Mnemosyne memory provider for Hermes Agent — local-first, zero-cloud memory
Project description
Mnemosyne for Hermes
Local-first memory provider for Hermes Agent. 23 tools. Zero cloud. Zero latency.
Mnemosyne is a Hermes-native memory provider that stores everything locally — SQLite with vector search, hybrid recall, episodic consolidation, and a temporal knowledge graph. No API keys. No cloud. No network calls. Your memory stays on your machine.
Quick Start
pip install mnemosyne-hermes
hermes memory setup # select "mnemosyne"
# Or manually:
hermes config set memory.provider mnemosyne
That's it. Hermes discovers the plugin automatically.
What You Get
- 23 memory tools.
remember,recall,sleep,validate,graph_query,triple_add,scratchpad_write, and more. All surfaced through the Hermes tool system. - Hybrid search. Vector similarity + FTS5 full-text + temporal scoring. Tunable per-query.
- Episodic consolidation.
mnemosyne_sleepcompresses working memory into long-term summaries — keeps context small, recall sharp. - Knowledge graph.
mnemosyne_triple_add/mnemosyne_triple_queryfor structured facts.mnemosyne_graph_querytraverses linked memories via BFS. - Multi-agent validation.
mnemosyne_validatelets agents attest, update, or invalidate each other's memories with provenance tracking. - Shared surface.
mnemosyne_shared_rememberstores compact cross-agent metadata.
Configuration
No required config. Everything defaults to ~/.mnemosyne/. Optional overrides:
| Variable | Default | Description |
|---|---|---|
MNEMOSYNE_HOME |
~/.mnemosyne |
Storage directory |
MNEMOSYNE_VEC_WEIGHT |
0.5 |
Vector similarity weight in hybrid recall |
MNEMOSYNE_FTS_WEIGHT |
0.3 |
Full-text search weight |
MNEMOSYNE_IMPORTANCE_WEIGHT |
0.2 |
Importance score weight |
MNEMOSYNE_AUTO_SLEEP_ENABLED |
false |
Auto-consolidate after N turns |
MNEMOSYNE_AUTO_SLEEP_THRESHOLD |
50 |
Turns between auto-consolidation |
MNEMOSYNE_PROFILE_ISOLATION |
false |
Separate DB per Hermes profile |
Links
- Mnemosyne GitHub — core library, benchmarks, docs, BEAM ICLR 2026
- Hermes Memory Providers — full comparison table
- Hermes Plugin Guide — developer docs
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file mnemosyne_hermes-0.1.0.tar.gz.
File metadata
- Download URL: mnemosyne_hermes-0.1.0.tar.gz
- Upload date:
- Size: 35.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.4
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
76ed3e61abf24d4ac839ada4c8f8fc07ca8b42664d4876cd241815235b6828f3
|
|
| MD5 |
3700329e4c85978f932ec904eb19372f
|
|
| BLAKE2b-256 |
ed53883b22721f9538627c6a859964be8bf60cd9879e39b404dafcf06623706a
|
File details
Details for the file mnemosyne_hermes-0.1.0-py3-none-any.whl.
File metadata
- Download URL: mnemosyne_hermes-0.1.0-py3-none-any.whl
- Upload date:
- Size: 36.2 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.4
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
dd10b081821be000549aa4a46fd46b51e82af0f7ea3ab0b80d90f5f36759214a
|
|
| MD5 |
2c4b951d1d740cefcf22417dc94f7476
|
|
| BLAKE2b-256 |
d7e0713bb19e4fdea02243511caf4d008de403aec64eafdb32f770e921415024
|