Asynchronous API and Stability Improvements

After some users ran into issues processing documents > 150 pages, we made some changes to our infrastructure to support the build out of an asynchronous API. This API allows you to make a single request to launch a large job, then poll our API endpoint for the job status until it is completed. We processed multiple thousand page documents in minutes with this update.

We also made some improvements to the stability and error handling of the API. We added more descriptive errors and have triggers saved such that new errors will be handled quickly.

Here is an example of usage for our API in Python.


import requests
import time

headers = {"authorization": "Bearer <token>"}
response = requests.post(
    "https://api.reducto.ai/chunk_url_async",
    headers=headers,
    params={"document_url": "https://ragchunking.s3.amazonaws.com/sampledoc.pdf"},
)

response = response.json()

while True:
    response = requests.get(
        "https://api.reducto.ai/" + response["Tracking"]["URL"],
        headers=headers,
    )
    response = response.json()
    if response["status"] == "completed":
        break
    time.sleep(15)

print(response["Results"])

Can read up on the full documentation for the API here:

Launch Async Job

Asynchronous processing

https://docxtract.readme.io/reference/chunk_url_async_chunk_url_async_post

Job Status

After launching a job with Chunk Url Async, you can check on the status of your request with this endpoint.The response can take two different shapes. If the job is in progress:JSON{ "Status": "Processing"} If the job is successful:JSON{ "Status": "Finished", "Url": "https://reducto-processing-bucke...

https://docxtract.readme.io/reference/get_status_status__job_id__get/