OpenCode in Home Assistant Extremely Slow Communicating with Ollama

OpenCode version: 2.5.0

Operating system: Windows 11 Pro

Steps to reproduce:

Install Ollama on a Windows PC then connect to it remotely from OpenCode in theHome Assistant app. Issue any query or instruction which takes minutes to complete.

Description

I installed Ollama on a MINISFORUM Mini PC AI X1 Pro running Windows 11 and have 48GB of RAM dedicated to GPU. Queries on models up to 30B e.g. qwen3:30b take a matter of seconds when run in Ollama. The same queries in OpenCode take many minutes. This speed degradation occurs with all of the local models I've tried. I have not tried a cloud model in Ollama. I have Home Assistant running on an HA Yellow with a Raspberry Pi 5 (CM5) and 8 GB of RAM and SSD HDD. I've tried numerous configuration options none of which improved the speed. I've listed my current configuration below.

A simple instruction like "Turn on the office lamp" takes about 4 minutes.

This instruction took over 8 minutes, I lost track of time. AI suggested using a sunlight sensor instead of fixed times. I don't use fixed times there is a dawn and dusk template used. Not sure how it came up with fixed times.

interface_mode: terminal
terminal_theme: breeze
font_size: 14
cursor_style: block
cursor_blink: false
focus_mode: false
mcp_enabled: true
mcp_tool_profile: full
lsp_enabled: true
screenshot_enabled: false
access_token: ''
native_ha_mcp_enabled: false
native_ha_mcp_api_id: assist
home_briefing_enabled: true
decision_notes_enabled: true
restrict_sensitive_files: true
addon_access_enabled: false
cpu_mode: auto
enable_server: false
cors_origins: []
ppq_private_enabled: false
ppq_api_key: ''
z2m_url: ''
z2m_mqtt_topic: zigbee2mqtt
serial_devices: []
enable_openchamber_lan: false
env_vars: []
opencode_config: |-
  {
    "provider": {
      "ollama-ai-pc": {
        "npm": "@ai-sdk/openai-compatible",
        "name": "Ollama - AI-PC",
        "options": {
          "baseURL": "http://10.99.1.123:11434/v1"
        },
        "models": {
          "mistral-small:24b": {
            "name": "Mistral Small 24B - AI-PC",
            "limit": {
              "context": 32768,
              "output": 4096
            }
          },
          "qwen3:30b": {
            "name": "Qwen3 30B - AI-PC",
            "limit": {
              "context": 32768,
              "output": 4096
            }
          }
        }
      }
    }
  }
user_hooks_enabled: false

Plugins

No response

OpenCode version

No response

Steps to reproduce

No response

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response