Hermes Memory Stack - Setup Guide¶
Source: Chukwuemeka001/hermes-memory-stack Stars: 0 ⭐ | License: Apache-2.0 Created: June 24, 2026 Category: Agent Infrastructure / Memory Management Tests: 228
1. What It Is¶
A modular memory OS layer for Hermes Agent that addresses the full memory lifecycle: semantic session retrieval, conservative auto-extraction, temporal versioning, and remediation for users with existing memory overload.
Component Status¶
| Component | Status | Description |
|---|---|---|
| Semantic retrieval | ✅ Built + verified | Session search via semantic indexing |
| Auto-extraction | ✅ Built + verified | Conservative memory auto-extraction with intake gate |
| Temporal versioning | ✅ Built + verified | Version-tracked memory with migration support |
| Export installer | 🟡 Draft | One-command install via install.sh |
| Curator/dreaming maintenance | 🟡 Planned | Remediation plan only |
| Remediation Area 1: state.db | ✅ Built + tested | State database remediation (explicit-run only) |
| Remediation Area 2: MEMORY.md audit | ✅ Built + tested | Memory file audit (read-only) |
| Remediation Area 3: pointer rewrite | ✅ Built + tested | Memory pointer rewrite (dry-run/render only) |
| Remediation Area 4: temporal migration | ✅ Built + tested | Temporal memory migration (verify read-only; sync gated) |
2. Prerequisites¶
- Hermes Agent with existing memory (state.db, MEMORY.md)
- Python 3.8+
3. Installation¶
git clone #repo-unavailable.git
cd hermes-memory-stack
bash install.sh
Verify¶
# Check semantic index
python3 scripts/semantic_index.py --health
# Run memory audit (read-only, safe)
python3 scripts/memory_audit.py
4. Core Modules¶
Semantic Retrieval (scripts/semantic_index.py)¶
Converts session history into searchable embeddings for natural-language recall. Run:
python3 scripts/semantic_query.py "what did we decide about the RAG architecture"
Auto-Extraction (scripts/memory_auto_extract*.py)¶
Conservatively extracts durable facts from sessions and writes them to memory. Uses hermes_memory_intake_gate.py to prevent garbage extraction.
Temporal Versioning (scripts/temporal_memory.py)¶
Tracks memory entries with timestamps and supports migration between versions:
python3 scripts/temporal_migrate.py --from v1 --to v2
Remediation Tools¶
For users with existing memory that needs cleanup:
| Tool | Function | Safety |
|---|---|---|
state_db_remediate.py |
Clean and optimize state.db | Explicit-run only |
memory_audit.py |
Audit MEMORY.md for issues | Read-only |
memory_rewrite.py |
Rewrite memory pointers | Dry-run/render only |
temporal_migrate_onboard.py |
Migrate to temporal memory | Verify read-only; sync gated |
5. CorpusIQ Use Cases¶
| Use Case | Application |
|---|---|
| Multi-agent memory sync | Semantic retrieval across CorpusIQ's agent fleet |
| Session knowledge extraction | Auto-extract operator insights and bug reports from agent sessions |
| Memory hygiene | Audit and remediate bloated MEMORY.md files across agent profiles |
| Temporal rollback | Version-track agent memories for compliance and debugging |
| Onboarding new agents | Install memory stack on new agent profiles for consistent recall |
6. Troubleshooting¶
| Symptom | Cause | Fix |
|---|---|---|
| Semantic index empty | Index not built | Run scripts/semantic_reindex.sh |
| Auto-extraction silent | Intake gate blocking | Check gate thresholds in hermes_memory_intake_gate.py |
| Migration fails | Schema version mismatch | Run temporal_migrate.py to align versions |
| Remediation tools won't run | Safety gates active | All remediation tools have explicit-run or dry-run safeties - review output before committing |
← Skills Catalog | June 24 Evening Discovery →
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.