Fabric v1.4.447 introduces support for Claude Opus 4.7, upgrades the Anthropic SDK to v1.37.0, and enables the 1M context window beta for the new model.
MemPalace v3.3.1 adds five new language locales, improves script-aware entity detection, hardens thread safety, and strengthens privacy defaults for local memory workflows.
crewAI 1.14.2 introduces checkpoint resume, diff and prune commands, checkpoint forking with lineage tracking, deploy validation improvements, and multiple security and reliability fixes.
Mem0 has published ts-v3.0.0-beta.2 for its Node SDK, signaling an early beta update for TypeScript users as the project continues refining its developer tooling.
Docling v2.89.0 adds explicit TikZ handling in the LaTeX backend, fixes OCR and DOCX parsing issues, improves pipeline cache stability, and trims Markdown debug overhead.
Claude Mem Release v12.1.6 resolves a critical regression that caused observation and summary writes to fail on Claude Code 2.1.109+ due to malformed setting flag handling in the bundled Agent SDK.
Daytona v0.167.0 introduces per-sandbox resource limits in region quotas, shifts code execution and environment processing from SDKs to the daemon, and expands documentation for OpenAI Agents SDK workflows and sandbox pause controls.
LobeHub has published nightly build v0.0.0-nightly.pr13850.8503 for PR #13850, focused on desktop validation, UI checks, and performance testing in a non-production environment.
LobeHub v2.1.52-canary.3 introduces a targeted fix for Claude Opus 4.7 by stripping unsupported temperature and top_p parameters, improving compatibility in the latest canary build.
LobeHub v2.1.52-canary.4 introduces two fixes focused on assistant markdown animation and localized system role injection, offering a small but meaningful usability update for testers.