AI Data Platform
FoundryDB extends your managed databases with a set of AI-focused data services. They sit inside the same platform you already use: no extra infrastructure, no additional accounts, and no data leaving European zones unless your chosen model provider routes outside the EU.
What's Included
| Capability | What it does |
|---|---|
| Vector search | Typed, read-only pgvector similarity search brokered through the controller. No raw database credentials needed for query-time access. |
| Embedding pipelines | Managed jobs that watch a source table and write embeddings into a target table automatically. Supports continuous, scheduled, and manual modes. |
| Inference proxy | An OpenAI-compatible /inference/v1 endpoint backed by your own provider API keys. Supports OpenAI, Anthropic, Mistral, Azure OpenAI, and Groq. Keys are stored encrypted at rest and never returned by the API. |
| AI Query (NL2SQL) | Translate a natural language question into read-only SQL and optionally execute it. Works with PostgreSQL and MySQL services. |
| Data Explorer | Brokered read-only query and schema introspection for your managed databases. Suitable for dashboards and internal tooling without exposing database credentials. |
How the Pieces Fit Together
The common thread across all five capabilities is the read-only data plane: every query path runs inside a READ ONLY transaction with a statement timeout and a row cap. Nothing in this section can mutate your data.
A typical AI workflow looks like this:
- Configure an embedding pipeline to vectorize new rows as they arrive.
- Use the inference proxy to call your embedding model without hard-coding provider keys in application code.
- Query the resulting vector column using the vector search endpoint to find semantically similar rows.
- Use AI Query to let non-technical users ask questions in plain language.
- Use the Data Explorer to inspect raw results or troubleshoot data quality.
Prerequisites
- A running PostgreSQL service (vector search, embedding pipelines, NL2SQL, Data Explorer).
- A running MySQL service for NL2SQL and Data Explorer on MySQL.
- An organization owner or admin role to configure inference provider keys.
- Embedding pipelines and inference proxy require provider API keys (OpenAI, Cohere, Anthropic, Mistral, etc.).
Sections
- Vector Search — run pgvector similarity queries over the API.
- Inference Proxy — bring your own provider keys, route through FoundryDB.
- Embedding Pipelines — automate row vectorization.
- AI Query and Data Explorer — natural language queries and brokered read-only access.