Forge adds guardrails that boost small LLMs to near‑perfect task accuracy
Forge is an open‑source reliability layer for self‑hosted LLM tool‑calling that lifts an 8B model’s accuracy on agentic tasks from 53% to 99%. It offers a drop‑in proxy, workflow runner, or middleware, and supports Ollama, llama.cpp, vLLM and Anthropic backends.