We've added an interactive terminal UI that activates automatically in real terminals to provide real-time monitoring of metrics like uptime, memory usage, and cache hit ratios. This new interface reuses existing in-process getters for data accuracy, while maintaining backward compatibility for piped or non-interactive environments to ensure CI and logging workflows remain uninterrupted. Users can toggle this off with --no-tui or the CACHEPOT_NO_TUI environment variable.

This update introduces a hybrid HNSW index for collections larger than 256 vectors, achieving sub-linear query times and ~3.9x faster performance compared to the previous brute-force approach. Additionally, we have added a RESP-compatible CLI and benchmarking tool, implemented TLS support for secure connections, and updated our documentation with a proposed clustering design. These changes mark a significant step forward in both production utility and architectural maturity.

This update introduces atomic transaction blocks (MULTI/EXEC/DISCARD) and optimistic concurrency control via WATCH, allowing for safer state updates. The implementation includes per-key version tracking to detect concurrent modifications, ensuring that updates are atomic and consistent. By keeping watch-version tracking gated behind an atomic active-count, the system maintains high performance for non-transactional writes on the hot path.

This significant release introduces a robust embedded web dashboard for real-time observability, including tools for browser, workbench, and profiling access. The backend now features AOF (Append Only File) persistence, a fully integrated RESP protocol parser, and expanded engine capabilities like SCAN, RENAME support, and advanced server introspection via commands like MONITOR and SLOWLOG. These enhancements dramatically improve the debuggability and data durability of the cache server.

Cache-Pot v0.1.0 is here, bringing a fast, single-binary Redis-compatible data store specifically designed for AI agents. This initial release includes support for core RESP2 commands, built-in vector search, a semantic cache to reduce LLM costs, and a native MCP server for agent tool integration. It is ready for your local AI experimentation and development workflows.

Cache-Pot is introduced as a fast, drop-in Redis-compatible data store specifically designed for AI applications. Beyond standard key-value storage, it includes native vector search, semantic caching to optimize LLM calls, and a built-in MCP server for AI agents to easily read, write, and search data. It aims to provide an accessible, lightweight alternative to standard Redis for AI-focused workflows.

Cache-Pot is a new, fast in-memory data store designed specifically for AI agents and applications. It acts as a drop-in Redis replacement that adds built-in semantic caching, vector search capabilities, and a native Model Context Protocol (MCP) server for direct agent integration. It's built as a single, dependency-free Go binary for easy deployment.