engram-mem 0.5.18

Memory traces for AI agents - Think like human

engram

Memory traces for AI agents — Think like humans.

Dual-memory AI system combining episodic (vector) + semantic (graph) memory with LLM reasoning. Entity-gated ingestion ensures only meaningful data is stored. Enterprise-ready with multi-tenancy, auth, caching, observability, and Docker deployment.

Exposes CLI, MCP (Claude Code / OpenClaw), HTTP API (/api/v1/), and WebSocket (/ws) interfaces.

pip install engram-mem

---

Features

Core Memory

• Episodic Memory — Qdrant vector store (embedded or server), semantic similarity search, Ebbinghaus decay, activation-based scoring, topic-key upsert
• Semantic Graph — NetworkX MultiDiGraph, typed entities and relationships, SQLite (default) or PostgreSQL backend, weighted edges
• Reasoning Engine — LLM synthesis (Gemini via litellm), dual-memory context fusion, constitution-guarded prompts
• Recall Pipeline — Query decision, temporal+pronoun entity resolution, parallel multi-source search, dedup, composite scoring
• Entity-Gated Ingestion — Only stores messages with extracted entities; skips noise (system prompts, trivial messages)
• Auto Memory — Detect and persist save-worthy messages automatically, poisoning guard for injection prevention
• Meeting Ledger — Structured meeting records with decisions, action items, attendees, topics
• Feedback Loop — Confidence scoring (+0.15/-0.2), importance adjustment, auto-delete on 3x negative feedback
• Graph Visualization — Interactive entity relationship explorer with dark theme, search, click-to-inspect (vis-network)

Intelligence Layer

• Temporal Resolution — 28 Vietnamese+English date patterns resolve "hom nay/yesterday" to ISO dates before storing
• Pronoun Resolution — "anh ay/he/she" to named entity from graph context, LLM-based fallback
• Fusion Formatter — Group recall results by type [preference]/[fact]/[lesson] for structured LLM context
• Memory Consolidation — Jaccard clustering + LLM summarization reduces redundancy

Enterprise

• Multi-Surface — CLI (Typer), MCP Server (stdio), HTTP API (FastAPI), WebSocket, Web UI
• Authentication — JWT + API keys with RBAC (ADMIN, AGENT, READER), optional, disabled by default
• Multi-Tenancy — Isolated per-tenant stores, contextvar propagation, row-level PostgreSQL isolation
• Caching — Redis-backed result caching with per-endpoint TTLs
• Rate Limiting — Sliding-window per-tenant limits, fail_open option
• Audit Trail — Structured before/after JSONL log for every episodic mutation
• Resource Tiers — 4-tier LLM degradation (FULL > STANDARD > BASIC > READONLY), 60s auto-recovery
• Data Constitution — 3-law LLM governance (namespace isolation, no fabrication, audit rights), SHA-256 tamper detection
• Consolidation Scheduler — Asyncio background tasks (cleanup daily, consolidate 6h, decay daily), tier-aware
• Key Rotation — Failover/round-robin for embedding API keys (GEMINI_API_KEY + GEMINI_API_KEY_FALLBACK)
• Observability — OpenTelemetry + JSONL audit logging (optional)
• Deployment — Docker Compose, Kubernetes-ready, health checks
• Backup/Restore — Memory snapshots, point-in-time recovery
• Benchmark Suite — p50/p95/p99 latency measurements for all endpoints

---

Architecture

flowchart TD
    CLI["CLI (Typer)"]
    MCP["MCP (stdio)"]
    HTTP["HTTP API /api/v1/"]
    WS["WebSocket /ws"]

    CLI & MCP & HTTP & WS --> Auth["Auth Middleware\n(JWT + RBAC, optional)"]
    Auth --> Tenant["TenantContext (ContextVar)"]
    Tenant --> Recall["Recall Pipeline\n(decision > resolve > search > feedback)"]
    Recall --> Episodic["EpisodicStore\n(Qdrant)"]
    Recall --> Semantic["SemanticGraph\n(NetworkX + SQLite/PG)"]
    Episodic & Semantic --> Reasoning["Reasoning Engine\n(Gemini via litellm)"]
    Episodic --> Cache["Redis Cache (optional)"]
    WS --> EventBus["Event Bus\n(push events)"]

---

Quick Start

# Install from PyPI
pip install engram-mem

# Or from source
git clone https://github.com/docaohieu2808/Engram-Mem.git
cd engram && pip install -e .

# Initialize config
engram init

# Set API key
export GEMINI_API_KEY="your-key"

# Start daemon (background HTTP server + watcher)
engram start

# Store a memory
engram remember "Deployed v2.1 to production at 14:00 - caused 503 spike"

# Search memories
engram recall "production incidents"

# Browse all data (episodic + semantic)
engram dump

# Reason across all memory
engram think "What deployment issues have we had?"

Requirements: Python 3.11+, GEMINI_API_KEY for LLM reasoning and embeddings. Basic storage works without it.

