[Release] openclaw/openclaw v2026.4.23: openclaw 2026.4.23

OpenClaw v2026.4.23 introduces significant enhancements to image generation across multiple providers (OpenAI, OpenRouter), adds optional forked context for subagents, implements per-call timeout support for generation tools, and includes configurable local embedding context sizes. The release also fixes numerous issues related to Codex harness routing, media handling, provider authentication, and streaming behavior across multiple platforms including WhatsApp, Slack, and Telegram.

Key Points

• •Enable image generation without API keys using Codex OAuth for OpenAI and OpenRouter, expanding provider flexibility
• •Add optional forked context for subagent sessions to allow child agents to inherit requester transcripts while maintaining isolated default sessions
• •Implement per-call `timeoutMs` support for image, video, music, and TTS generation tools to handle long-running provider requests
• •Introduce configurable `memorySearch.local.contextSize` (default 4096) for tuning local embeddings on constrained hosts
• •Fix Codex harness routing for native `request_user_input` prompts and preserve queued follow-up answers
• •Resolve Windows npm-installed `codex.cmd` shims through PATHEXT for proper model routing
• •Suppress verbose tool/plan progress on Slack non-DM surfaces to prevent internal traces from leaking into rooms
• •Fix block streaming to prevent duplicate replies when partial delivery aborts match final assembled text
• •Preserve image attachments for text-only models by offloading as media references instead of dropping them
• •Honor explicit image-model configuration before native-vision skips, including agent defaults and provider-specific settings

Workflow Diagram

Concepts