csm Releases

A critical security patch addressing a specific CVE vulnerability.

Feb 24, 2026

Feb 21, 2026

This release introduces major integrations with Neo4j and Elasticsearch, along with new agents for data analysis and web search, and adds sandbox support for DB-GPT.

Feb 11, 2026

A major update adding batch execution capabilities and a Laboratory interface, while stabilizing function calling and supporting legacy Chat interfaces.

Jan 22, 2026

This release introduces the ability to pass OpenAPI tags to mounted endpoints and refactors logging by replacing print statements with a proper logger to reduce output noise.

Nov 24, 2025

Nov 15, 2025

Introduces DSQ (DeepSeek Quantization) to significantly reduce VRAM/RAM usage for DeepSeek-OCR and PaddleOCR-VL by applying compact patches to heavy linear layers while preserving critical components in float precision.

Nov 14, 2025

Fixes a critical memory issue on Blackwell GPUs where the RoPE (Rotary Positional Embedding) cache was incorrectly expanding to an excessively large size.

Nov 10, 2025

Adds the PaddleOCR-VL backend, implements lazy loading for weights to reduce startup time, and standardizes sampling controls across the CLI and server.

Nov 9, 2025

This release adds support for GLM4.5 models and BurnCloud as a new AI provider, while fixing critical bugs related to serialization, database connections, and document parsing.

Oct 24, 2025

The update ensures compatibility with older versions of Gradio Lite, specifically handling versions prior to 5.46.0 to maintain broader support.

Sep 17, 2025

This release focuses on critical bug fixes including an Async Client issue, input handling errors, and UI component stability, alongside a license update and dependency relaxation.

Sep 2, 2025

The release introduces an immersive UI experience, adds a new compiled JavaScript library to the source tree, and includes a demonstration for GPT-OSS.

Aug 12, 2025

This release focuses on ML capabilities by adding support for MLX inference and new data sources like GaussDB, alongside improvements to embeddings and API integrations.

Jul 25, 2025

Users can now programmatically close connections using Python code, providing better control over the application lifecycle and resource management.

Jul 7, 2025

This release adds support for new AI models including OpenAI o1, Gemini 2.0, and GPT-4o-2024-11-20, while enhancing infrastructure capabilities with Redis Sentinel and Cluster support.

Feb 2, 2025

Jan 15, 2025

Dec 24, 2024

Introduces support for new providers including SiliconFlow, Groq, and xAI, alongside OIDC authentication and enhanced Ollama embedding capabilities.

Dec 22, 2024

Nov 21, 2024