Troubleshooting
Common errors and their exact fixes. If your issue isn't listed here, email [email protected] with the full error message and your Node.js version.
Licence Errors
The key was not found in Polar. Common causes: you used a sandbox key (sandbox keys only work against the sandbox API, not production), the key was typed incorrectly, or the key was revoked after a refund.
Copy the key exactly from your Polar purchase email. If you cannot find your email, retrieve it at polar.sh → Orders → License Keys.
Your licence is already active on 3 machines (the maximum). To activate on a new machine, deactivate an old one first.
node -e "require('vektor-slipstream/vektor-licence').deactivateMachine(process.env.VEKTOR_LICENCE_KEY)"
Go to polar.sh → customer portal → License Keys → remove an activation manually.
Need more than 3 machines? Enterprise licences (10 seats) available — contact [email protected].
Set the VEKTOR_LICENCE_KEY environment variable so it is always available:
export VEKTOR_LICENCE_KEY=YOUR-LICENCE-KEY-HERE
The cache file is at ~/.vektor/licence.json. If missing or corrupted, delete it and re-enter your key at the prompt.
Check your internet connection. Slipstream uses a 30-day offline cache — once activated, it continues working without network access for up to 30 days.
You haven't passed a licence key to createMemory(). Either pass it directly or set the environment variable:
export VEKTOR_LICENCE_KEY=VEKTOR-XXXX-XXXX-XXXX
const memory = await createMemory({ agentId: 'my-agent', licenceKey: process.env.VEKTOR_LICENCE_KEY, // ← add this });
The key doesn't exist in Polar's system. Check three things:
1. You're using the key from your purchase confirmation email — not a placeholder like VEKTOR-XXXX-XXXX-XXXX.
2. The key is complete — Polar keys are typically 36+ characters.
3. Your purchase was for Vektor Slipstream specifically. Keys are product-scoped.
If you believe your key is correct, email [email protected] with your order reference.
Vektor couldn't reach the Polar API (no internet, firewall, etc.) but found a valid cached validation from a previous run. The engine will start normally. The cache is valid for 30 days from the last successful validation.
ONNX / Embedding Errors
The ONNX model wasn't included in the install. This usually means the package was installed from a broken version. Reinstall:
npm uninstall -g vektor-slipstream npm install -g ./vektor-slipstream-1.5.8.tgz
Verify the model is present after install:
ls node_modules/vektor-slipstream/models/
You should see model_quantized.onnx (~23MB).
The ONNX runtime couldn't load. This is usually a Node.js version issue. Verify you're on Node 18+:
node --version # must be ≥ 18.0.0
If on an ARM machine (Raspberry Pi, Apple Silicon), try forcing CPU execution by setting the environment variable:
ONNX_EP=cpu node your-agent.js
SQLite / Database Errors
Two processes are writing to the same .db file simultaneously. Common causes:
1. Two agent instances with the same dbPath running at once.
2. A previous process crashed mid-write and left a .db-wal lock file.
# Remove stale lock files
rm memory.db-wal memory.db-shm
If running multiple agents, give each its own database file:
const agent1 = await createMemory({ agentId: 'agent-1', dbPath: './agent1.db' }); const agent2 = await createMemory({ agentId: 'agent-2', dbPath: './agent2.db' });
The sqlite-vec vector extension couldn't load for your platform. This is non-fatal — the engine will run without it, but vector similarity search falls back to JS-based cosine similarity (slower on large DBs).
To restore native vector support, ensure the platform package is installed:
# Windows npm install sqlite-vec-windows-x64 # macOS Apple Silicon npm install sqlite-vec-darwin-arm64 # Linux npm install sqlite-vec-linux-x64
Memory / Database Issues
Low memory count or Claude seems to have forgotten everything
Symptom: vektor_status shows far fewer memories than expected, or Claude has no recall of past sessions.
Cause: Each Claude surface (Desktop, web, VS Code) spawns its own MCP server process. If VEKTOR_DB_PATH is not set, each process creates a new DB in the default location, fragmenting your memories across multiple files.
Fix: Set VEKTOR_DB_PATH to a single absolute path in your MCP config env block:
"env": { "VEKTOR_DB_PATH": "/absolute/path/to/your/vektor-memory.db" }
Find your real DB:
find ~ -name "*.db" -path "*vektor*" 2>/dev/null
Merge fragment DBs into one:
npx vektor-slipstream merge-dbs --source ./fragment.db --target ./master.db
General
The package isn't installed globally. Download the .tgz from the Downloads page and run:
npm install -g ./vektor-slipstream-1.5.8.tgz
If you see Cannot find module 'vektor-memory' — that package name is deprecated. The correct package is vektor-slipstream.
node --version), and operating system. Response within 24 hours — 6 months support included with your purchase.