engram-ms 0.0.2

Engram memory system for AI agents — engram-named distribution backed by memorytrace.

engram-ms

The engram-named distribution of the engram memory system for AI agents — SQLite + FTS5 backend with optional MCP, LLM, and semantic-search integrations.

engram-ms is a thin alias on top of memorytrace, which carries the real engine code. Installing engram-ms automatically pulls in memorytrace, so both names give you the same engine.

Install

pip install engram-ms

Optional extras (forwarded to the corresponding memorytrace extras):

pip install "engram-ms[llm]"        # Anthropic + OpenAI clients
pip install "engram-ms[semantic]"   # numpy + embeddings
pip install "engram-ms[mcp]"        # Model Context Protocol server
pip install "engram-ms[full]"       # everything above

Requires Python 3.9+.

Usage

from engram_ms import MemoryEngine, EngramConfig

with MemoryEngine(EngramConfig()) as engine:
    engine.create_entity("Alice", entity_type="person")
    engine.add_fact("Alice", "role", "engineer")
    print(engine.timeline("Alice"))

The full API lives under the canonical engram namespace:

from engram.engine import MemoryEngine
from engram.integrations.sdk import EngramSDK

Relationship to memorytrace

Same engine, two distribution names:

Distribution	Role
engram-ms	Engram-named alias (preferred going forward)
memorytrace	Original/legacy distribution; still maintained

Either install gives you the engram import path, the engram / engram-advanced CLIs, and the same SDK. Pick whichever name you prefer in requirements.txt.

License

Released under the MIT License.