Frequently asked questions about SuperLocalMemory V3.

SuperLocalMemory is a persistent memory system for AI assistants. It stores your decisions, bug fixes, project context, and preferences locally, then automatically provides them to your AI in future sessions. Your AI stops forgetting you.

Yes. SuperLocalMemory is open-source (GNU Affero General Public License v3.0 or later) and completely free. No usage limits, no credit system, no subscription. Forever.

Core memory is SQLite-backed inside the configured SLM data root. That root also contains configuration, logs, queues, models, and derived state. Mode C sends configured query or enrichment content to its provider; optional connectors, backup, and downloads have their own network behavior in every mode.

Run slm connect --list for the release's documented client names. MCP-compatible clients can also be configured manually, but a client is considered verified only when it passes the release integration matrix.

Mode A and Mode B work fully offline. Mode C requires internet for the cloud LLM.

• Python 3.11+ (required for V3 engine)
• Node.js 14+ (if installing via npm)
• Any supported IDE
• For Mode B: Ollama with a pulled model
• For Mode C: API key for your cloud LLM provider

# npm (recommended)
npm install -g superlocalmemory
slm setup
slm warmup    # Optional — pre-download embedding model

# or inside an activated Python virtual environment
python3 -m venv .venv
source .venv/bin/activate  # Windows PowerShell: .venv\Scripts\Activate.ps1
python -m pip install superlocalmemory
slm setup

npm install -g superlocalmemory@latest
# or, while the SLM virtual environment is active:
python -m pip install --upgrade superlocalmemory

No. Run slm migrate after updating. All memories, profiles, and settings are preserved. A backup is created automatically. See Migration from V2 for details.

When you start a conversation in your IDE, SuperLocalMemory automatically retrieves relevant memories and injects them into your AI's context. You do not need to call "recall" explicitly — it happens in the background via the MCP server.

slm remember "The deploy script needs AWS_REGION set to us-east-1"

• queryable means raw evidence and the SQLite relational/FTS projection are durable and recallable.
• enriching means a lease-owning worker is running configured derivation stages.
• complete means every declared derivation and configured projector succeeded.
• failed retains the raw evidence, error, attempt count, and retry timing; it is not silent data loss.

slm recall "deploy configuration"

slm trace "deploy configuration"

This shows per-channel scores (Semantic, BM25, Entity Graph, Temporal) for each result.

slm forget "search query"     # Delete matching memories (with confirmation)

• Mode A if you need privacy, compliance, or offline operation
• Mode B if you want composed answers and have a capable machine (16GB+ RAM)
• Mode C if you want maximum accuracy and cloud access is acceptable

Yes: slm mode a, slm mode b, or slm mode c. Your memories are shared across all modes.

The V3 paper provides published LoCoMo evidence carried into V3.7 architecture:

• 60.4% Mode A Raw across 10 conversations / 1,276 questions with zero-LLM answer construction.
• 74.8% Mode A Retrieval across the same scope with local retrieval and GPT-4.1-mini answer synthesis.
• 87.7% Mode C on Conv-30 / 81 questions with cloud embeddings and GPT-4.1-mini answer generation and judge.

The figures retain their original protocol scope; they are not a newly rerun 3.7 package benchmark. See the linked preprint for category results, ablations, and limitations.

No. Your database is a local file on your machine. It is not synced, uploaded, or shared with anyone — including us.

No software package certifies the complete deployment. SLM supplies local storage, erasure, provenance, retention, access-policy, and audit controls; applicability and sufficiency depend on the operator, use case, configuration, providers, and surrounding systems.

The database is a standard SQLite file at ~/.superlocalmemory/memory.db. You can copy it, back it up, or query it directly.

slm forget "query" deletes matching memories. To delete everything, remove the database: rm ~/.superlocalmemory/memory.db.

1. Check that SuperLocalMemory is running: slm status
2. Check that you have stored memories: slm recall "test"
3. Verify your IDE connection: restart the IDE after configuring MCP
4. Check the active profile: slm profile list

Try more specific queries. Use slm trace "query" to see which channels contribute — this helps diagnose whether the issue is semantic, keyword, or entity matching.

Use manual configuration. See IDE Setup for per-IDE config paths.

Open an issue at github.com/qualixar/superlocalmemory/issues.

Part of Qualixar | Created by Varun Pratap Bhardwaj