[Release] openclaw/openclaw v2026.4.14: openclaw 2026.4.14

OpenClaw v2026.4.14 is a quality release focused on model provider improvements, particularly for GPT-5 family support and channel provider fixes. The release includes forward-compatibility for gpt-5.4-pro, enhanced Telegram forum topic handling, and significant performance refactors to the core codebase. Key fixes address Ollama timeout handling, Codex provider catalog validation, media tool model references, security allowlist enforcement, and various provider integrations including Slack, Google, and Discord.

Key Points

• •Added forward-compatibility support for gpt-5.4-pro with Codex pricing/limits and catalog visibility before upstream adoption
• •Fixed Ollama timeout inheritance by forwarding configured embedded-run timeout to global undici stream timeout tuning
• •Resolved Codex provider catalog validation by including apiKey in output to prevent silent model drops from models.json
• •Normalized model references in media tools (image/PDF) to properly recognize Ollama vision models
• •Enhanced Slack security by applying global allowFrom owner allowlist to channel block-action and modal interactive events
• •Implemented fail-closed behavior for local attachment path resolution via realpath to prevent allowlist bypass
• •Prevented model-facing gateway tool from enabling security-sensitive flags via config.patch/apply calls
• •Fixed Telegram forum topic persistence to session sidecar store for context retention across restarts
• •Improved Ollama streaming completions with stream_options.include_usage to report real usage metrics
• •Optimized doctor/plugins performance by caching external preferOver catalog lookups within each auto-enable pass

Workflow Diagram

Concepts