Ai

AstrBot v4.23.6 Adds Token Usage Stats and Firecrawl Search Tools

AstrBot v4.23.6 Adds Token Usage Stats and Firecrawl Search Tools

AstrBot v4.23.6 delivers a focused product update centered on observability, web tool expansion, and compatibility fixes across AI provider and messaging workflows. The release adds a new /stats command for session-level token tracking, introduces Firecrawl-based web search and page extraction capabilities, and tightens several edge-case behaviors involving reasoning fields, image handling, TTS output, and provider interoperability.

What Changed

The headline addition in v4.23.6 is the new /stats command, which lets users inspect token consumption for the current session. The breakdown includes total usage as well as input from cache, other input, and output, giving operators a clearer view into model cost and interaction patterns.

Another notable enhancement is support for Firecrawl web search and webpage extraction tools. This expands AstrBot’s ability to work with live web content by adding search-result processing, page extraction, session management, request validation, and related test coverage.

On the experience side, the release improves the Provider configuration interface with better performance, more responsive rendering, and refinements to typography and layout. The statistics page also gets a logic update: TPM calculations now count output tokens only, with the wording adjusted to reflect that change more accurately.

AstrBot also improves compatibility in OpenAI-compatible provider handling by filtering empty assistant messages more consistently. Both streaming and non-streaming paths now treat empty strings and empty list content in a unified way, reducing the chance that strict providers will reject message history.

Why It Matters

For teams operating AI assistants in production, the new token statistics command is especially useful because it gives immediate visibility into usage patterns without requiring external monitoring. That makes it easier to troubleshoot prompt inflation, estimate costs, and understand how much of a session is being driven by cached versus fresh input.

The Firecrawl integration is equally important because web search and extraction are becoming core capabilities for modern AI agents. By baking these tools directly into the release, AstrBot strengthens its position as a more capable orchestration layer for assistants that need to retrieve, parse, and act on live web content.

The bug fixes in v4.23.6 are also meaningful for reliability. The update addresses DeepSeek v4 reasoning-content handling, corrects an OpenRouter reasoning field naming issue, improves oversized image compression behavior, changes MiniMax TTS default output from MP3 to WAV to avoid QQ platform conversion errors, and fixes image sending behavior in the Computer sandbox. Together, these changes reduce integration friction across providers and channels.

Overall, AstrBot v4.23.6 is less about sweeping architectural change and more about making the platform more observable, more web-capable, and more dependable in edge-case production scenarios. For existing users, the practical value lies in tighter token visibility, improved provider compatibility, and a smoother operational experience across AI and messaging workflows.

Official Source: https://github.com/AstrBotDevs/AstrBot/releases/tag/v4.23.6