Description

Telegram images are received by the gateway but not passed to the configured imageModel when using a local Ollama vision model.

Environment

• Clawdbot version: 2026.1.24-3
• OS: Windows 11
• Ollama version: (run ollama --version)
• Node version: 22.16.0

Configuration

{
  "agents": {
    "defaults": {
      "imageModel": {
        "primary": "ollama/qwen2.5vl:7b"
      },
      "model": {
        "primary": "ollama/qwen3-coder:latest"
      }
    }
  },
  "models": {
    "providers": {
      "ollama": {
        "api": "openai-completions",
        "baseUrl": "http://127.0.0.1:11434/v1",
        "models": [
          {
            "id": "qwen2.5vl:7b",
            "input": ["text", "image"]
          }
        ]
      }
    }
  }
}

Steps to Reproduce

1. Configure Ollama with qwen2.5vl:7b vision model
2. Set imageModel.primary to ollama/qwen2.5vl:7b
3. Send an image via Telegram DM to the bot
4. Bot responds with text model (qwen3-coder) saying it cannot see images

Expected Behavior

When an image is sent via Telegram, the gateway should:

1. Detect the image attachment
2. Switch to the configured imageModel (qwen2.5vl:7b)
3. Pass the image to the vision model for analysis

Actual Behavior

• The text model (qwen3-coder) responds instead of the vision model
• Logs show no mention of media, photo, image, or imageModel
• The bot says "I'm receiving images but cannot see them"

Verification

The Ollama /v1/chat/completions endpoint does support images. This works correctly:

$imageBase64 = [Convert]::ToBase64String([System.IO.File]::ReadAllBytes("image.png"))
$body = @{
    model = "qwen2.5vl:7b"
    messages = @(@{
        role = "user"
        content = @(
            @{ type = "image_url"; image_url = @{ url = "data:image/png;base64,$imageBase64" } },
            @{ type = "text"; text = "What do you see?" }
        )
    })
} | ConvertTo-Json -Depth 10

Invoke-RestMethod -Uri "http://127.0.0.1:11434/v1/chat/completions" -Method Post -ContentType "application/json" -Body $body
# Returns 450+ prompt tokens, confirming image was processed

Logs

embedded run start: runId=... provider=ollama model=qwen3-coder:latest thinking=off messageChannel=telegram

No logs showing image detection or imageModel switching.

Related

• [telegram] No user ack on media fetch failure #4662 - [telegram] No user ack on media fetch failure