Enofirm-Temp/extra-addons/AI/llm_pgvector

LLM pgvector Integration

PostgreSQL-native vector storage using pgvector extension.

Module Type: 🗄️ Vector Store (PostgreSQL Native)

Architecture

┌───────────────────────────────────────────────────────────────┐
│                    Used By (RAG Modules)                      │
│        ┌───────────────┐           ┌───────────────┐         │
│        │ llm_knowledge │           │llm_assistant  │         │
│        │   (RAG)       │           │  (with RAG)   │         │
│        └───────┬───────┘           └───────┬───────┘         │
└────────────────┼───────────────────────────┼─────────────────┘
                 └─────────────┬─────────────┘
                               ▼
              ┌───────────────────────────────────────────┐
              │             llm_store                     │
              │        (Vector Store API)                 │
              └─────────────────────┬─────────────────────┘
                                    │
                                    ▼
              ┌───────────────────────────────────────────┐
              │      ★ llm_pgvector (This Module) ★       │
              │         pgvector Implementation           │
              │  🐘 PostgreSQL │ Native │ No Extra Server │
              └─────────────────────┬─────────────────────┘
                                    │
                                    ▼
              ┌───────────────────────────────────────────┐
              │             PostgreSQL + pgvector         │
              │           (Your Odoo Database)            │
              └───────────────────────────────────────────┘

Installation

What to Install

For RAG with PostgreSQL vectors:

# 1. Install pgvector extension on PostgreSQL
# See: https://github.com/pgvector/pgvector

# 2. Install the Odoo module
odoo-bin -d your_db -i llm_knowledge,llm_pgvector

Auto-Installed Dependencies

• llm (core infrastructure)
• llm_store (vector store abstraction)

Why Choose pgvector?

Feature	pgvector
Integration	🐘 Uses your Odoo PostgreSQL
Extra Server	❌ Not needed
Simplicity	✅ No external dependencies
Scale	📊 Good for moderate datasets

Vector Store Comparison

Feature	llm_pgvector	llm_qdrant	llm_chroma
Server	🐘 PostgreSQL	🔷 Qdrant server	🌈 Chroma server
Setup	Easy	Moderate	Moderate
Scale	Medium	High	Medium
Best For	Simple RAG	Large scale	Development

Common Setups

I want to...	Install
Simple RAG	llm_knowledge + llm_pgvector
Chat + RAG	llm_assistant + llm_openai + llm_knowledge + llm_pgvector

Features

• Native PostgreSQL vector storage
• Cosine similarity search
• Collection-specific indices
• Metadata filtering
• Uses existing Odoo database connection

Configuration

1. Ensure pgvector extension is installed on your PostgreSQL server
2. Install the module
3. Configure vector store in LLM > Configuration > Vector Stores
4. Set up knowledge base with pgvector as the storage backend

Creating Indices for Better Performance

Models inheriting from EmbeddingMixin can organize their embeddings into collections and create collection-specific indices:

class DocumentChunk(models.Model):
    _name = 'document.chunk'
    _inherit = ['llm.embedding.mixin']

    def ensure_collection_index(self, collection_id=None):
        """Ensure a vector index exists for the specified collection."""
        embedding_model = self.env['llm.model'].search([
            ('model_use', '=', 'embedding'),
        ], limit=1)

        sample_embedding = embedding_model.generate_embedding("")
        dimensions = len(sample_embedding)

        self.create_embedding_index(
            collection_id=collection_id,
            dimensions=dimensions,
            force=False
        )

Requirements

• Odoo 18.0+
• PostgreSQL with pgvector extension
• Python packages: pgvector, numpy

License

LGPL-3