Ariadne is a memory provider for Hermes Agent: fast vector + keyword search, hybrid retrieval, knowledge graph, and deduplication. It exposes its own ariadne_* memory tools.

---

Option 1 — Autonomous Setup (Recommended)

Paste this entire block into Hermes as a single message. It contains everything Hermes needs — what Ariadne is, where to get it, how to install, and how to migrate:

Replace my memory provider with Ariadne. Here's everything you need:

ARIADNE
- PyPI: https://pypi.org/project/ariadne-memory/
- GitHub: https://github.com/kyssta-exe/Ariadne
- Docs: https://ariadne.mantes.net
- Plugin: https://github.com/kyssta-exe/Ariadne/tree/main/plugin

WHAT IT IS
Ariadne is a local memory system for AI agents. It uses FAISS for vector search, FTS5 for keywords, Reciprocal Rank Fusion to combine both, a knowledge graph with typed edges and multi-hop traversal, MinHash LSH deduplication, and an Ebbinghaus forgetting curve for retention. No cloud, no daemon, no API keys.

INSTALL
1. Install the Python package: pip install ariadne-memory
2. Download the plugin files from https://github.com/kyssta-exe/Ariadne/tree/main/plugin and save them to ~/.hermes/plugins/ariadne/__init__.py and ~/.hermes/plugins/ariadne/plugin.yaml
   (Or clone the repo and copy: git clone https://github.com/kyssta-exe/Ariadne.git /tmp/ariadne && cp -r /tmp/ariadne/plugin ~/.hermes/plugins/ariadne)
3. Set the provider: hermes config set memory.provider ariadne
4. Migrate existing memories: if ~/.hermes/mnemosyne/data/mnemosyne.db exists, export all memories from it and import them into ~/.hermes/ariadne/memory.db using the ariadne CLI or the ariadne_import tool
5. Restart: hermes restart

DO THIS NOW. Install it, migrate my memories, and switch to Ariadne as my memory provider. After restarting, run ariadne_stats to confirm Ariadne is active.

Hermes will:

1. Install ariadne from PyPI
2. Download and place the plugin files
3. Switch the config to memory.provider: ariadne
4. Export your existing Mnemosyne memories and import them into Ariadne
5. Restart itself
6. Verify with ariadne_stats

After it finishes, send "Check memory status" — you should see "engine": "Ariadne" in the output.

---

Option 2 — Manual Setup

Step-by-step if you prefer to do it yourself.

Step 1 — Install the Python package

pip install ariadne-memory

Step 2 — Download the plugin

git clone https://github.com/kyssta-exe/Ariadne.git /tmp/ariadne-repo
mkdir -p ~/.hermes/plugins/ariadne
cp /tmp/ariadne-repo/plugin/__init__.py ~/.hermes/plugins/ariadne/
cp /tmp/ariadne-repo/plugin/plugin.yaml ~/.hermes/plugins/ariadne/
rm -rf /tmp/ariadne-repo

Step 3 — Switch the provider

hermes config set memory.provider ariadne

Or edit ~/.hermes/config.yaml directly:

memory:
  provider: ariadne

Step 4 — Migrate existing memories (optional)

If you were using Mnemosyne before and have memories to keep, export them to JSON and import with the CLI:

ariadne migrate ~/mnemosyne_export.json --db-path ~/.hermes/ariadne/memory.db

Step 5 — Restart

hermes restart

Step 6 — Verify

Send to Hermes: "Check memory status"

You should see:

{
  "engine": "Ariadne v0.1.2",
  "active_memories": 234,
  "graph_nodes": 58,
  "graph_edges": 60
}

If it says engine: Ariadne — you're done.

---

What Changes

	Before (Mnemosyne)	After (Ariadne)
Vector search	sqlite-vec (row scan)	FAISS (Flat → IVF)
Hybrid search	❌	FTS5 + FAISS + RRF
Dedup	❌	MinHash LSH near-duplicates
Knowledge graph	Basic edges	Typed edges + multi-hop traversal
Retention	❌	Ebbinghaus forgetting curve

Ariadne exposes its own memory tools (ariadne_remember, ariadne_recall, ariadne_graph_query, ariadne_sleep, etc.) through the standard Hermes MemoryProvider interface, so the natural-language experience is unchanged — you just talk to Hermes and it manages memory for you.

---

Reverting

Switch back to Mnemosyne:

hermes config set memory.provider mnemosyne
hermes restart

Or manually:

mv ~/.hermes/plugins/ariadne ~/.hermes/plugins/ariadne.disabled
# Edit ~/.hermes/config.yaml: set provider back to mnemosyne
hermes restart

---

File Layout

~/.hermes/
├── plugins/
│   └── ariadne/
│       ├── __init__.py     # MemoryProvider implementation
│       └── plugin.yaml     # Plugin metadata + tool schemas
└── ariadne/
    ├── memory.db           # SQLite: memories, graph, metadata, embeddings
    └── shared/
        └── memory.db       # Cross-agent shared surface

The FAISS vector index is rebuilt from the embeddings stored in memory.db on startup, so there is no separate index file to back up or keep in sync.

Back up with: cp -r ~/.hermes/ariadne ~/backup/ariadne-$(date +%Y%m%d)