databend-aiserver

Databend AI Server extends any data warehouse with AI-ready UDFs, seamlessly fusing object storage, embeddings, and SQL pipelines.

Quickstart

uv sync
uv run databend-aiserver --port 8815

AI Functions

Function	Signature	Purpose	Output
ai_list_files	(stage_location, max_files)	List objects in a stage for inspection/sampling.	Table with file details (path, size, etc.)
ai_embed_1024	(text)	Generate 1024-dim embeddings (default: Qwen).	VECTOR(1024)
ai_parse_document	(stage_location, path)	Parse docs (PDF, DOCX, Images, etc.) to Markdown.	VARIANT (pages, metadata)

Usage

1. Register Functions in Databend

CREATE OR REPLACE FUNCTION ai_list_files(stage_location STAGE_LOCATION, pattern VARCHAR, max_files INT)
RETURNS TABLE (stage_name VARCHAR, path VARCHAR, uri VARCHAR, size UINT64, last_modified VARCHAR, etag VARCHAR, content_type VARCHAR)
LANGUAGE PYTHON HANDLER = 'ai_list_files' ADDRESS = '<your-ai-server-address>';

CREATE OR REPLACE FUNCTION ai_embed_1024(text VARCHAR)
RETURNS VECTOR(1024)
LANGUAGE PYTHON HANDLER = 'ai_embed_1024' ADDRESS = '<your-ai-server-address>';

CREATE OR REPLACE FUNCTION ai_parse_document(stage_location STAGE_LOCATION, file_path VARCHAR)
RETURNS VARIANT
LANGUAGE PYTHON HANDLER = 'ai_parse_document' ADDRESS = '<your-ai-server-address>';

2. Run Queries

-- Setup Stage
CREATE CONNECTION my_s3_conn STORAGE_TYPE = 's3' ACCESS_KEY_ID = '...' SECRET_ACCESS_KEY = '...';
CREATE STAGE docs_stage URL='s3://load/files/' CONNECTION = (CONNECTION_NAME = 'my_s3_conn');

-- Execute AI Functions
SELECT * FROM ai_list_files(@docs_stage, 50);
SELECT ai_embed_1024(doc_body) FROM docs_tbl;
SELECT ai_parse_document(@docs_stage, 'reports/q1.pdf');

Development

# Run full test suite
uv run pytest

---

Built by the Databend team.