mnemo

Local-first AI memory layer with a knowledge graph and CLI

TLDR

Store a memory chunk for later retrieval

$ mnemo ingest "I prefer Neovim with dark mode"

Search for relevant context to inject into a prompt

$ mnemo search "what editor do I use?"

List extracted entities

$ mnemo entities

Show one entity and its graph neighbors

$ mnemo entity <uuid> --neighbors

Check server and database health

$ mnemo health

Wipe all stored memory (with confirmation)

$ mnemo wipe --yes

SYNOPSIS

mnemo subcommand [arguments] [options]

mnemo is the command-line client for a local AI memory service. It talks to mnemo-api (default `http://localhost:8080`), which ingests text, uses an OpenAI-compatible LLM to extract entities and relationships, stores them in SQLite, and maintains an in-memory petgraph knowledge graph for multi-hop retrieval.Typical workflow: run mnemo-api (or `docker compose up` from the project), ingest conversations or notes with mnemo ingest, then call mnemo search to obtain a ranked context_prompt string suitable for injection into any LLM application. Backends include Ollama (fully local), OpenAI, Anthropic, or any OpenAI-compatible API.The server exposes REST endpoints (`/ingest`, `/retrieve`, `/entities`, `/chunks`, `/search`, `/stats`, `/wipe`). The CLI is a thin wrapper around those endpoints using blocking HTTP.

PARAMETERS

ingest text

POST text to `/ingest` for entity extraction and graph storage.

search query

Retrieve ranked memory context for a natural-language query.

entities

List stored entities (paginated on the server).

entity uuid [--neighbors]

Show one entity; optional graph neighbor traversal.

chunks

List raw memory chunks.

health

Report API, database, and LLM connectivity.

stats

Print entity, chunk, and graph counts plus uptime.

wipe [--yes]

Delete all memory; skips the confirmation prompt when --yes is set.

--server url

Use a non-default API base URL (default `http://localhost:8080`).

CONFIGURATION

The API reads environment variables (or `mnemo.example.toml` via --config):

Variable	Default	Purpose
MNEMO_DB_PATH	mnemo.db	SQLite database file
MNEMO_PORT	8080	API listen port
MNEMO_LLM_BASE_URL	http://localhost:11434/v1	OpenAI-compatible LLM endpoint
MNEMO_LLM_MODEL	llama3	Model for extraction
MNEMO_LLM_API_KEY	ollama	API key (any value works for Ollama)
MNEMO_LLM_PROVIDER	ollama	ollama, openai, anthropic, or custom

Environment variables override TOML values.

CAVEATS

mnemo requires mnemo-api to be running. Entity extraction quality depends on the configured LLM. mnemo wipe is irreversible. Install the CLI with `cargo install --path crates/mnemo-cli` from the upstream repository.