context-chef-middleware

What problem does it solve?

This Skill helps you stop AI SDK prompts from growing too large by automatically compressing chat history, truncating oversized tool outputs, and keeping token usage within your configured budget.

Core Features & Use Cases

• Zero-code-change wrapping: wrap your existing Vercel AI SDK model once and keep using generateText or streamText as before.
• History compression: use a cheaper model to summarize older messages when the prompt exceeds the token budget.
• Tool result truncation: preserve head/tail portions of large tool outputs (optionally with storage for retrieval) to reduce hallucination risk and context bloat.
• Optional prompt enhancements: inject dynamic state and apply a transform hook after compression for RAG or metadata.

Quick Start

Ask the Skill to integrate context-chef middleware into your Vercel AI SDK v6+ project, then reuse your current generateText/streamText calls while wrapping your model with withContextChef using compression and truncation.