Sókrates Wiki

Hermes Agent Documentation Ingestion

2 min read

Hermes Agent Documentation Ingestion

Summary

The process of scraping, converting, and organizing the external Hermes Agent documentation into the local Sokrates repository. This initiative was undertaken to provide the development team with a searchable, offline-accessible, and modifiable reference for the Hermes Agent’s architecture, features, and configuration patterns.

Details

To facilitate the integration and customization of the Hermes Agent within the Sokrates project, the team executed a comprehensive documentation ingestion task. Using parallelized scraping agents, approximately 23 pages of documentation were retrieved and mirrored into the local directory structure at docs/hermes-agent/. This local documentation serves as the primary reference for modifying the agent’s behavior and understanding its interaction with the broader Sokrates stack.

Documentation Structure

The ingested documentation is organized into five primary categories:

  1. Getting Started: Covers installation, quickstart guides, and learning paths.
  2. User Guide: Detailed breakdowns of core features located in docs/hermes-agent/user-guide/features/.
    • Tools & Toolsets: Documentation on terminal backends, container security, background processes, and sudo execution.
    • Memory: Management of persistent memory via MEMORY.md and USER.md files, session search capabilities, and Honcho integration.
    • Skills System: The SKILL.md format, progressive disclosure of capabilities, and agent-managed skill hubs.
    • MCP (Model Context Protocol): Integration patterns for stdio/HTTP servers, tool filtering (include/exclude), and instructions for running Hermes itself as an MCP server.
    • Voice Mode: Configuration for CLI voice loops, Telegram/Discord gateways, and STT/TTS provider setups.
    • Personality & SOUL.md: The identity system, built-in presets, and the logic governing the prompt stack order.
    • Context Files: Discovery priority and the AGENTS.md hierarchy used for environmental awareness.
  3. Guides: Practical implementation patterns, including use-mcp-with-hermes.md, use-voice-mode-with-hermes.md, and use-soul-with-hermes.md.
  4. Developer Guide: Technical internals for extending the agent.
    • Architecture: Maps the agent loop, prompt system, provider resolution, and messaging gateways.
    • Adding Tools: Documents the “3-file modification pattern” (tool file, toolsets.py, and model_tools.py) and the async handler contract.
    • Creating Skills: Provides the decision matrix for choosing between a skill and a tool, and the directory structure for publication.
  5. Reference: Includes a full CLI command reference (covering chat, gateway, mcp, profiles, etc.) and troubleshooting FAQs.

Technical Implementation

The ingestion was performed by three agents working in parallel to ensure consistency across the site structure. While most pages were captured in full, the architecture documentation (architecture.md) was noted as being potentially summarized due to the processing model of the fetch tool, requiring manual verification against the source if deep internal mapping is required.

Graph View