Hermes History Ingest - Setup Guide

Source: ar9av/obsidian-wiki (Community) Skill: hermes-history-ingest · Installs: 2.1K+ · Category: Memory / Knowledge Management Platform: Linux, macOS, Windows

Hermes History Ingest extracts knowledge from your Hermes agent history and distills it into an Obsidian wiki. It mines past sessions for durable knowledge, imports ~/.hermes/memories, and tracks what's been ingested via a manifest file to avoid duplicates.

Installation

npx skills add ar9av/obsidian-wiki@hermes-history-ingest

Prerequisites

Requirement	Details
Obsidian vault	Existing vault with .obsidian-wiki config or CWD-based setup
Hermes data	~/.hermes/memories/ and optionally session transcripts
Python 3.7+	For manifest and file processing

Configuration

The skill resolves configuration via a priority chain:

1. Inline @name override
2. Walk up CWD for .env file
3. ~/.obsidian-wiki/config
4. Interactive prompt setup

Key config values: - OBSIDIAN_VAULT_PATH - where the Obsidian wiki lives - HERMES_HISTORY_PATH - defaults to ~/.hermes

Ingest Modes

Append Mode (Default)

Checks .manifest.json at the vault root. Only processes: - Files not in the manifest (new memories, new session logs) - Files with modification time newer than ingested_at

Use for regular syncs.

Full Mode

Processes everything regardless of manifest. Use after wiki-rebuild or for a fresh import.

Data Sources

Ranked by knowledge value:

Priority	Source	Content
1 (Highest)	~/.hermes/memories/*.md	Curated persistent agent knowledge
2	~/.hermes/sessions/**/*.jsonl	Structured turn-by-turn transcripts
3 (Lowest)	~/.hermes/config.yaml	Metadata only (model prefs, paths)

Skip: .hub/ internals, skills/ directory (source material, not user knowledge).

Workflow

1. Survey: Scan HERMES_HISTORY_PATH and compare against .manifest.json
2. Classify: Mark each file as New, Modified, or Unchanged
3. Extract: Mine durable knowledge from new/modified files
4. Distill: Write to Obsidian wiki with proper links and tags
5. Update: Update .manifest.json with new ingested_at timestamps

Verification

After running: - Check .manifest.json for updated timestamps - Verify new wiki pages in Obsidian vault - Confirm no duplicate entries from previous runs

Related Skills

• Agent Memory Setup
• Memory Merger (GitHub Copilot)

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects