PDFAgent
A production-ready PDF agent for reading, analysing and conversing with your PDF built for business-critical use cases on retrieval-augmented generation.
Got questions/need support? Join our discord here: https://discord.gg/a3K9c8GRGt
Features
✅ Built-in observability (via Twilix dashboard)
Installation
You can install it easily using the following:
pip install pdfagent
QuickStart
You can get started quickly by grabbing your API key from https://app.twilix.io/.
from PDFAgent import PDFAgent
agent = PDFAgent(
name="AgentExample",
api_key="XXX"
)
Inserting a PDF
You can insert a PDF in just 1 line of code - under the hood, we take care of OCR, preprocessing, vectorising, splitting and storage.
agent.insert_pdf(
"https://www.w3.org/WAI/WCAG21/working-examples/pdf-table/table.pdf"
)
Once it's inserted, we recommend giving it 30 to 60 seconds to properly index.
If you have a local, PDF, we recommend uploading and inserting via the https://app.twilix.io dashboard.
Now ask questions!
agent.ask("What is this PDF about?")
Built-in observatory
Once you set it up - you get a free monitoring and observatory! This includes:
- Monitoring input queries
- Monitoring output queries
- Latency
- Retrieval results
Ask for a co-pilot analysis!
agent.copilot("How many participants are there in total?")
Powerful templating for any desired output:
You can then use out-of-the-box templating where you by inserting a
{reference}
so that users can get a clean abstraction.
agent.template("""How many participants are there in total? Please respond in a JSON with the key `total_participants`.
{reference}
JSON:""")
See More
Interested in exploring further? Find out more from https://docs.twilix.io.