Help developers stop wasting LLM tokens.

I built SuperCompress: open-source prompt compression for production AI apps. It cuts long chat history, RAG chunks, support logs, and tool traces before the LLM call. Why it matters: - fewer input tokens - lower prefill cost - less context-window pressure - no GPU required Try it: https://supercompress.dev/dashboard?signup=1

Show HN: SuperCompress, open-source prompt compression for LLM apps I built a small query-aware prompt compression engine for production AI apps. It runs before inference and removes low-value context from RAG chunks, chat history, tool traces, support transcripts, and logs. The point is not summarization. It selects context against the current user request, then returns the compressed prompt plus token-savings and risk metadata. Free API key: https://supercompress.dev/dashboard?signup=1 Repo: https://gitlab.com/arjunkshah/supercompress

I made an open-source prompt compressor for AI apps with growing context. If your app sends RAG chunks, chat history, tool output, support tickets, or JSON blobs into an LLM, SuperCompress can sit before the model call and trim the prompt while keeping query-relevant evidence. It has a browser demo, hosted API, Python client, and integration examples. Free API key: https://supercompress.dev/dashboard?signup=1

Subject: Cut LLM context cost before inference Hi, I built SuperCompress, an open-source prompt compression layer for AI apps with long context. It compresses chat history, RAG chunks, support logs, tool traces, and JSON context before the model call, then returns token-savings and risk metadata. Free hosted API key: https://supercompress.dev/dashboard?signup=1 Repo: https://gitlab.com/arjunkshah/supercompress Best, Arjun