Issue Station
Sign In
Tools hereToolbox

openclaw - 💡(How to fix) Fix [Bug]: Upgrade to 2026.3.28 leaves assistant runtime unstable until re-onboard / re-add API key [1 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#57383•Fetched 2026-04-08 01:50:23
View on GitHub
Comments
0
Participants
1
Timeline
1
Reactions
0
Author
Manet147-clManet147-cl
Participants
Manet147-clManet147-cl
Timeline (top)
labeled ×1

After upgrading OpenClaw to 2026.3.28 on macOS, the gateway stayed up and the config still validated, but the assistant runtime became unstable/unusable until I reconfigured the setup and re-added my OpenAI API key. This did not look like data loss. Workspace files, memory, sessions, and cron jobs were still present. The problem appeared to be in runtime config/auth/fallback migration after the update.

Root Cause

Impact observed:

  • assistant became unreliable/unusable until reconfiguration
  • provider fallback chain cascaded across failures
  • existing auth state became effectively untrustworthy
  • recovery required re-onboarding/reconfiguration and re-adding the OpenAI API key
  • cron jobs and gateway stayed present, which made the issue harder to diagnose because the system looked healthy at a surface level
RAW_BUFFERClick to expand / collapse

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

After upgrading OpenClaw to 2026.3.28 on macOS, the gateway stayed up and the config still validated, but the assistant runtime became unstable/unusable until I reconfigured the setup and re-added my OpenAI API key. This did not look like data loss. Workspace files, memory, sessions, and cron jobs were still present. The problem appeared to be in runtime config/auth/fallback migration after the update.

Steps to reproduce

  1. Start from an existing OpenClaw setup that was working before upgrade, with:

    • previously configured providers/auth
    • a non-trivial fallback chain (for example Anthropic + OpenAI Codex + OpenAI)
    • existing sessions, cron jobs, memory, and Control UI usage

  2. Upgrade OpenClaw to 2026.3.28 on macOS.

  3. Restart the gateway / open the Control UI and try to use the assistant normally.

  4. Observe that:

    • gateway appears healthy
    • openclaw config validate still passes
    • workspace files, memory, cron jobs, and sessions are still present

  5. Send a normal prompt in the assistant and inspect logs.

  6. Observe runtime instability such as:

    • fallback cascade across providers/models
    • Anthropic rate-limit failures
    • OpenAI Codex token refresh/auth failures
    • model execution errors related to unsupported thinking values like none

  7. Attempt to recover normal assistant behavior.

  8. Notice that the practical recovery path is to:

    • re-run onboarding or reconfigure providers
    • re-add the OpenAI API key
    • simplify/fix the fallback model setup

  9. After that, observe that the assistant becomes usable again (for example on openai/gpt-5.1-codex).

Expected behavior

After upgrading: existing valid provider config should continue working, or fail in a clearer/self-healing way stale fallback/auth state should not poison the assistant runtime unsupported thinking defaults should be migrated automatically or sanitized the assistant should remain usable without needing a fresh onboard unless credentials are actually missing

Actual behavior

After upgrading: system looked healthy at a surface level but assistant runtime became unstable/unusable recovery required reconfiguration / re-adding API key

OpenClaw version

2026.3.28

Operating system

macOS 26.3.1 (arm64)

Install method

No response

Model

openai/gpt-5.1-codex anthropic/claude-sonnet-4-6

Provider / routing chain

anthropic/claude-sonnet-4-6 -> openai-codex/gpt-5.2 -> openai/gpt-5.1-codex -> openai-codex/gpt-5.4

Additional provider/model setup details

This was not a fresh install. It was an existing working setup upgraded in place to 2026.3.28.

Before/during failure:

  • mixed provider setup already existed
  • fallback chain was active across Anthropic + OpenAI Codex + OpenAI
  • gateway service stayed up
  • config remained valid
  • runtime showed provider auth and thinking-compatibility issues

After recovery:

  • fallback chain was simplified
  • OpenAI API key had to be re-added/reconfirmed
  • stable usage resumed on openai/gpt-5.1-codex

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

High

The upgrade did not appear to delete data, but it left the assistant runtime unstable enough to disrupt normal use.

Impact observed:

  • assistant became unreliable/unusable until reconfiguration
  • provider fallback chain cascaded across failures
  • existing auth state became effectively untrustworthy
  • recovery required re-onboarding/reconfiguration and re-adding the OpenAI API key
  • cron jobs and gateway stayed present, which made the issue harder to diagnose because the system looked healthy at a surface level

Severity rationale:

  • not destructive to stored data
  • but high operational impact because the assistant could no longer be used normally until manually repaired

extent analysis

Fix Plan

To resolve the issue, follow these steps:

  • Reconfigure the provider setup to simplify the fallback chain.
  • Re-add the OpenAI API key to ensure proper authentication.
  • Update the thinking defaults to supported values.

Example code snippet to update the provider configuration:

import openclaw

# Simplify the fallback chain
providers = [
    {"name": "anthropic/claude-sonnet-4-6", "priority": 1},
    {"name": "openai/gpt-5.1-codex", "priority": 2}
]

# Update the provider configuration
openclaw.config.update_providers(providers)

# Re-add the OpenAI API key
openclaw.config.update_api_key("openai", "YOUR_API_KEY")

# Update the thinking defaults
openclaw.config.update_thinking_defaults({"default": "openai/gpt-5.1-codex"})

Replace YOUR_API_KEY with your actual OpenAI API key.

Verification

After applying the fix, verify that the assistant runtime is stable and usable by:

  • Sending a normal prompt and checking the logs for any errors.
  • Verifying that the fallback chain is working correctly.
  • Checking that the cron jobs and gateway are still present and functional.

Extra Tips

To prevent similar issues in the future, consider:

  • Regularly backing up your configuration and data.
  • Testing updates in a non-production environment before applying them to your main setup.
  • Monitoring your system's logs and performance to quickly identify any potential issues.

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

After upgrading: existing valid provider config should continue working, or fail in a clearer/self-healing way stale fallback/auth state should not poison the assistant runtime unsupported thinking defaults should be migrated automatically or sanitized the assistant should remain usable without needing a fresh onboard unless credentials are actually missing

#api#serialization error#model compatibility#GPU setup#container setup

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]: Upgrade to 2026.3.28 leaves assistant runtime unstable until re-onboard / re-add API key [1 participants]