Hedra Python Library

The Hedra Python library provides convenient access to the Hedra APIs from Python.

Table of Contents

• Installation
• Reference
• Usage
• Custom base URL
• Async Client
• Exception Handling
• Streaming
• Pagination
• File Uploads
• Advanced
  • Access Raw Response Data
  • Retries
  • Timeouts
  • Custom Client
• Contributing

Installation

pip install hedra-sdk

Reference

A full reference for this library is available here.

Usage

Instantiate and use the client with the following:

import time

from hedra import Hedra, InputMinimaxH3

client = Hedra(
    api_key="<value>",
)

submitted = client.jobs.submit_minimax_h3(
    input=InputMinimaxH3(
        prompt="a fox sprinting across fresh snow",
        aspect_ratio="16:9",
        resolution="768p",
        duration_ms=6000,
    ),
)

# Poll until the job reaches a terminal state, then fetch the result envelope.
status = client.jobs.get_status(submitted.job_id)
while status.status in ("IN_QUEUE", "IN_PROGRESS"):
    time.sleep(2)
    status = client.jobs.get_status(submitted.job_id)

result = client.jobs.get(submitted.job_id)
for output in result.outputs or []:
    print(output.url)

Every model has its own submit method — submit_minimax_h3, submit_kling_o3, submit_veo3 and so on — each taking the typed input model for that model (InputMinimaxH3, InputKlingO3, …). The reference lists all of them. To run a model by its public id instead, with an untyped input dict that the API validates at submit time, use client.jobs.submit(model, input={...}).

Instead of polling you can follow the job over server-sent events; see Streaming.

The client authenticates with Authorization: Bearer <api key>; an API key is the <key_id>:<secret> credential from the Hedra console. The API key can also be provided via the HEDRA_API_KEY environment variable, in which case api_key may be omitted:

from hedra import Hedra

# Reads HEDRA_API_KEY from the environment
client = Hedra()

Custom base URL

The client targets https://api.hedra.com/v3 (HedraEnvironment.PRODUCTION). Override it with base_url if you need to point elsewhere (e.g. a mock server in tests):

from hedra import Hedra

client = Hedra(
    api_key="<value>",
    base_url="http://localhost:8000/v3",
)

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 hedra import AsyncHedra, InputMinimaxH3

client = AsyncHedra(
    api_key="<value>",
)


async def main() -> None:
    await client.jobs.submit_minimax_h3(
        input=InputMinimaxH3(
            prompt="a fox sprinting across fresh snow",
            aspect_ratio="16:9",
            resolution="768p",
            duration_ms=6000,
        ),
    )


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 hedra.core.api_error import ApiError

try:
    client.jobs.submit_minimax_h3(...)
except ApiError as e:
    print(e.status_code)
    print(e.body)

Streaming

client.jobs.stream(job_id) follows a job over server-sent events instead of polling. It returns a generator that yields a StatusResponse for every status frame and a JobLogItem for every log frame, and it returns once the job reaches a terminal state:

from hedra import Hedra, StatusResponse

client = Hedra(
    api_key="<value>",
)

# submitted is the SubmitResponse returned by any submit call
for event in client.jobs.stream(submitted.job_id):
    if isinstance(event, StatusResponse):
        print(event.status, event.progress)
    else:
        print(event.level, event.message)

A dropped connection is resumed automatically from the last event id. Tune that with the stream_reconnection_enabled and max_stream_reconnection_attempts options, on the client or per request via request_options.

Pagination

Paginated requests will return a SyncPager or AsyncPager, which can be used as generators for the underlying object.

from hedra import Hedra

client = Hedra(
    api_key="<value>",
)

for job in client.jobs.list(limit=50):
    print(job.job_id, job.status)

# You can also iterate through pages and access the typed response per page
pager = client.jobs.list(...)
for page in pager.iter_pages():
    print(page.response)  # access the typed response for each page
    for item in page:
        print(item)

File Uploads

Media inputs (start_image, end_image, images, audio, video, …) take either a public URL or a file you uploaded first. client.files.upload stores the bytes and returns a presigned URL that is the file's handle for the next hour; pass it back verbatim as a url source:

from hedra import Hedra, InputMinimaxH3, InputMinimaxH3StartImage_Url

client = Hedra(
    api_key="<value>",
)

with open("frame.png", "rb") as f:
    upload = client.files.upload(file=f)

client.jobs.submit_minimax_h3(
    input=InputMinimaxH3(
        prompt="the fox turns toward the camera",
        resolution="768p",
        duration_ms=6000,
        start_image=InputMinimaxH3StartImage_Url(url=upload.url),
    ),
)

file accepts an open binary file, raw bytes, or a (filename, content, content_type) tuple.

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 hedra import Hedra

client = Hedra(...)
response = client.jobs.with_raw_response.submit_minimax_h3(...)
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.jobs.submit_minimax_h3(..., 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 hedra import Hedra

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

# Override timeout for a specific method
client.jobs.submit_minimax_h3(..., request_options={
    "timeout": 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 hedra import Hedra

client = Hedra(
    ...,
    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!