep": true,
"patch": true,
"glob": true,
"todowrite": true,
"todoread": true,
"webfetch": true
}
}

And this one causes that issue right:

  "provider": {
    "ollama": {
      "name": "Ollama (local)",
      "npm": "ollama-ai-provider-v2"
      "options": {
        "baseURL": "http://local-llm:11434"
      },
      "models": {
        "gemma3:4b": {
          "name": "gemma3:4b"
        }
      }
    }
  },

Hm is there any reason you can't use the openai compatible one? I think a patch needs to be made to this ollama provider to work for your usecase if this is breaking^

And this one causes that issue right:
  "provider": {
    "ollama": {
      "name": "Ollama (local)",
      "npm": "ollama-ai-provider-v2"
      "options": {
        "baseURL": "http://local-llm:11434"
      },
      "models": {
        "gemma3:4b": {
          "name": "gemma3:4b"
        }
      }
    }
  },
Hm is there any reason you can't use the openai compatible one? I think a patch needs to be made to this ollama provider to work for your usecase if this is breaking^

Here is the text in English:

I was able to fix the issue. The solution was to keep using "npm": "@ai-sdk/openai-compatible", but update the Ollama baseURL to "http://ollama-server:11434/v1". After that, everything worked correctly.

Awesome

Navigation Menu

Repository navigation

App is calling the wrong endpoint for the Ollama provider when using local models #4341

Description

OpenCode version

Steps to reproduce

Screenshot and/or share link

Operating System

Terminal

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

App is calling the wrong endpoint for the Ollama provider when using local models #4341

Description

Description

OpenCode version

Steps to reproduce

Screenshot and/or share link

Operating System

Terminal

Activity

rekram1-node commented on Nov 14, 2025

filaupe commented on Nov 14, 2025

rekram1-node commented on Nov 14, 2025

filaupe commented on Nov 14, 2025

rekram1-node commented on Nov 15, 2025

filaupe commented on Nov 16, 2025

rekram1-node commented on Nov 16, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions