Skip to content

thread.start fails with "Internal error" (502 command_failed) when using Ox Alpha via omp harness #2460

Description

@fabricioereche

Summary

Starting a thread with the Ox Alpha model (via OpenRouter) through omp as the harness fails immediately: every turn shows a generic "Internal error" in the UI and thread.start fails server-side with HTTP 502 / JSON-RPC error -32000 ("Internal error", command_failed, retryable: false). Expected the thread to start and the model to respond.

Versions and environment

  • bb 0.39.0 via npx bb-app@latest
  • Windows 11 Pro, running under WSL2
  • omp 18.0.5 as the harness
  • Provider: OpenRouter, model: Ox Alpha

Steps to reproduce

  1. Run npx bb-app@latest under WSL2 on Windows 11.
  2. Connect omp 18.0.5 as the harness with OpenRouter / Ox Alpha as the provider/model.
  3. Send any message in a new thread (e.g. hello).

Result: every message fails with "Command thread.start failed — Internal error". Reproduced on 4 consecutive messages in the same thread.

Expected vs actual

Expected: thread starts and Ox Alpha responds.

Actual — UI (screenshot): every turn renders an error row "Command thread.start failed" with the generic message Internal error:

Actual — bb terminal log:

[21:13:11] WARN: [server] Live ready turn command failed {"threadId":"thr_5xjai772cq"}
    err: {
      "type": "ApiError",
      "message": "Internal error",
      "stack":
          Error: Internal error
              at callHostOnlineRpcWithRetry (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/server/dist/start-server.js:48317:11)
              at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
              at async runLiveHostCommand (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/server/dist/start-server.js:221632:20)
      "status": 502,
      "body": {
        "type": "Object",
        "message": "Internal error",
        "code": "command_failed",
        "retryable": false
      }
    }
[21:13:11] WARN: [host-daemon] online host RPC failed {"serverUrl":"http://127.0.0.1:38886","type":"thread.start"}
    err: {
      "type": "hB",
      "message": "Internal error",
      "stack":
          JsonRpcResponseError: Internal error
              at MZr (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:49:3384)
              at gtt (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:50:678)
              at rt (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:53:3215)
              at ro (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:53:4829)
              at Object.handleStdoutLine (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:52:16978)
              at Object.onLine (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:52:3906)
              at s (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:49:280)
              at Socket.<anonymous> (file:///home/fabricioereche/.npm/_npx/614ebd23ff24de90/node_modules/bb-app/host-daemon/dist/daemon-bundle.mjs:49:51)
              at Socket.emit (node:events:518:28)
              at addChunk (node:internal/streams/readable:561:12)
      "code": -32000,
      "name": "JsonRpcResponseError"
    }

Evidence

  • Thread id: thr_5xjai772cq
  • Screenshot of the thread timeline showing four consecutive "Command thread.start failed / Internal error" turns (attached).
  • The host daemon's JSON-RPC thread.start call returns code -32000 ("Internal error") from the host worker; the server surfaces it as HTTP 502 command_failed, non-retryable.

What you ruled out

  • Not transient: failed on 4 consecutive messages in the same thread.
  • The failure is at thread startup (thread.start), before any model response — the provider request itself never appears to be reached.

Suggested priority and effort (optional)

Blocks all use of Ox Alpha via omp on this setup; no workaround found.

AGENT GENERATED

Metadata

Metadata

Assignees

No one assigned

    Labels

    ompOh My Piprovider-acpBuilt-in plugin: provider-acpprovidersCross-provider bridges, models, login

    Type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions