FAISS RAG Backend
by Kitworks Systems
850,00 €
| Odoo Apps Dependencies | base, base_automation, bus, http_routing, mail |
| Community Apps Dependencies | Show |
| Technical Name | kw_ai_rag_faiss |
| License | OPL-1 |
| Website | https://kitworks.systems/ |
| Versions | 17.0 18.0 19.0 |
AI Connector — RAG Backend
FAISS — In-Memory Vector Backend
Meta's battle-tested library for billion-scale approximate nearest-neighbour search. Runs entirely in Odoo server memory — no sidecar process needed.
IVF
Index Type
in-proc
Deployment
GPU
Optional Accel.
When to Use FAISS
Maximum speed
In-process ANN search delivers the lowest latency of any backend. Ideal for real-time assistants where every millisecond counts.
On-premises air-gapped deployments
Runs without any external network calls. Suitable for secure government, banking, or defence environments.
GPU-accelerated search
FAISS GPU indexes enable billion-scale searches on CUDA hardware. Uncommon for Odoo but available when needed.
Technical Specifications
| Property | Value | Notes |
|---|---|---|
| FAISS version | 1.7+ | Installed via pip into Odoo virtualenv |
| Index type | IVFFlat | Configurable: Flat, HNSW32 |
| Similarity metric | L2 / Inner product | Cosine via normalization |
| Persistence | File (index.faiss) | Written to Odoo filestore |
| External service | None | Runs inside Odoo process |
Requirements
- ✓ kw_ai — AI Connector core
- ✓ kw_ai_rag — RAG orchestration module
- ✓ faiss-cpu (or faiss-gpu) Python package installed in Odoo virtualenv
- ✓ Any embedding module
Fastest in-process vector search for Odoo
No extra services. Runs inside Odoo — air-gap friendly.
Book a Demo