Teradata

Run queries on Teradata

Overview

The Teradata tools allow you to run SQL queries on a Teradata database and to build and search vector stores on Teradata Vantage Cloud for RAG-style workflows.

Key Features

• TERADATA_RUN_QUERY
  • Execute SQL queries on a Teradata database.
  • Retrieve results in a structured format.
• VANTAGE_CLOUD_VECTOR_STORE_CREATE
  • Create a VectorStore on Vantage Cloud from files in Jinba storage, with document ingestion and chunking.
• VANTAGE_CLOUD_VECTOR_STORE_ADD_DOCUMENT
  • Add documents from files in Jinba storage to an existing VectorStore on Vantage Cloud.
• VANTAGE_CLOUD_VECTOR_SEARCH
  • Perform semantic similarity search on a Vantage Cloud VectorStore.

Authentication

For TERADATA_RUN_QUERY, you will need your Teradata database credentials: host, user, and password.

The three VANTAGE_CLOUD_* tools share the same config parameters:

Parameter	Required	Description
host	No	Vantage Cloud host address (default: host.docker.internal, keep the default when using an SSH tunnel on the host machine)
dbs_port	No	Database port number (default: 1025)
username	Yes	Database username
password	Yes	Database password
base_url	Yes	UES URI (the URL without the trailing /open-analytics)
pat_token	Yes	Personal Access Token (for CCP authentication)
pem_content	Yes	Contents of the PEM private key (paste the text starting with -----BEGIN PRIVATE KEY----- as-is)
pem_kid	No	Key ID of the PEM key (defaults to username; must match the key name registered on the Teradata side)

Note: Treat API keys as sensitive information and never commit them to public repositories.

Vantage Cloud Vector Tools

VANTAGE_CLOUD_VECTOR_STORE_CREATE

Create a VectorStore on Vantage Cloud from files in Jinba storage.

Input:

Parameter	Type	Required	Description
vector_store_name	string	Yes	Name of the VectorStore to create
file_urls	string	Yes	Comma-separated (or newline-separated) HTTPS URLs of files to ingest as documents (maximum 10 files)
embeddings_model	string	No	Embeddings model to use for vectorization (default: amazon.titan-embed-text-v1)
search_algorithm	string	No	Search algorithm: VECTORDISTANCE or HNSW (default: VECTORDISTANCE)
top_k	number	No	Number of top results for search (default: 5)
object_names	string	No	Object name for data storage (defaults to vector_store_name + _data)
data_columns	string	No	Data column name for text content (default: chunks)
vector_column	string	No	Column name for vector embeddings (default: Embedding)
chunk_size	number	No	Chunk size for document splitting (default: 500)

Output: result with name (created VectorStore), status, and message.

VANTAGE_CLOUD_VECTOR_STORE_ADD_DOCUMENT

Add documents to an existing VectorStore on Vantage Cloud.

Input:

Parameter	Type	Required	Description
vector_store_name	string	Yes	Name of the existing VectorStore to add documents to
file_urls	string	Yes	Comma-separated (or newline-separated) HTTPS URLs of files to add as documents (maximum 10 files)

Output: result with name, status, and message.

VANTAGE_CLOUD_VECTOR_SEARCH

Perform semantic similarity search on a Vantage Cloud VectorStore.

Input:

Parameter	Type	Required	Description
vector_store_name	string	Yes	Name of the VectorStore to search
query	string	Yes	Natural language query for similarity search
top_k	number	No	Number of similar results to return (default: 5)

Output: result with similar_objects (each containing score and, when available, DataBaseName, TableName, TD_ID, TD_FILENAME, chunks, index_label) and total_count.

Example: Create a VectorStore and Search It

- id: create_vector_store
  tool: VANTAGE_CLOUD_VECTOR_STORE_CREATE
  config:
    - name: username
      value: "{{secrets.VANTAGE_USERNAME}}"
    - name: password
      value: "{{secrets.VANTAGE_PASSWORD}}"
    - name: base_url
      value: "{{secrets.VANTAGE_BASE_URL}}"
    - name: pat_token
      value: "{{secrets.VANTAGE_PAT_TOKEN}}"
    - name: pem_content
      value: "{{secrets.VANTAGE_PEM_CONTENT}}"
  input:
    - name: vector_store_name
      value: "product_docs"
    - name: file_urls
      value: "{{steps.upload_manual.result.file_url}}"

- id: vector_search
  tool: VANTAGE_CLOUD_VECTOR_SEARCH
  config:
    - name: username
      value: "{{secrets.VANTAGE_USERNAME}}"
    - name: password
      value: "{{secrets.VANTAGE_PASSWORD}}"
    - name: base_url
      value: "{{secrets.VANTAGE_BASE_URL}}"
    - name: pat_token
      value: "{{secrets.VANTAGE_PAT_TOKEN}}"
    - name: pem_content
      value: "{{secrets.VANTAGE_PEM_CONTENT}}"
  input:
    - name: vector_store_name
      value: "product_docs"
    - name: query
      value: "How do I reset my password?"
    - name: top_k
      value: 5

Example: Run Teradata Query

- id: run_query
  tool: TERADATA_RUN_QUERY
  config:
    - name: host
      value: 100.21.165.28
    - name: user
      value: carnot
    - name: password
      value: carnot#123
  input:
    - name: query
      value: "

        \          with examples as (

        \              select 1 as id, 'jon' as name

        \              union all

        \              select 2 as id, 'jane' as name

        \          )

        \          select * from examples;

        \          "