---

Integrations

Claude Code (MCP)

Add to ~/.claude.json:

{
  "mcpServers": {
    "engram": {
      "command": "engram-mcp",
      "env": { "GEMINI_API_KEY": "your-key" }
    }
  }
}

OpenClaw

Install the engram skill, then enable session watcher in ~/.engram/config.yaml:

capture:
  openclaw:
    enabled: true
    sessions_dir: ~/.openclaw/workspace/sessions

HTTP API

# Start server
engram serve --port 8765

# Store memory
curl -X POST http://localhost:8765/api/v1/remember \
  -H "Content-Type: application/json" \
  -d '{"content": "Deployed v1.0", "memory_type": "fact", "priority": 8}'

# Search
curl "http://localhost:8765/api/v1/recall?query=deployment&limit=5"

# Reason
curl -X POST http://localhost:8765/api/v1/think \
  -H "Content-Type: application/json" \
  -d '{"question": "What deployment issues have we had?"}'

# Meeting ledger
curl -X POST http://localhost:8765/api/v1/meeting-ledger \
  -H "Content-Type: application/json" \
  -d '{"title": "Sprint Review", "decisions": ["Ship v2"], "action_items": ["Update docs"]}'

---

CLI Reference

Memory Operations

# Store with options
engram remember <content> [--type fact|decision|preference|todo|error|context|workflow|meeting_ledger]
                          [--priority 1-10] [--tags tag1,tag2] [--expires 2h|1d|7d]
                          [--topic-key unique-key]

# Search
engram recall <query> [--limit 5] [--type <type>] [--tags tag1,tag2]
              [--resolve-entities] [--resolve-temporal]

# Smart query (auto-routes to recall or think)
engram ask <question>

# Reason across all memory
engram think <question>
engram summarize [--count 20] [--save]

Semantic Graph

engram add node <name> --type <NodeType>
engram add edge <from_key> <to_key> --relation <relation>
engram remove node <key>
engram query [<keyword>] [--type <NodeType>] [--related-to <name>] [--format table|json]
engram autolink-orphans [--apply] [--min-co-mentions 3]
engram graph                          # Open interactive graph visualization

Browse & Export

engram status                         # Summary counts
engram dump                           # Rich tables: all memories, nodes, edges
engram dump --format json             # Full JSON export
engram decay [--limit 20]             # Ebbinghaus decay report

System

engram init                           # Initialize config
engram start                          # Start daemon (HTTP server + watcher)
engram stop                           # Stop daemon
engram serve [--host 0.0.0.0] [--port 8765]  # Foreground server
engram watch [--daemon]               # Watch inbox + OpenClaw sessions
engram health                         # Full system health check
engram resource-status                # Resource tier (FULL/STANDARD/BASIC/READONLY)
engram queue-status                   # Embedding queue status
engram scheduler-status               # Background task schedule
engram constitution-status            # 3-law governance + SHA-256 hash

Maintenance

engram cleanup                        # Delete expired memories
engram consolidate [--limit 50]       # LLM-driven memory consolidation
engram ingest <file.json> [--dry-run] # Ingest chat JSON
engram backup                         # Export memory snapshot
engram restore <file>                 # Import snapshot
engram config show / get <key> / set <key> <value>
engram feedback <id> --positive|--negative

---

MCP Tools

Tool	Description
engram_remember	Store memory with type, priority, tags, namespace
engram_recall	Search episodic memories (compact format by default)
engram_think	Reason across episodic + semantic memory via LLM
engram_status	Show memory statistics
engram_get_memory	Retrieve full memory content by ID or prefix
engram_timeline	Chronological context around a memory
engram_add_entity	Add entity node to knowledge graph
engram_add_relation	Add relationship edge between entities
engram_query_graph	Query knowledge graph
engram_ingest	Dual ingest: extract entities + store memories
engram_meeting_ledger	Record structured meeting (decisions, action items)
engram_feedback	Record positive/negative feedback on memories
engram_auto_feedback	Auto-detect feedback from conversation
engram_cleanup	Delete all expired memories
engram_cleanup_dedup	Deduplicate similar memories by cosine similarity
engram_summarize	Summarize recent N memories via LLM
engram_session_start	Begin new conversation session
engram_session_end	End active session with optional summary
engram_session_summary	Get summary of completed session
engram_session_context	Retrieve memories from active session
engram_ask	Smart query — auto-routes to recall or think

