openclaw - 💡(How to fix) Fix [Bug]: openai-codex/gpt-5.5 returns 'No text output' via --local on v2026.5.7 (regression of #76464) [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
openclaw/openclaw#80086Fetched 2026-05-11 03:18:54
View on GitHub
Comments
1
Participants
2
Timeline
2
Reactions
2
Timeline (top)
closed ×1commented ×1

openclaw infer model run --model openai-codex/gpt-5.5 --local returns Error: No text output returned for provider "openai-codex" model "gpt-5.5" on v2026.5.7, despite #76464 being closed and PR #76470 merged on 2026-05-06. The same call succeeds with --gateway, and the agent loop succeeds (verified on a separate Linuxbrew host where openai-codex is the configured primary).

Error Message

$ openclaw --version 2026.5.7 $ openclaw plugins list --json | jq -r '.[] | select(.name=="@openclaw/codex") | "(.name) (.version) (.status)"' @openclaw/codex 2026.5.7 loaded $ openclaw models auth list | grep -i openai-codex

Bug: --local fails

$ openclaw infer model run --model openai-codex/gpt-5.5 --local --prompt "Reply with PONG only" Error: No text output returned for provider "openai-codex" model "gpt-5.5".

Same model, --gateway works

$ openclaw infer model run --model openai-codex/gpt-5.5 --gateway --prompt "Reply with PONG only" PONG

Root Cause

openclaw infer model run --model openai-codex/gpt-5.5 --local returns Error: No text output returned for provider "openai-codex" model "gpt-5.5" on v2026.5.7, despite #76464 being closed and PR #76470 merged on 2026-05-06. The same call succeeds with --gateway, and the agent loop succeeds (verified on a separate Linuxbrew host where openai-codex is the configured primary).

Fix Action

Fix / Workaround

The CLI returns Error: No text output returned for provider "openai-codex" model "gpt-5.5" from the --local path. The --gateway path returns clean text using the same OAuth profile, same plugin version, same model. The agent loop also returns clean text. Only the in-process --local dispatch path emits the empty-response wrapper error.

openclaw -> openai-codex (OAuth) -> Codex Responses API. No proxy, no gateway in the routing chain (--local runs in-process in the CLI's own pi-runtime instance; --gateway dispatches via the running gateway and works).

  • agents.defaults.agentRuntime.id: pi
  • Codex plugin version: @openclaw/[email protected]
  • Two OAuth profiles present: openai-codex:[email protected] (healthy until 2026-05-19) and openai-codex:default (recently expired). The [email protected] profile is the one used at probe time.
  • Behrooz host fallback chain has openai-codex/gpt-5.5 at FB1; chain dispatches via gateway and works.

Code Example

$ openclaw --version
2026.5.7
$ openclaw plugins list --json | jq -r '.[] | select(.name=="@openclaw/codex") | "\(.name) \(.version) \(.status)"'
@openclaw/codex 2026.5.7 loaded
$ openclaw models auth list | grep -i openai-codex
- openai-codex:kaveh@360innovate.co.uk (kaveh@360innovate.co.uk) [openai-codex/oauth; expires 2026-05-19T04:18:59.041Z]

# Bug: --local fails
$ openclaw infer model run --model openai-codex/gpt-5.5 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.5".

# Same model, --gateway works
$ openclaw infer model run --model openai-codex/gpt-5.5 --gateway --prompt "Reply with PONG only"
PONG

---

$ openclaw infer model run --model openai-codex/gpt-5.5 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.5".

$ openclaw infer model run --model openai-codex/gpt-5.5 --gateway --prompt "Reply with PONG only"
PONG

$ openclaw infer model run --model openai-codex/gpt-5.4 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.4".
RAW_BUFFERClick to expand / collapse

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

openclaw infer model run --model openai-codex/gpt-5.5 --local returns Error: No text output returned for provider "openai-codex" model "gpt-5.5" on v2026.5.7, despite #76464 being closed and PR #76470 merged on 2026-05-06. The same call succeeds with --gateway, and the agent loop succeeds (verified on a separate Linuxbrew host where openai-codex is the configured primary).

