Bring intelligence to your data, not your data to the cloud. Run high-performance LLM inference entirely within your VPC with a single SQL extension.
-- pgInfer: Summarize 1k rows in one query
SELECT ai_live.summarize(content)
FROM production_logs
WHERE created_at > now() - interval '1 hour';Choose the mode that fits your security, speed, and cost requirements
The Speed Pillar
Immediate access to frontier models like GPT-4 and Claude. Perfect for rapid prototyping and testing logic before moving to production.
The Security Pillar
Total data residency. Run inference on local models (via Ollama/vLLM) inside your firewall. Zero data transit. Built for regulated industries.
The Efficiency Pillar
Optimize for scale. Use async Batch APIs to process millions of rows at 50% lower cost than real-time inference. Perfect for non-urgent processing.
Manage AI logic with the same tools you use for your database
No Python glue code, no ETL pipelines, and no new microservices to monitor. It's just Postgres.
Manage your AI logic using standard SQL migrations and version control.
Leverage existing Postgres RBAC. Permissions are handled at the database level—if they can't see the row, they can't run the AI.
Built-in audit tables track all prompt/response interactions for compliance without tracking individual user identities.
See how pgInfer eliminates complexity
| Feature | pgInfer | Traditional Stack |
|---|---|---|
| Data Residency | ||
| Zero External APIs | ||
| Native SQL Integration | ||
| No ETL Required | ||
| Built-in Audit Trails | ||
| VPC/On-Prem Support |
Enterprise-grade security and compliance
The extension code is open-audit, ensuring no backdoors. Your data flows through transparent, verifiable paths.
Automatic redaction of sensitive patterns before data hits the model. Protect customer information at the SQL level.
Every inference is logged to a system table for compliance officers. Track usage, costs, and intent using SQL.
Control which database users can trigger AI functions. Fine-grained permissions for enterprise security.
We spent years building fragile AI pipelines that broke every time a schema changed. We got tired of "glue code" and the security risks of moving production data to the cloud. V1 turns PostgreSQL into a native AI engine.
Choose the plan that fits your security and scale requirements
20 records/day limit. Live Proxy mode. BYO-Key.
Unlimited inference. Live Proxy & Batch modes. BYO-Key.
Unlimited. Sovereign AI Mode (Local/VPC). Air-gapped license. Priority SLA.
The Architecture of Data Residency: Traditional AI implementation requires a "Data-to-Model" approach, forcing organizations to transmit sensitive PII and proprietary data across public networks to third-party providers. pgInfer fundamentally inverts this model by bringing Sovereign Intelligence directly into the database.
Zero-Infrastructure Deployment: pgInfer eliminates the "Integration Tax" usually associated with AI projects. There are no new microservices to manage, no complex ETL pipelines to maintain, and no external API dependencies that create fragility or latency.
Auditability and Regulatory Compliance: Designed for highly regulated sectors such as finance and healthcare, pgInfer provides a transparent, privacy-first audit trail. Every inference operation is recorded into native SQL audit tables, providing an immutable record for compliance officers.