Tools & AI

LLM Settings

Configure your LLM provider, model and behavior.

Connection
Endpoint and authentication used to contact the LLM provider.
Example: http://host.docker.internal:11434/v1/ when running Ollama locally alongside the pgAssistant Docker image. Make sure to set OLLAMA_HOST=0.0.0.0:11434.
Use only for a trusted internal LLM endpoint with a self-signed certificate. Disabling verification makes HTTPS connections vulnerable to interception.
Model
Select the model returned by the configured endpoint.
SQL behavior
Guidelines used when pgAssistant asks the LLM to review SQL or table definitions.
These guidelines are available as {{GUIDELINES}} in the naming conventions prompt.
RFC prompt
Template used to analyze table structures against RFCs and standards.
Must contain {{DDL}}. Leave empty or reset to use the original pgAssistant template.
Naming prompt
Template used to check SQL naming conventions on table definitions.
Must contain {{DDL}}. Use {{GUIDELINES}} if the prompt should include the SQL conventions URL.