Steps to reproduce

$ openclaw --version
2026.5.7
$ openclaw plugins list --json | jq -r '.[] | select(.name=="@openclaw/codex") | "\(.name) \(.version) \(.status)"'
@openclaw/codex 2026.5.7 loaded
$ openclaw models auth list | grep -i openai-codex
- openai-codex:[email protected] ([email protected]) [openai-codex/oauth; expires 2026-05-19T04:18:59.041Z]

# Bug: --local fails
$ openclaw infer model run --model openai-codex/gpt-5.5 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.5".

# Same model, --gateway works
$ openclaw infer model run --model openai-codex/gpt-5.5 --gateway --prompt "Reply with PONG only"
PONG

Same --local failure also reproduces with openai-codex/gpt-5.4.

Expected behavior

Per #76464 / PR #76470 (merged 2026-05-06, before v2026.5.7 was published 2026-05-07), infer model run --local should return text from openai-codex/* models when a healthy OAuth profile is present. The PR fixed the instructions are required Codex Responses API error by populating systemPrompt in the local-capability path.

Actual behavior

The CLI returns Error: No text output returned for provider "openai-codex" model "gpt-5.5" from the --local path. The --gateway path returns clean text using the same OAuth profile, same plugin version, same model. The agent loop also returns clean text. Only the in-process --local dispatch path emits the empty-response wrapper error.

OpenClaw version

2026.5.7 (eeef486)

Operating system

Ubuntu 24.04

Install method

npm global at /usr/local, system Node v22.22.2 (apt, held)

Model

openai-codex/gpt-5.5 (also reproduces on openai-codex/gpt-5.4)

Provider / routing chain

openclaw -> openai-codex (OAuth) -> Codex Responses API. No proxy, no gateway in the routing chain (--local runs in-process in the CLI's own pi-runtime instance; --gateway dispatches via the running gateway and works).

Additional provider/model setup details

  • agents.defaults.agentRuntime.id: pi
  • Codex plugin version: @openclaw/[email protected]
  • Two OAuth profiles present: openai-codex:[email protected] (healthy until 2026-05-19) and openai-codex:default (recently expired). The [email protected] profile is the one used at probe time.
  • Behrooz host fallback chain has openai-codex/gpt-5.5 at FB1; chain dispatches via gateway and works.

Logs, screenshots, and evidence

$ openclaw infer model run --model openai-codex/gpt-5.5 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.5".

$ openclaw infer model run --model openai-codex/gpt-5.5 --gateway --prompt "Reply with PONG only"
PONG

$ openclaw infer model run --model openai-codex/gpt-5.4 --local --prompt "Reply with PONG only"
Error: No text output returned for provider "openai-codex" model "gpt-5.4".

Note: openai-codex/gpt-5.3-codex-spark returns a clean error (gpt-5.3-codex-spark is no longer exposed by the OpenAI or Codex catalogs. Use openai/gpt-5.5.) — this is unrelated to the present bug and is correct behavior for a decommissioned model.

Impact and severity

  • Affected: anyone using infer model run --local as a diagnostic primitive against openai-codex/* models on v2026.5.7. Includes per-host probe scripts that default to --local.
  • Severity: low (workaround is --gateway), but is a regression of a fix that was explicitly delivered in v2026.5.5 / merged 2026-05-06.
  • Frequency: 100% of --local calls on this host. Reproduces on multiple hosts (this report from Ubuntu 24.04 apt-Node host; cross-checked against a Linuxbrew node@24 host where the same agent configuration uses codex as primary).
  • Consequence: probe-driven health checks silently skip codex models. Hosts using codex appear "broken" in --local CLI probes despite working fine via gateway.

Additional information

  • Last known good version per #76464 / PR #76470 commit history: should be everything from 2026.5.5 onward.
  • First known bad: 2026.5.7 on this host. v2026.5.5 / 5.6 not directly tested for this surface.
  • Workaround: use --gateway or the agent path (openclaw agent --message ...).
  • Related: this bug does NOT affect fallback chains (which dispatch via gateway), so production behavior is unaffected. Only the diagnostic primitive is broken.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

Per #76464 / PR #76470 (merged 2026-05-06, before v2026.5.7 was published 2026-05-07), infer model run --local should return text from openai-codex/* models when a healthy OAuth profile is present. The PR fixed the instructions are required Codex Responses API error by populating systemPrompt in the local-capability path.

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

openclaw - 💡(How to fix) Fix [Bug]: openai-codex/gpt-5.5 returns 'No text output' via --local on v2026.5.7 (regression of #76464) [1 comments, 2 participants]