pgvector

Pro

DevDb finds vector columns (pgvector) in PostgreSQL and Neon databases. It shows embeddings in a readable form and runs similarity searches.

Requirements

• A PostgreSQL or Neon database with the vector extension.
• A table with a vector column.
• A DevDb Pro license.

Find similar rows

1. Open a table with a vector column.
2. In a vector cell, click the Find similar rows action.
3. The similarity search panel opens. Set the options, then run the search.

Source

Select where the reference vector comes from:

Source	Description
This row	The vector of the cell that you selected
Paste vector	A vector that you paste, for example [0.12, 0.44, -0.03]
Query text	Text that DevDb sends to an embedding endpoint

Options

Option	Description
Metric	Cosine (<=>), L2 (<->), Inner product (<#>) or L1 (<+>)
Top-k	Number of rows to return
Metadata filter	Optional SQL condition, for example tenant_id = 42 AND lang = 'en'
ef_search (Advanced)	HNSW search size. A larger value gives better recall and slower queries.

pgvector returns the negative inner product for <#>. DevDb changes the sign so that a larger value means more similar.

The panel shows the SQL that DevDb ran. Click Copy SQL to use it in your code.

Search by text

To search with text, add an embedding endpoint. DevDb sends your text to the endpoint, gets a vector, and searches with it.

WARNING

Use the same embedding model that created the vectors in your table. Vectors from different models do not compare.

Embedding endpoints

1. In the similarity search panel, select Query text.
2. Click Add an embedding endpoint to search by text, or click Manage endpoints.
3. Enter the details:

Field	Description
Label	A name for the endpoint
Provider	OpenAI-compatible, Ollama or Custom
URL	The endpoint URL
Model	The embedding model, for example text-embedding-3-small or nomic-embed-text
API key	Optional. Needed by most hosted services.

Supported endpoints:

Provider	URL example
OpenAI-compatible /v1/embeddings (OpenAI, Azure OpenAI, LM Studio, Together and similar)	https://api.openai.com/v1/embeddings
Ollama /api/embeddings	http://localhost:11434/api/embeddings

Endpoint security

• The URL must use https. DevDb accepts http only for localhost, 127.0.0.1 and ::1.
• DevDb keeps API keys in the secret storage of your editor. It does not send them to the DevDb panel, and it does not show them again after you save.
• DevDb sends an API key only to the origin (scheme, host and port) that you saved it for.