Advanced PGVector + Postgres nodes for n8n with full CRUD control
npm install n8n-nodes-pgvector-advancedAdvanced PGVector nodes for n8n with full CRUD control. No more limitations of the built-in node.
View Full Documentation | Installation | Quick Start | Docker Guide
---
``bash`
bash <(curl -fsSL https://raw.githubusercontent.com/aaron777collins/BetterPGVectorN8N/main/install.sh)
That's it! The installer auto-detects your setup and does the right thing.
What does the installer do?
The script detects your n8n environment and offers the best option:
| Your Setup | What It Does |
|------------|--------------|
| Docker Compose | Creates persistent setup that survives rebuilds |
| Running Container | Installs directly into the container |
| npm/Local Install | Adds to ~/.n8n/nodes |
| Nothing Found | Creates fresh n8n + pgvector stack |
Manual install options
`bashDownload installer first
curl -fsSL https://raw.githubusercontent.com/aaron777collins/BetterPGVectorN8N/main/install.sh -o install.sh
chmod +x install.sh
Or via n8n UI: Settings → Community Nodes → Install →
Updating to latest version
bash
Update existing installation
./install.sh --updateOr run the installer again - it auto-updates on container restart
docker compose restart n8n
`The Docker init script automatically checks for updates on each container start.
---
Why Use This?
| Built-in PGVector Node | This Package |
|------------------------|--------------|
| Insert only | Full CRUD (Upsert, Query, Delete, Get) |
| No stable IDs | External IDs for reliable syncing |
| Basic queries | Filters, pagination, multiple distance metrics |
| Single inserts | Batch operations (1000+ embeddings) |
| Manual schema | Auto table/index creation |
| No AI Agent tools | AI Agent Tool for RAG workflows |
---
Quick Start
$3
In n8n: Credentials → Add → Postgres
Host: your-postgres-host
Port: 5432
Database: your_db
User: your_user
Password: your_password
$3
Add a PGVector Advanced node with:
- Operation:
- Admin Operation:
- Dimensions: (or your embedding size)$3
`
Operation: Upsert
Collection: my_documents
External ID: doc-123
Content: "Your document text"
Embedding: [0.1, 0.2, 0.3, ...]
Metadata: {"category": "tech", "author": "Jane"}
$3
Operation: Query
Collection: my_documents
Query Embedding: [0.1, 0.2, ...]
Top K: 10
Distance Metric: cosine
---
Operations
$3
json
{
"operation": "upsert",
"collection": "documents",
"externalId": "doc-123",
"content": "Document text",
"metadata": {"category": "tech"},
"embedding": [0.1, 0.2, ...]
}
`Batch mode: Map fields from input items for bulk inserts.
$3
json
{
"operation": "query",
"collection": "documents",
"queryEmbedding": [0.1, 0.2, ...],
"topK": 10,
"distanceMetric": "cosine",
"metadataFilter": {"category": "tech"}
}
`$3
json
{
"operation": "delete",
"collection": "documents",
"deleteBy": "externalId",
"deleteExternalIds": "doc-1, doc-2"
}
`Delete by:
, externalId, or metadata filter.$3
`json
{
"operation": "get",
"collection": "documents",
"getBy": "externalId",
"getExternalIds": "doc-1, doc-2"
}
`$3
| Operation | What It Does |
|-----------|--------------|
|
| Creates table + indexes if missing |
| createIndex | Adds HNSW or IVFFlat vector index |
| dropCollection | Deletes all records in a collection |---
Distance Metrics
| Metric | Best For |
|--------|----------|
| Cosine | Text embeddings (OpenAI, Cohere, etc.) |
| L2 | When absolute distance matters |
| Inner Product | Pre-normalized vectors |
---
Example Workflows
Semantic Search:
1. Parse documents → Generate embeddings (OpenAI) → Upsert to PGVector → Query similar
Deduplication:
1. Query existing → If similarity > threshold, skip → Else upsert
Sync from External System:
1. Use
externalId to upsert → Automatically updates existing or inserts new---
Docker Persistence (How It Works)
The installer creates a custom Dockerfile that auto-installs community nodes on startup:
`
n8n/
├── Dockerfile # Extends official n8n image
└── init-nodes.sh # Installs packages on container start
To add more packages, edit
:`sh
PACKAGES="n8n-nodes-pgvector-advanced other-package"
`Then:
---
Troubleshooting
"pgvector extension not found"
sql
CREATE EXTENSION IF NOT EXISTS vector;
`Slow queries?
- Create an HNSW index via Admin → Create Index
- Use metadata filters to reduce search space
Dimension mismatch?
- All embeddings in a collection must have the same dimensions
- Use different collections for different embedding models
---
Development
bash
git clone https://github.com/aaron777collins/BetterPGVectorN8N.git
cd BetterPGVectorN8N
npm install
npm run build
npm test
`---
AI Agent Tool
The PGVector Store Tool lets n8n AI Agents manage knowledge with natural operations:
| Operation | What it does |
|-----------|--------------|
| Remember | Store new info, update by ID or similar concept |
| Recall | Search with similarity threshold |
| Forget | Delete by exact ID |
| Forget Similar | Delete by concept (with dry-run safety) |
| Lookup | Get by exact ID |
$3
Each operation has n8n settings (safety/behavior) and AI parameters (data):
| n8n Config | AI Provides |
|------------|-------------|
| Collection, thresholds, ID hints | Content, IDs, queries |
| Dry run mode, result limits | Metadata filters |
Example - Remember operation:
- n8n:
, Auto-Generate ID = ON
- AI: , id (optional), metadataExample - Forget Similar operation:
- n8n:
, Dry Run = ON
- AI: (what to delete)---
MCP Server for External AI Agents
Use with Claude, GPT, or other AI agents outside n8n.
`bash
Run directly
npx n8n-nodes-pgvector-advancedOr install globally
npm install -g n8n-nodes-pgvector-advanced
pgvector-mcp
`Available tools:
, pgvector_query, pgvector_delete, pgvector_get, pgvector_adminSet environment variables:
, PGPORT, PGDATABASE, PGUSER, PGPASSWORD`---
| Guide | Description |
|-------|-------------|
| Installation | All installation methods |
| Quick Start | Get running in 5 minutes |
| Operations | Full operations reference |
| AI Agent Tools | Use with n8n AI Agents |
| MCP Server | Use with external AI (Claude) |
| Docker Guide | Persistent Docker setup |
| Troubleshooting | Common issues & fixes |
---
MIT
---
Made with ❤️ for the n8n community