All plugins

entropicmem

Durable local memory for Hermes agents: SQLite FTS5 facts with cited, explainable recall; a Markdown vault with knowledge-graph export and an interactive visual graph; and turn-prefetch context injection with reason tokens. Write policy auto-redacts high-confidence secrets at write time (other PII types are warn-only findings) and stored-content injection screening flags and warns on recall and prefetch payloads (flag-and-warn, fail-open by design, not a blocking filter). Includes a 34-command CLI, destructive-op gates with automatic backups, and an append-only audit log. Disclosure — all memory data (vault markdown, memory.db, index.db, backups) stays local under HERMES_HOME/entropicmem/; a shared publish DB defaults to HERMES_HOME/entropicmem-shared/. The CLI init command appends ENTROPICMEM_* settings to HERMES_HOME/.env (append-only). Vector embeddings are off by default; only with embeddings_enabled: true (provider) or ENTROPICMEM_EMBEDDINGS=1 / query --semantic (CLI) and the optional sentence-transformers extra installed does first use download all-MiniLM-L6-v2 weights from huggingface.co; otherwise the provider has no network egress. The CLI ingest command fetches a user-supplied URL; CLI graph export writes to ./export in the cwd by default and its HTML loads D3/marked from d3js.org and cdn.jsdelivr.net when opened; CLI graph serve starts a loopback HTTP server. entropicmem_patch_core edits the plugin's own Core-memory markdown (Persona/User_Profile), not Hermes core, and is off by default.

Ufonik882Memoryv2.8.1Updated 1 month ago

About

Durable local memory for Hermes agents: SQLite FTS5 facts with cited, explainable recall; a Markdown vault with knowledge-graph export and an interactive visual graph; and turn-prefetch context injection with reason tokens. Write policy auto-redacts high-confidence secrets at write time (other PII types are warn-only findings) and stored-content injection screening flags and warns on recall and prefetch payloads (flag-and-warn, fail-open by design, not a blocking filter). Includes a 34-command CLI, destructive-op gates with automatic backups, and an append-only audit log. Disclosure — all memory data (vault markdown, memory.db, index.db, backups) stays local under HERMES_HOME/entropicmem/; a shared publish DB defaults to HERMES_HOME/entropicmem-shared/. The CLI init command appends ENTROPICMEM_* settings to HERMES_HOME/.env (append-only). Vector embeddings are off by default; only with embeddings_enabled: true (provider) or ENTROPICMEM_EMBEDDINGS=1 / query --semantic (CLI) and the optional sentence-transformers extra installed does first use download all-MiniLM-L6-v2 weights from huggingface.co; otherwise the provider has no network egress. The CLI ingest command fetches a user-supplied URL; CLI graph export writes to ./export in the cwd by default and its HTML loads D3/marked from d3js.org and cdn.jsdelivr.net when opened; CLI graph serve starts a loopback HTTP server. entropicmem_patch_core edits the plugin's own Core-memory markdown (Persona/User_Profile), not Hermes core, and is off by default.

EntropicMem MemoryProvider

Hermes Agent memory provider: local SQLite facts, vault retrieval, and prefetch injection, wired as memory.provider: entropicmem. The engine and CLI live in scripts/ inside this directory, so the plugin is self-contained for catalog installs.

Install

Catalog (recommended):

hermes plugins install entropicmem

Development checkout:

ln -sf "/path/to/EntropicMem/plugins/entropicmem" ~/.hermes/plugins/entropicmem

Activate

# ~/.hermes/config.yaml
memory:
  provider: entropicmem

Then bootstrap a vault:

python3 ~/.hermes/plugins/entropicmem/scripts/entropicmem.py init

Tools

| Tool | Purpose |
|------|---------|
| entropicmem_remember | Store a durable fact (policy + PII sanitized) |
| entropicmem_recall | Search facts, FTS5 + hybrid vector fusion, why_retrieved on every hit |
| entropicmem_query | Hybrid vault retrieval with citations (wikilinks + FTS) |
| entropicmem_patch_core | Surgical Persona / User Profile edits (opt-in via core_memory_writable) |
| entropicmem_stats | Memory statistics: fact count, domain distribution, DB path |
| entropicmem_get | Retrieve a single stored fact by entropic_id |
| entropicmem_consolidate | Archive old, low-access facts (gated) |

Hooks

on_memory_write, on_session_switch, on_session_end, on_turn_start, on_pre_compress. Behavior and config keys: docs/ARCHITECTURE.md.

Memory-Context Injection (Critical)

When active, Hermes injects prefetched memories wrapped in <memory-context> tags at the start of each turn. These are system-injected context, not user input. The agent MUST:

  • Use the facts silently to inform answers
  • NEVER acknowledge the block itself
  • NEVER say "thanks for sharing," "I notice you've shared," or similar
  • NEVER ask "what do you want me to do with this context"
  • If a memory-context block appears with no actual user message, ask the user what they need

Smart Context Management

Prefetch relevance filtering, token budgets, and dedup windows optimize context injection. Configure under plugins.entropicmem in ~/.hermes/config.yaml; defaults and the full option list are in skills/entropicmem/references/HERMES_INTEGRATION.md.

README from the reviewed commit 7e02412; it updates when the author re-pins.

More by Ufonik88

2 plugins by Ufonik88 in the catalog.

Ufonik88vaultknoxSecret-leak protection for Hermes chats. Scans inbound messages and outbound replies for API keys, tokens, and passwords with a 28-pattern detector registry, redacts them before they reach session storage or the user, and injects safe secret-handling rules so the agent never asks for secrets in chat. Zero dependencies, fully offline detection; pairs with the optional hermes-vault CLI for encrypted storage. Disclosure — rewrites inbound user messages (regex secret redaction, may false-positive on password-like strings), injects a secret-handling prompt each conversation and rewrites assistant replies that ask for secrets; outbound replies are value-redacted with the same registry (regex-based, so a shape no detector covers passes through), with findings logged as detector name plus SHA-256 fingerprint and never the raw value. The vaultknox tool is inert unless the separate vaultknox pip package is installed, its write gate is a tool argument the model sets, and a vault policy that authorises raw consume_token returns the plaintext to the model.Communitytools1