llm-wiki

name	llm-wiki
description	Use when configuring or using llmwiki as a shared agent knowledge base, including API-free markdown mode, OpenAI-compatible extraction, local/OAuth bridge extraction, and optional embedding endpoints.
version	0.1.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["llmwiki","knowledge-base","embeddings","rag","markdown"],"related_skills":["hermes-memory-stack"]}}

llmwiki for Hermes Memory Stack

Overview

llmwiki is the shared agent knowledge layer. It is for structured, reusable knowledge that multiple agents can consult: project architecture, operating rules, manuals, concepts, and references.

It can run in API-free markdown mode or with OpenAI-compatible extraction and embedding endpoints.

Modes

1. API-free markdown mode

   • Use markdown files, wikilinks, indexes, and Hermes file search.
   • No external API required.
   • Semantic search and automatic extraction are limited.

2. Extraction only

   • Use a chat/completions endpoint to extract concepts and summaries.
   • Embeddings disabled or handled separately.

3. Local/OAuth bridge extraction

   • Route extraction through a local OpenAI-compatible bridge that represents an existing provider/OAuth login.

4. Hybrid

   • Extraction endpoint and embedding endpoint can be different.
   • Example: local bridge for extraction, LiteLLM for embeddings.

Secret Handling

Keep keys in ~/.hermes/.env, never in wiki files or committed config.

Recommended env names:

LLMWIKI_API_KEY=
LLMWIKI_EMBEDDINGS_API_KEY=

NotebookLM Boundary

NotebookLM is not a replacement for llmwiki. Use llmwiki for internal/shared agent knowledge. Use NotebookLM only for optional source-grounded external research packs.

Verification Checklist

• API-free mode works with plain markdown and Hermes file tools.
• If extraction is enabled, the configured endpoint responds.
• If embeddings are enabled, the embedding endpoint and model are compatible.
• Secrets are only in .env.