Activeloop

Tensor Query Language (TQL)

A high-performance, SQL-like query engine (implemented in C++ inside Deep Lake) for filtering datasets and running hybrid embedding-plus-attribute search. TQL queries are executed through the SDK via `ds.query()` or `vector_store.search(query=)`, and require an authenticated Activeloop account (not available on purely local datasets). This is a query language surfaced through the SDK, not a standalone REST API.

GitHub OpenAPI

Documentation

📖
Documentation
https://docs-v3.activeloop.ai/examples/tql

Specifications

⚙
OpenAPI
https://raw.githubusercontent.com/api-evangelist/activeloop/refs/heads/main/openapi/activeloop-openapi.yml

Other Resources

🔗
GitHub
https://github.com/activeloopai/deeplake

OpenAPI Specification

activeloop-openapi.yml Raw ↑ 
openapi: 3.0.1
info:
  title: Activeloop Deep Lake API
  description: >-
    Deep Lake is a database for AI built by Activeloop. Its PRIMARY interface is
    the open-source `deeplake` Python SDK paired with the Tensor Query Language
    (TQL) - datasets and embeddings are created, versioned, queried, and streamed
    through Python, not through a broad REST API. The only documented HTTP/REST
    surface is the alpha Managed Database query endpoint at
    https://app.activeloop.ai/api/query/v1, which accepts a TQL query string and
    returns results for datasets hosted in the Activeloop Managed Tensor Database.
    That single endpoint is modeled below; no other REST endpoints are fabricated.
    For dataset creation, ingestion, version control, vector store operations, and
    most queries, use the Deep Lake Python SDK (https://docs.deeplake.ai/) and TQL.
  termsOfService: https://www.activeloop.ai/terms-of-service/
  contact:
    name: Activeloop Support
    url: https://www.activeloop.ai/
  license:
    name: Apache-2.0
    url: https://www.apache.org/licenses/LICENSE-2.0
  version: 'v1-alpha'
servers:
  - url: https://app.activeloop.ai
    description: Activeloop Cloud (Managed Tensor Database)
paths:
  /api/query/v1:
    post:
      operationId: runManagedQuery
      tags:
        - Managed Database
      summary: Run a TQL query against the Managed Tensor Database (Alpha).
      description: >-
        Executes a Tensor Query Language (TQL) query against a dataset hosted in
        the Activeloop Managed Tensor Database (datasets created with
        `runtime={"tensor_db": True}`). This endpoint is documented as Alpha and
        its syntax may change without announcement. Most Deep Lake usage runs
        through the Python SDK rather than this endpoint.
      security:
        - bearerAuth: []
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/QueryRequest'
      responses:
        '200':
          description: Query executed successfully; results returned as JSON.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/QueryResponse'
        '401':
          description: Missing or invalid ACTIVELOOP_TOKEN.
        '400':
          description: Malformed TQL query or request body.
components:
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: >-
        Activeloop API token passed as `Authorization: Bearer {ACTIVELOOP_TOKEN}`.
  schemas:
    QueryRequest:
      type: object
      required:
        - query
      properties:
        query:
          type: string
          description: >-
            A Tensor Query Language (TQL) statement, e.g.
            `select * from "hub://org_id/dataset_name" where ...`.
          example: select * from "hub://activeloop/dataset" limit 10
    QueryResponse:
      type: object
      description: >-
        JSON results of the query. The exact response shape is not formally
        documented by Activeloop (Alpha); results are returned as query rows.
      additionalProperties: true