Reducto Python SDK

Warnings

• breaking The `file` parameter for the `client.upload()` method in `v0.22.0` and later now expects binary file-like objects (e.g., `Path`, `bytes`, or a `(filename, contents, media_type)` tuple) instead of a `str` file path.
  Fix: Update your file upload calls to pass `pathlib.Path('your_file.pdf')`, `open('your_file.pdf', 'rb').read()`, or `('filename.pdf', b'content', 'application/pdf')` instead of just the string path.
• gotcha Uploaded files to Reducto's servers (identified by `reducto://` URIs) are temporary and expire after 24 hours. Attempting to use an expired `file_id` will result in an error.
  Fix: Re-upload the file if you need to process it again after 24 hours, or implement a mechanism to manage `file_id` lifecycles if persistent storage is required.
• gotcha Direct file uploads via `client.upload()` are limited to 100MB. For files larger than 100MB (up to 5GB), you must use the presigned URL upload method.
  Fix: For large files, refer to the Reducto documentation on 'Presigned URL Upload' to handle the upload process correctly.
• gotcha The SDK automatically reads the API key from the `REDUCTO_API_KEY` environment variable by default. If it's not set, client initialization might fail or requests will be unauthenticated.
  Fix: Always set the `REDUCTO_API_KEY` environment variable or explicitly pass `api_key='YOUR_KEY'` when initializing the `Reducto` client.
• gotcha While the async client uses `httpx` by default, for improved concurrency performance, it's recommended to install `aiohttp` and configure the client to use it.
  Fix: Install with `pip install reductoai[aiohttp]` and then initialize `AsyncReducto` with `http_client=DefaultAioHttpClient()` if `aiohttp` is desired.

Install

• pip install reductoai Install stable version
• pip install reductoai[aiohttp] Install with aiohttp for enhanced async performance

Imports

• Reducto

  from reducto import Reducto

• AsyncReducto

  from reducto import AsyncReducto

  Use for asynchronous API calls.
• Path

  from pathlib import Path

  Commonly used for specifying local files for upload.

Quickstart

This quickstart demonstrates how to initialize the `Reducto` client, upload a local file, and then parse its content. It assumes your Reducto API key is set as an environment variable `REDUCTO_API_KEY`. A dummy file is created for demonstration purposes and then cleaned up.

import os
from pathlib import Path
from reducto import Reducto

# Ensure REDUCTO_API_KEY is set in your environment
# export REDUCTO_API_KEY="your_api_key_here"
api_key = os.environ.get('REDUCTO_API_KEY', '')

if not api_key:
    print("Warning: REDUCTO_API_KEY environment variable not set.")
    print("Please set it to run the quickstart example.")
    exit(1)

# Initialize the client
client = Reducto(api_key=api_key)

# Create a dummy file for upload example
dummy_file_path = Path("dummy_document.txt")
dummy_file_path.write_text("This is a dummy document for Reducto.")

try:
    # Upload a document
    upload_response = client.upload(file=dummy_file_path)
    print(f"Uploaded file with ID: {upload_response.file_id}")

    # Parse the document using the uploaded file_id
    parse_result = client.parse.run(input=upload_response.file_id)
    
    # Access the extracted content (example: print first chunk content)
    if parse_result.result and parse_result.result.chunks:
        print(f"First chunk content: {parse_result.result.chunks[0].content}")
    else:
        print("No chunks found in the parsed result.")

except Exception as e:
    print(f"An error occurred: {e}")
finally:
    # Clean up the dummy file
    if dummy_file_path.exists():
        dummy_file_path.unlink()