Chunkr AI Python Client

0.3.7 · active · verified Wed Apr 15

Chunkr AI provides a Python client for its open-source document intelligence platform, offering API services for document layout analysis, OCR, and semantic chunking. It transforms complex documents like PDFs, PPTs, Word files, and images into structured, RAG/LLM-ready data, aiming for high-quality output and improved AI application performance. The current version is 0.3.7, and the project shows active development with regular updates and blog posts on new features and models.

Warnings

gotcha The Python SDK is currently in alpha and requires the `--pre` flag for installation. This indicates that the API might be subject to changes before a stable release.
Fix: Always install with `pip install chunkr-ai --pre`. Consult the official documentation for the latest installation instructions and API stability updates.
breaking There are two distinct versions: an open-source AGPL self-hosted version and a fully managed Cloud API. They use different underlying models (community/open-source vs. proprietary in-house), leading to differences in accuracy, speed, and available features (e.g., Excel support is Cloud API exclusive).
Fix: Carefully consider your use case. For production workloads and higher performance, the Cloud API is recommended. Ensure you are using the correct client and configuration for the chosen platform. If self-hosting, be aware of the capabilities and limitations of the open-source models.
gotcha API key is required for authentication with the Chunkr AI Cloud API. Failing to provide a valid key will result in authentication errors.
Fix: Obtain an API key from your Chunkr AI dashboard after creating an account. Set it as an environment variable (e.g., `CHUNKR_API_KEY`) and pass it to the `Chunkr` client upon initialization.
gotcha Suboptimal chunking strategies can lead to increased AI costs, reduced retrieval accuracy, and inconsistent LLM responses. While Chunkr aims for intelligent chunking, users should be aware of how different strategies impact their RAG systems.
Fix: Review Chunkr's documentation on custom chunking strategies, VLM processing, and configuration options. Experiment with `ChunkProcessing`, `Configuration`, and `Tokenizer` parameters to optimize chunk size and content for your specific use case and LLM.

Install

pip install chunkr-ai --pre Install pre-release version

Imports

Chunkr
```
from chunkr_ai import Chunkr
```

ChunkProcessing

from chunkr_ai.models import ChunkProcessing

Configuration

from chunkr_ai.models import Configuration

Tokenizer
```
from chunkr_ai.models import Tokenizer
```

Quickstart

This quickstart demonstrates how to initialize the Chunkr client and submit a document for processing. It assumes you have an API key set as an environment variable. After submission, you can monitor the task status or retrieve the output through the Chunkr AI dashboard or further API calls.

import os
from chunkr_ai import Chunkr
from chunkr_ai.models import ChunkProcessing, Configuration, Tokenizer

# Ensure your Chunkr API key is set as an environment variable CHUNKR_API_KEY
api_key = os.environ.get('CHUNKR_API_KEY', '')
if not api_key:
    print("Warning: CHUNKR_API_KEY environment variable not set. The API call will likely fail.")

chunkr = Chunkr(api_key=api_key)

# Example of processing a document (replace with your document URL or file path)
# This example uses default chunking strategies.
try:
    task = chunkr.parse_document(file_url="https://example.com/document.pdf")
    print(f"Document processing task submitted with ID: {task.task_id}")

    # You can poll for the task status or set up webhooks
    # For a simple quickstart, we'll just acknowledge submission.
    print("Check Chunkr AI dashboard or use get_task_output for results.")

except Exception as e:
    print(f"An error occurred: {e}")

view raw JSON →