name: llmslim title: LLMSlim repo: https://github.com/Thanatos9404/llmslim sha: 34c6f80a28fce47309f6a18bffe0df7bf40d296f subdir: plugins/hermes description: "Context engine that compresses older tool output and turns locally with LLMSlim instead of asking a model for a summary: no LLM call, deterministic output, and the start of the prompt is sent unchanged, so prompt caching keeps working. Select it with context.engine: llmslim. Disclosure: installs llmslim==0.12.0 from PyPI (with numpy, scikit-learn and jsonschema); registers no tools, hooks or middleware; makes no network calls, runs no shell commands, needs no credentials, writes no files and sends no telemetry; reads its settings from config.yaml through Hermes's config loader. Token counts use a character estimate unless tiktoken is already installed in Hermes's environment, in which case tiktoken may download its encoding file once." maintainer: Thanatos9404 tier: community category: general requires_hermes: ">=0.21.7" docs_url: https://github.com/Thanatos9404/llmslim/blob/main/docs/integrations/HERMES.md version: "0.12.0" capabilities: provides_tools: [] provides_hooks: [] provides_middleware: [] requires_env: []