[Release] openclaw/openclaw v2026.4.25-beta.4: openclaw 2026.4.25-beta.4

OpenClaw v2026.4.25-beta.4 introduces major upgrades to text-to-speech (TTS) with new providers and chat-scoped controls, migrates plugin startup paths to a cold persisted registry for faster performance, expands OpenTelemetry observability across model calls and tool loops, hardens browser automation with safer URLs and iframe awareness, and strengthens installation/update processes across multiple platforms. The release also adds PWA/Web Push support to the control UI and includes new CLI tools for plugin management and diagnostics.

Key Points

• •TTS system receives comprehensive upgrade: `/tts latest` read-aloud, chat-scoped auto-TTS toggles (`/tts chat on|off|default`), per-agent and per-account overrides, plus 7 new providers (Azure Speech, Xiaomi, Local CLI, Inworld, Volcengine, ElevenLabs v3)
• •Plugin registry migrates to cold persisted storage, eliminating broad manifest scans during startup and making plugin updates, repairs, and provider discovery deterministic and faster
• •OpenTelemetry observability expands across model calls, token usage, tool loops, harness runs, exec processes, outbound delivery, context assembly, and memory pressure with bounded low-cardinality attributes
• •Browser automation improvements: safe tab URLs in agent responses, CDP-native role snapshots with iframe awareness, headless one-shot launch mode, and deeper browser doctor probes for slow hosts (e.g., Raspberry Pi)
• •Control UI adds PWA install support and Web Push notifications for Gateway chat, plus new TUI setup flows with context mode selection and shorter startup greeting
• •Installation/update hardening covers Windows, macOS, Linux, Docker, bundled plugin runtime dependencies, Node service restarts, LaunchAgent token rotation, and mixed-version gateway verification
• •Plugin hooks expanded with before-agent-finalize hooks, cron `jobId` context, bounded native permission fingerprints, and Codex MCP relay support
• •CLI enhancements: `openclaw plugins registry` for persisted-registry inspection, `openclaw browser start --headless` for one-shot launch, and faster `models list --all --provider` via installed plugin index
• •Diagnostics improvements: signal-specific OTLP endpoint overrides, W3C traceparent header propagation from model-call trace context, bounded telemetry exporter health diagnostics, and new bundled Prometheus plugin
• •Discord voice channel LLM override support via `channels.discord.voice.model` config, and generic `--background` flag for image generation CLI commands

Workflow Diagram

Concepts

Artifacts (6)

# Per-agent TTS override
agents:
  list:
    - id: agent-1
      tts:
        provider: azure-speech
        voice: en-US-AriaNeural

# Channel and account TTS overrides
channels:
  feishu:
    accounts:
      account-123:
        tts:
          provider: elevenlabs-v3
          voice: custom-voice-id

# Chat-scoped auto-TTS control
# Usage: /tts chat on|off|default
# Usage: /tts latest (read-aloud support)
# Usage: /tts audio, /tts status

# Safe tab URLs in agent responses
# CDP-native role snapshot fallback with iframe-aware refs

# One-shot local managed browser launch
openclaw browser start --headless

# Deep browser diagnostics for slow hosts
openclaw browser doctor --deep

# Raise CDP readiness timeouts for slower hosts (e.g., Raspberry Pi)
# Configure in config file: browser.cdp.readiness_timeout_ms

# Inspect persisted registry
openclaw plugins registry

# List plugins from cold persisted registry (default)
openclaw plugins list

# Refresh plugin index and cold registry
openclaw doctor --fix

# Explicit registry repair without startup rescan
openclaw plugins registry --refresh

# Speed up model listing for static manifest-backed providers
openclaw models list --all --provider <provider-id>

# Signal-specific OTLP endpoint overrides
diagnostics:
  otel:
    traces:
      otlp_endpoint: http://localhost:4317
    metrics:
      otlp_endpoint: http://localhost:4318
    logs:
      otlp_endpoint: http://localhost:4319

# Enable latest GenAI semantic conventions
env:
  OTEL_SEMCONV_STABILITY_OPT_IN: gen_ai_latest_experimental

# W3C traceparent propagation from model-call trace context
# Automatically propagated to provider transports

# Override LLM for Discord voice channel responses
channels:
  discord:
    voice:
      model: gpt-4-turbo  # Override default LLM
      # STT and TTS remain on existing media settings

# Generic background flag for image generation
openclaw infer image generate --background "blue gradient"
openclaw infer image edit --background "white"

# OpenAI-specific alias (still supported)
openclaw infer image generate --openai-background "blue gradient"

# FAL image generation with output format
openclaw infer image generate --output-format png|jpeg