visual hearth
Hey Krill/team — quick question. Is the new native Codex runtime config expected to be working now, or is it still being worked on?
I migrated Telegram direct-chat agents on macOS from the stable PI route:
openai-codex/gpt-5.5 + OpenClaw Pi Default
to native Codex:
model.primary = "openai/gpt-5.5"
agentRuntime.id = "codex"
Fresh canary sessions sometimes show the expected state:
Model: openai/gpt-5.5
Runtime: OpenAI Codex
agentHarnessId: codex
fallbackUsed: false
But in real Telegram direct sessions, some agents become slow or non-responsive and show a mixed state:
Model: openai-codex/gpt-5.5
Runtime: OpenAI Codex
Observed logs include:
codex app-server turn idle timed out waiting for completion
selected model unavailable
I also saw fallback/auto-pin behavior:
openai/gpt-5.5 -> openai-codex/gpt-5.5
When I tried migrating multiple agents to the native Codex setup, the problem became worse: several Telegram direct-chat agents hung or became very slow, so I had to revert them back to the stable openai-codex/gpt-5.5 + OpenClaw Pi Default route.
Setup: macOS 26.4.1, OpenClaw 2026.5.7 eeef486, npm global install via Homebrew prefix, Node v25.9.0.
Is there a recommended config or workaround for this?