RAGIE Python Client SDK

Warnings

• breaking The SDK is currently in beta. Breaking changes may occur between versions without a major version update. Users are strongly advised to pin their usage to a specific package version to prevent unexpected breaks.
  Fix: Pin the `ragie` package version in your `requirements.txt` or `pyproject.toml` (e.g., `ragie==1.15.1`).
• gotcha Improper resource management can lead to open HTTP connections and memory leaks in long-lived applications. The `Ragie` client manages underlying `httpx` clients.
  Fix: Always initialize the `Ragie` client using a context manager (`with Ragie(...) as client:`). This ensures that HTTP connections and other resources are properly closed and released.
• gotcha Documents ingested into Ragie are not immediately ready for retrieval. They go through a lifecycle (e.g., `pending`, `partitioning`, `indexed`, `ready`). Attempting to retrieve from a document that is not yet 'ready' will yield incomplete or no results.
  Fix: Implement a polling mechanism or listen for webhooks to confirm the document's status is 'ready' before attempting retrievals.
• gotcha Invalid or missing API keys are a common source of authentication failures (401 errors). Additionally, exceeding plan-based rate limits or quotas (e.g., page processing, retrieval limits) can result in 402 or 429 errors.
  Fix: Ensure `RAGIE_API_KEY` is correctly set and valid. Monitor your usage against plan limits and consider upgrading your plan if necessary. Refer to the Ragie documentation for error codes and resolutions.

Install

• pip install ragie Pip
• poetry add ragie Poetry

Imports

• Ragie

  from ragie import Ragie

Quickstart

This quickstart demonstrates how to initialize the Ragie client, create a raw text document, wait for its processing to complete (indicated by the 'ready' status), and then perform a retrieval query to get relevant information based on the ingested document. Ensure you have `RAGIE_API_KEY` set as an environment variable.

import os
from ragie import Ragie
from dotenv import load_dotenv

load_dotenv() # Load environment variables from .env file

RAGIE_API_KEY = os.environ.get("RAGIE_API_KEY", "")

if not RAGIE_API_KEY:
    raise ValueError("RAGIE_API_KEY environment variable not set. Get one at https://secure.ragie.ai/api-keys")

try:
    # Initialize the Ragie client using a context manager for resource cleanup
    with Ragie(auth=RAGIE_API_KEY) as ragie_client:
        # Create a raw document
        print("Creating a raw document...")
        document = ragie_client.documents.create_raw(request={
            "data": "The quick brown fox jumps over the lazy dog. A common phrase used to test typewriters."
        })
        print(f"Document created with ID: {document.id}")

        # Wait for the document to be ready (simplified polling for demonstration)
        print("Waiting for document to be ready...")
        while True:
            doc_status = ragie_client.documents.get(document_id=document.id)
            if doc_status.status == "ready":
                print(f"Document {document.id} is ready.")
                break
            # In a real application, you'd add a delay (e.g., time.sleep(1)) and/or timeout

        # Retrieve chunks based on a query
        print("Retrieving chunks...")
        retrieval = ragie_client.retrievals.retrieve(request={"query": "What is a common phrase?"})
        if retrieval.chunks:
            for chunk in retrieval.chunks:
                print(f"Retrieved chunk: {chunk.text}")
        else:
            print("No chunks retrieved for the query.")

except Exception as e:
    print(f"An error occurred: {e}")
    # Consider handling specific ragie.models.RagieError for API-related issues