ThoughtSpot Python SDK

Python client for ThoughtSpot's v2 REST APIs.

The SDK is async-first (built on httpx) and ships a synchronous variant of every method. It is compatible with Python 3.9 and above.

Installation

pip install thoughtspot-rest-api-sdk

Usage

Create a Configuration, pass it to an ApiClient, and make calls through ThoughtSpotRestApi (which exposes every endpoint) or a focused per-tag class such as UsersApi or MetadataApi.

import asyncio
from thoughtspot_rest_api_sdk import Configuration, ThoughtSpotRestApi

BASE_URL = "https://your-cluster.thoughtspot.cloud"

async def main():
    config = Configuration(host=BASE_URL, access_token="YOUR_BEARER_TOKEN")

    # Pass the Configuration directly; the client manages its own connection
    # pool. (To share one pool across several API classes, build an ApiClient
    # explicitly and pass that instead — see "Per-tag API classes".)
    async with ThoughtSpotRestApi(configuration=config) as client:

        # Current user
        user = await client.get_current_user_info()
        print(user.name)

        # Search with a request body (dict or the typed request model)
        users = await client.search_users({"record_offset": 0, "record_size": 10})
        for u in users:
            print(u.name)

asyncio.run(main())

Synchronous usage

Every method has a blocking *_sync variant. No event loop or await is required, which makes the SDK usable from synchronous code (Django, Flask, scripts, notebooks):

from thoughtspot_rest_api_sdk import Configuration, ThoughtSpotRestApi

config = Configuration(host=BASE_URL, access_token="YOUR_BEARER_TOKEN")
client = ThoughtSpotRestApi(configuration=config)

user = client.get_current_user_info_sync()
print(user.name)

Per-tag API classes

ThoughtSpotRestApi aggregates all endpoints. For a focused surface, instantiate a per-tag class against the same ApiClient — it shares the same connection pool and authentication:

from thoughtspot_rest_api_sdk import ApiClient, Configuration, UsersApi, MetadataApi

async with ApiClient(config) as api_client:
    users = UsersApi(api_client)
    metadata = MetadataApi(api_client)
    await users.search_users({"record_offset": 0, "record_size": 10})

Accessing response status and headers

Each method has a *_with_http_info variant (and a *_sync_with_http_info variant) that returns the status code, headers, and deserialized data:

response = await client.get_current_user_info_with_http_info()
print(response.status_code)
print(response.headers)
print(response.data)

Authentication

Set access_token on the Configuration. It accepts a static token string, the built-in token provider, or your own callable — three options for the three common cases.

1. Static token — when you already hold a bearer token:

config = Configuration(host=BASE_URL, access_token="YOUR_BEARER_TOKEN")

2. Built-in provider — ThoughtSpotTokenProvider mints a token from /auth/token/full and refreshes it transparently. Supply exactly one of password (basic auth) or secret_key (trusted authentication). It caches the token and only re-mints when the cached one is near expiry, and collapses concurrent refreshes into a single call:

from thoughtspot_rest_api_sdk import Configuration, ThoughtSpotTokenProvider

provider = ThoughtSpotTokenProvider(BASE_URL, "USERNAME", password="PASSWORD")
config = Configuration(host=BASE_URL, access_token=provider)

3. Your own callable — for any other identity source. It is invoked on every request (so it doubles as the refresh hook) and may be sync or async:

async def token_supplier() -> str:
    return await my_identity_provider.fetch_bearer()

config = Configuration(host=BASE_URL, access_token=token_supplier)

Streaming

Endpoints that return a server-sent event (SSE) stream expose an additional *_stream async generator that yields events as they arrive:

from thoughtspot_rest_api_sdk.models import SendAgentConversationMessageStreamingRequest

async for event in client.send_agent_conversation_message_streaming_stream(
    conversation_identifier="CONVERSATION_ID",
    send_agent_conversation_message_streaming_request=(
        SendAgentConversationMessageStreamingRequest(messages=["Hello"])
    ),
):
    if event.get("type") == "text-chunk":
        print(event.get("content", ""), end="", flush=True)

File uploads

Multipart endpoints (for example DBT and Style Customization) accept file content as raw bytes, a (filename, bytes) tuple, or an open file handle:

with open("project.zip", "rb") as f:
    await client.dbt_connection(
        connection_name="my-connection",
        database_name="MY_DB",
        import_type="ZIP_FILE",
        file_content=f,
    )

File downloads

Export endpoints return binary content as bytes. Write it to disk:

from pathlib import Path

data = await client.export_liveboard_report(metadata_identifier="LIVEBOARD_ID", file_format="PDF")
Path("report.pdf").write_bytes(data)

Error handling

API errors raise typed exceptions. Catch them by HTTP status, or catch the base ApiException:

from thoughtspot_rest_api_sdk.exceptions import (
    ApiException,                  # base class
    BadRequestException,           # 400
    UnauthorizedException,         # 401
    ForbiddenException,            # 403
    NotFoundException,             # 404
    ConflictException,             # 409
    UnprocessableEntityException,  # 422
    ServiceException,              # 5xx
)

try:
    await client.search_metadata({"metadata": [{"type": "LIVEBOARD"}]})
except UnauthorizedException:
    ...  # refresh credentials and retry
except ApiException as e:
    print(e.status, e.reason, e.body, e.headers)

Every exception exposes .status, .reason, .body, .data, and .headers.

Retries

Retries are off by default. Set retries on the Configuration to enable automatic retries with exponential backoff and jitter:

config = Configuration(host=BASE_URL, access_token="...", retries=3)

When enabled, the SDK retries on retry_statuses (default 429, 502, 503, 504) and on network/timeout errors, with exponential backoff and jitter. Once the retry budget is exhausted the final response is returned, raising the usual typed exception (a 429 raises the base ApiException — branch on e.status == 429).

POST requests are retried too (ThoughtSpot uses POST for read endpoints). To avoid retrying non-idempotent calls, restrict the eligible methods:

config.retry_methods = {"GET", "PUT", "DELETE"}

Configuration

The default request timeout is 300s when nothing is configured. The timeout and header options are constructor arguments:

config = Configuration(
    host=BASE_URL,
    access_token="...",
    connect_timeout=5,
    read_timeout=60,
    default_headers={"X-My-App": "demo"},
)

A single call can still override these: await client.search_users({...}, _request_timeout=30, _headers={"X-Trace": "1"}). _request_timeout accepts a float (all phases) or a (connect, read) tuple of floats.

Note: For clusters with self-signed certificates, disable verification:

config = Configuration(host=BASE_URL, access_token="YOUR_BEARER_TOKEN", verify_ssl=False)

Updating configuration at runtime

To apply a new Configuration to an existing client — for example after rotating credentials or changing timeouts — call apply_configuration:

client.apply_configuration(Configuration(host=BASE_URL, access_token=NEW_TOKEN))

This rebuilds the underlying API client. For continuous token refresh you do not need this — set access_token to a callable (see Authentication); it is invoked on every request.

Live Playground

We recommend that you browse through the API playground before you start constructing your API requests. The playground offers an interactive portal with comprehensive information about the API endpoints and their request and response workflows.

Documentation for API endpoints

The full list of available methods is on the ThoughtSpotRestApi class in thoughtspot_rest_api_sdk/api/thought_spot_rest_api.py. Method names match the operation IDs in the REST API, in snake_case.

Author

This SDK is developed and maintained by ThoughtSpot.

• Website: www.thoughtspot.com
• GitHub: thoughtspot
• Email: support@thoughtspot.com

License

For full details, please refer to the license documentation.