truefoundry-gateway-sdk 0.0.2

No project description provided

Truefoundry Python Library

This library provides convenient access to the TrueFoundry Gateway agent API. The gateway is a stateful, multi-tenant runtime that streams agent responses over Server-Sent Events.

Call the gateway directly — not via the control-plane /api/llm proxy. The tenant name is part of the base URL (e.g. https://gateway.truefoundry.ai/<tenant>) and is applied to every request.

[!tip] You can ask questions about this SDK using DeepWiki

• Python:
• TypeScript:

Table of Contents

• Install
• Quickstart
• Releasing
• Installation
• Reference
• Usage
• Environments
• Async Client
• Exception Handling
• Streaming
• Advanced
  • Access Raw Response Data
  • Retries
  • Timeouts
  • Custom Client
• Contributing

Install

pip install truefoundry-gateway-sdk

Quickstart

from truefoundry_gateway_sdk import (
    AgentInputUserMessage,
    NamedAgentRunInput,
    TruefoundryGateway,
)

client = TruefoundryGateway(
    base_url="https://gateway.truefoundry.ai/<tenant>",
    token="<TFY_API_KEY>",
)

for event in client.agent.responses.create(
    request=NamedAgentRunInput(
        agent_name="my-agent",
        input=[AgentInputUserMessage(role="user", content="hi")],
    ),
):
    print(event)

AsyncTruefoundryGateway provides the same surface with async/await.

Releasing

Releases are cut from truefoundry/truefoundry-gateway-fern-config by pushing a v* tag there. The Fern workflow regenerates this repo onto a release-v<version> branch and publishes the wheel to PyPI in the same job.

Installation

pip install truefoundry-gateway-sdk

Reference

A full reference for this library is available here.

Usage

Instantiate and use the client with the following:

from truefoundry_gateway_sdk import TruefoundryGateway, AgentResponsesInlineAgent

client = TruefoundryGateway(
    token="<token>",
)

client.agents.responses.create(
    request=AgentResponsesInlineAgent(
        model="model",
    ),
)

Environments

This SDK allows you to configure different environments for API requests.

from truefoundry_gateway_sdk import TruefoundryGateway
from truefoundry_gateway_sdk.environment import TruefoundryGatewayEnvironment

client = TruefoundryGateway(
    environment=TruefoundryGatewayEnvironment.DEFAULT,
)

Async Client

The SDK also exports an async client so that you can make non-blocking calls to our API. Note that if you are constructing an Async httpx client class to pass into this client, use httpx.AsyncClient() instead of httpx.Client() (e.g. for the httpx_client parameter of this client).

import asyncio

from truefoundry_gateway_sdk import AsyncTruefoundryGateway

client = AsyncTruefoundryGateway(
    token="<token>",
)


async def main() -> None:
    await client.agents.responses.create(
        request=AgentResponsesInlineAgent(
            model="model",
        ),
    )


asyncio.run(main())

Exception Handling

When the API returns a non-success status code (4xx or 5xx response), a subclass of the following error will be thrown.

from truefoundry_gateway_sdk.core.api_error import ApiError

try:
    client.agents.responses.create(...)
except ApiError as e:
    print(e.status_code)
    print(e.body)

Streaming

The SDK supports streaming responses, as well, the response will be a generator that you can loop over.

from truefoundry_gateway_sdk import TruefoundryGateway, AgentResponsesInlineAgent

client = TruefoundryGateway(
    token="<token>",
)

client.agents.responses.create(
    request=AgentResponsesInlineAgent(
        model="model",
    ),
)

Advanced

Access Raw Response Data

The SDK provides access to raw response data, including headers, through the .with_raw_response property. The .with_raw_response property returns a "raw" client that can be used to access the .headers and .data attributes.

from truefoundry_gateway_sdk import TruefoundryGateway

client = TruefoundryGateway(...)
response = client.agents.responses.with_raw_response.create(...)
print(response.headers)  # access the response headers
print(response.status_code)  # access the response status code
print(response.data)  # access the underlying object

Retries

The SDK is instrumented with automatic retries with exponential backoff. A request will be retried as long as the request is deemed retryable and the number of retry attempts has not grown larger than the configured retry limit (default: 2).

Which status codes are retried depends on the retryStatusCodes generator configuration:

legacy (current default): retries on

• 408 (Timeout)
• 409 (Conflict)
• 429 (Too Many Requests)
• 5XX (All server errors, including 500)

recommended: retries on

• 408 (Timeout)
• 409 (Conflict)
• 429 (Too Many Requests)
• 502 (Bad Gateway)
• 503 (Service Unavailable)
• 504 (Gateway Timeout)

Use the max_retries request option to configure this behavior.

client.agents.responses.create(..., request_options={
    "max_retries": 1
})

Timeouts

The SDK defaults to a 60 second timeout. You can configure this with a timeout option at the client or request level.

from truefoundry_gateway_sdk import TruefoundryGateway

client = TruefoundryGateway(..., timeout=20.0)

# Override timeout for a specific method
client.agents.responses.create(..., request_options={
    "timeout_in_seconds": 1
})

Custom Client

You can override the httpx client to customize it for your use-case. Some common use-cases include support for proxies and transports.

import httpx
from truefoundry_gateway_sdk import TruefoundryGateway

client = TruefoundryGateway(
    ...,
    httpx_client=httpx.Client(
        proxy="http://my.test.proxy.example.com",
        transport=httpx.HTTPTransport(local_address="0.0.0.0"),
    ),
)

Contributing

While we value open-source contributions to this SDK, this library is generated programmatically. Additions made directly to this library would have to be moved over to our generation code, otherwise they would be overwritten upon the next generated release. Feel free to open a PR as a proof of concept, but know that we will not be able to merge it as-is. We suggest opening an issue first to discuss with us!

On the other hand, contributions to the README are always very welcome!