---

Configuration

Config file: ~/.engram/config.yaml — Priority: CLI flags > env vars > YAML > defaults

episodic:
  mode: embedded              # embedded (Qdrant in-process) or server
  path: ~/.engram/qdrant
  namespace: default

embedding:
  provider: gemini
  model: gemini-embedding-001
  key_strategy: failover      # failover or round-robin

semantic:
  provider: sqlite            # or postgresql
  path: ~/.engram/semantic.db

llm:
  provider: gemini
  model: gemini/gemini-2.0-flash
  api_key: ${GEMINI_API_KEY}

serve:
  host: 127.0.0.1
  port: 8765

capture:
  openclaw:
    enabled: false
    sessions_dir: ~/.openclaw/workspace/sessions
  claude_code:
    enabled: false
    sessions_dir: ~/.claude/projects

auth:
  enabled: false
cache:
  enabled: false
  redis_url: redis://localhost:6379/0
rate_limit:
  enabled: false
audit:
  enabled: false
  path: ~/.engram/audit.jsonl

---

API Reference

Start server: engram serve [--host 0.0.0.0] [--port 8765]

Method	Endpoint	Purpose
GET	/health	Liveness check
GET	/health/ready	Readiness probe
POST	/api/v1/remember	Store episodic memory
GET	/api/v1/recall	Search memories (?query=X&limit=5&offset=0)
POST	/api/v1/think	LLM reasoning across episodic + semantic
GET	/api/v1/query	Graph search (?keyword=X&node_type=Y&related_to=Z)
GET	/api/v1/memories	List/filter memories with pagination
GET	/api/v1/memories/export	Export memories as JSON
POST	/api/v1/meeting-ledger	Record structured meeting
POST	/api/v1/ingest	Extract entities + store memories
POST	/api/v1/feedback	Record feedback on a memory
GET	/api/v1/graph/data	Graph data JSON for visualization
GET	/graph	Interactive graph visualization UI
POST	/api/v1/cleanup	Delete expired memories (admin)
POST	/api/v1/cleanup/dedup	Deduplicate memories (admin)
POST	/api/v1/summarize	LLM summary of recent memories (admin)
GET	/api/v1/status	Memory statistics

---

WebSocket API

Connect via ws://host:8765/ws?token=JWT (token optional when auth disabled).

Commands:

Command	Payload
remember	{"content": "...", "priority": 7}
recall	{"query": "...", "limit": 5}
think	{"question": "..."}
feedback	{"memory_id": "abc123", "feedback": "positive"}
query	{"keyword": "PostgreSQL"}
ingest	{"messages": [...]}
status	{}

Push Events: memory_created, memory_updated, memory_deleted, feedback_recorded

---

Environment Variables

Variable	Purpose
GEMINI_API_KEY	LLM + embeddings (primary key)
GEMINI_API_KEY_FALLBACK	Secondary key for key rotation
ENGRAM_NAMESPACE	Memory namespace isolation
ENGRAM_AUTH_ENABLED	Enable JWT auth
ENGRAM_SEMANTIC_PROVIDER	sqlite or postgresql
ENGRAM_CACHE_ENABLED	Enable Redis caching
ENGRAM_AUDIT_ENABLED	Enable audit logs
ENGRAM_TELEMETRY_ENABLED	Enable OpenTelemetry

---

Docker

# Quick start
docker build -t engram:latest .
docker run -e GEMINI_API_KEY="your-key" -p 8765:8765 engram:latest

# Production with PostgreSQL + Redis
ENGRAM_AUTH_ENABLED=true \
ENGRAM_SEMANTIC_PROVIDER=postgresql \
ENGRAM_SEMANTIC_DSN=postgresql://user:pass@postgres:5432/engram \
ENGRAM_CACHE_ENABLED=true \
ENGRAM_CACHE_REDIS_URL=redis://redis:6379/0 \
docker compose up

---

Testing

pytest tests/ -v                      # All tests
pytest tests/ --cov=src/engram        # With coverage
pytest tests/ -k "recall or feedback" # Specific suites

894+ tests, 61%+ code coverage, CI/CD via GitHub Actions.

---

Documentation

• Project Overview & PDR
• System Architecture
• Code Standards
• Deployment Guide
• Codebase Summary
• Project Roadmap
• Changelog

---

License

MIT — Copyright (c) Do Cao Hieu