aurelio-sdk 0.0.16

Aurelio Platform SDK

Aurelio SDK

The Aurelio Platform SDK. API references

Installation

To install the Aurelio SDK, use pip or poetry:

pip install aurelio-sdk

Authentication

The SDK requires an API key for authentication. Get key from Aurelio Platform. Set your API key as an environment variable:

export AURELIO_API_KEY=your_api_key_here

Usage

See examples for more details.

Initializing the Client

from aurelio_sdk import AurelioClient
import os

client = AurelioClient(api_key=os.environ["AURELIO_API_KEY"])

or use asynchronous client:

from aurelio_sdk import AsyncAurelioClient

client = AsyncAurelioClient(api_key="your_api_key_here")

Chunk

from aurelio_sdk import ChunkingOptions, ChunkResponse

# All options are optional with default values
chunking_options = ChunkingOptions(
    chunker_type="semantic", max_chunk_length=400, window_size=5
)

response: ChunkResponse = client.chunk(
    content="Your text here to be chunked", processing_options=chunking_options
)

Extracting Text from Files

PDF Files

from aurelio_sdk import ExtractResponse

# From a local file
file_path = "path/to/your/file.pdf"

response_pdf_file: ExtractResponse = client.extract_file(
    file_path=file_path, quality="low", chunk=True, wait=-1
)

Video Files

from aurelio_sdk import ExtractResponse

# From a local file
file_path = "path/to/your/file.mp4"


response_video_file: ExtractResponse = client.extract_file(
    file_path=file_path, quality="low", chunk=True, wait=-1
)

Extracting Text from URLs

PDF URLs

from aurelio_sdk import ExtractResponse

# From URL
url = "https://arxiv.org/pdf/2408.15291"
response_pdf_url: ExtractResponse = client.extract_url(
    url=url, quality="low", chunk=True, wait=-1
)

Video URLs

from aurelio_sdk import ExtractResponse

# From URL
url = "https://storage.googleapis.com/gtv-videos-bucket/sample/ForBiggerMeltdowns.mp4"
response_video_url: ExtractResponse = client.extract_url(
    url=url, quality="low", chunk=True, wait=-1
)

Waiting for completion and checking document status

# Set wait time for large files with `high` quality
# Wait time is set to 10 seconds
response_pdf_url: ExtractResponse = client.extract_url(
    url="https://arxiv.org/pdf/2408.15291", quality="high", chunk=True, wait=10
)

# Get document status and response
document_response: ExtractResponse = client.get_document(
    document_id=response_pdf_file.document.id
)
print("Status:", document_response.status)

# Use a pre-built function, which helps to avoid long hanging requests (Recommended)
document_response = client.wait_for(
    document_id=response_pdf_file.document.id, wait=300
)

Embeddings

from aurelio_sdk import EmbeddingResponse

response: EmbeddingResponse = client.embedding(
    input="Your text here to be embedded",
    model="bm25")

# Or with a list of texts
response: EmbeddingResponse = client.embedding(
    input=["Your text here to be embedded", "Your text here to be embedded"]
)

Response Structure

The ExtractResponse object contains the following key information:

• status: The current status of the extraction task
• usage: Information about token usage, pages processed, and processing time
• message: Any relevant messages about the extraction process
• document: The extracted document information, including its ID
• chunks: The extracted text, divided into chunks if chunking was enabled

The EmbeddingResponse object contains the following key information:

• message: Any relevant messages about the embedding process
• model: The model name used for embedding
• usage: Information about token usage, pages processed, and processing time
• data: The embedded documents

Best Practices

1. Use appropriate wait times based on your use case and file sizes.
2. Use async client for better performance.
3. For large files or when processing might take longer, enable polling for long-hanging requests.
4. Always handle potential exceptions and check the status of the response.
5. Adjust the quality parameter based on your needs. "low" is faster but less accurate, while "high" is slower but more accurate.