Pinned Loading

1. tool-call-guard tool-call-guard Public

   Deny-by-default policy gate for AI agent tool calls — allowlists, validation, rate caps, approval hooks, audit trail. One policy model, enforced in JavaScript and Python.

   Python 1

2. vitest-mcp vitest-mcp Public

   Vitest harness and matchers for testing MCP servers — in-memory transport, real protocol path, millisecond tests.

   TypeScript

3. react-native-local-llm react-native-local-llm Public

   One interface for every on-device LLM runtime in React Native — Apple Foundation Models, ExecuTorch (Gemma), and llama.cpp-based Cactus — with device-aware model selection and unified streaming.

   TypeScript 1

4. llm-context-budget llm-context-budget Public

   Fit chat history into any model's context window — sliding-window packing, running summaries, overflow detection. Zero dependencies, no AI inside. Python sibling of @yanib/context-budget.

   Python

5. context-budget context-budget Public

   Fit chat history into any model's context window — sliding-window packing, running summaries, overflow detection. Zero dependencies, no AI inside.

   TypeScript

6. ai-sdk-cache ai-sdk-cache Public

   Record/replay cache middleware for the Vercel AI SDK — free dev iterations, offline deterministic tests. Streams included.

   TypeScript 2