OpenCode hangs indefinitely when using Ollama behind reverse proxy (streaming SSE not delivered)

[bgoniasa]

Description

opencode run hangs indefinitely when using a custom Ollama provider behind a reverse proxy (Easypanel/Traefik). The model responds correctly via non-streaming requests, but OpenCode uses streaming by default and never receives the SSE chunks.

Environment

• OpenCode: v1.18.7
• OS: Linux (Docker container)
• Model: qwen2.5-coder:7b via Ollama (remote, behind Easypanel reverse proxy)
• Provider config: @ai-sdk/openai-compatible with custom baseURL

Config

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Ollama Remote",
      "options": {
        "baseURL": "https://<host>/v1"
      },
      "models": {
        "qwen2.5-coder:7b": {
          "name": "Qwen 2.5 Coder 7B"
        }
      }
    }
  },
  "model": "ollama/qwen2.5-coder:7b"
}

What works

Non-streaming requests work perfectly:

curl -s https://<host>/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen2.5-coder:7b","messages":[{"role":"user","content":"say OK"}]}'
# Returns: {"choices":[{"message":{"content":"OK"}}]} OK

What doesn't work

Streaming requests return empty response through the reverse proxy:

curl -sN https://<host>/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen2.5-coder:7b","stream":true,"messages":[{"role":"user","content":"say OK"}]}'
# Returns: (empty)

OpenCode logs show it starts streaming but never receives any data:

level=INFO message=stream providerID=ollama modelID=qwen2.5-coder:7b small=false agent=build
level=INFO message="llm runtime selected" llm.runtime=ai-sdk llm.provider=ollama llm.model=qwen2.5-coder:7b
# ... hangs here indefinitely

Root cause

The reverse proxy (Easypanel, which uses Traefik) appears to buffer or not properly forward SSE chunks. This is a common issue with reverse proxies and Ollama deployments.

Feature request

Add a config option to disable streaming for a provider, or auto-fallback to non-streaming when streaming data isn't received within a timeout. This would make OpenCode compatible with Ollama deployments behind reverse proxies that don't handle SSE correctly.

Something like:

"options": {
  "baseURL": "https://<host>/v1",
  "stream": false
}

Workaround

None found. OpenCode is unusable with this Ollama setup.

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Is there a way to disable streaming mode? #785: Requests an option to disable streaming mode for proxy providers that don't support it — same feature request as described here
• OpenCode 1.4.3 hangs with local Ollama provider on simple prompts, while /v1/chat/completions works #22132: OpenCode hangs indefinitely with local Ollama provider via @ai-sdk/openai-compatible — nearly identical symptoms (same model qwen2.5-coder:7b, same hang behavior, same provider config)

If your issue is the same, please add a +1 or reaction there rather than opening a new one. If there's something distinct about your setup (the reverse proxy/SSE buffering angle specifically), feel free to note that in a comment on the existing issue.

[oxomichael]

I think I’m hitting the same problem, though my config is a bit different. I’m running llama.cpp behind an nginx reverse proxy. Everything responds fine on version 1.18.7, but upgrading to 1.18.8 causes requests to hang indefinitely.