Release

Open sourcing our new Chat With Your PDFs LLM app starter template

Use this open source template to rapidly build and ship a conversational AI bot that interacts with and answers questions about PDFs.
Share on social media

We are thrilled to announce the release of Inductor’s newest open source LLM application starter template: Chat With Your PDFs (GitHub repo here). This template is designed to empower developers to quickly build and ship a conversational AI bot that can interact with, answer questions about, and extract information from PDF documents. Whether you’re creating a knowledge base search bot, a customer support bot, a document review assistant, or simply want to build AI that can intelligently answer questions about a collection of PDFs, this starter template provides the foundation that you need.

Much like our previous starter template releases, this template goes beyond just providing the basic structure for an LLM application. It includes an end-to-end developer workflow tailored to the iterative nature of building production-ready LLM applications. Key features of this template include:

  • Robust test suites: Ensure the accuracy and reliability of your PDF chatbot with systematic testing that covers various interaction scenarios.
  • Automated experimentation and optimization: Quickly iterate on and refine your application by experimenting with different models, prompts, and configurations using Inductor hyperparameters.
  • Instant prototyping playground: A secure, auto-generated environment that enables you to prototype and share your PDF chatbot with your team, fully integrated with your test and experimentation setups.
  • Comprehensive logging: Monitor live traffic, understand user interactions, resolve issues, and continuously improve your application with integrated logging and observability.
  • PDF parsing and embedding: Leverage Unstructured for smart chunking and Sentence-Transformers for embedding generation, transforming static PDFs into dynamic, interactive knowledge sources. Complex documents, including those with images and tables, are processed and stored in a ChromaDB vector database for efficient retrieval based on user queries.

This Chat With Your PDFs template is your fast track to developing an LLM-powered bot that can handle complex document queries with ease. Whether you’re just starting or looking to scale, this template makes it simple to get up and running in minutes.

Beyond consumer AI: why enterprises need custom LLM applications

The rise of AI-powered knowledge assistants highlights the growing demand for intelligent systems that can seamlessly access and retrieve information. For example, products like ChatPDF, Liner, Eightify, and Phind leverage different data sources to enable chat assistants capable of querying and interacting with documents, web pages, videos, and research papers. These products enable users to extract key insights and summaries in real time, demonstrating the value of LLM-powered applications in various industries.

These products exemplify how AI can enhance productivity by transforming static content into dynamic, interactive resources.  Yet, while they may work well for general consumer use cases, enterprises often require more customized and complex solutions to meet their unique demands.  For example, businesses dealing with legal documents, HR policy documents, or specialized training materials may need to handle deeper domain-specific queries, process multiple data formats, and integrate seamlessly with existing internal systems – demands that off-the-shelf products often don’t meet.

This is where Inductor comes in. Enterprises often need flexible, tailor-made solutions, and Inductor provides a platform built specifically for LLM app development. Our tools enable developers to rapidly create production-ready AI applications that address unique business needs – for example for use cases such as internal knowledge management, legal compliance, or advanced research.

Unlocking the power of PDF parsing and embedding

At the heart of this starter template is its ability to transform static PDFs into dynamic, conversational experiences. By leveraging Unstructured for smart chunking and Sentence-Transformers for embedding generation, the template ensures that even complex documents are easy to interact with.

Once the PDF content is processed, it's stored in a ChromaDB vector database. This allows the application to efficiently retrieve the most relevant sections based on user queries, ensuring that the LLM has the context it needs to generate accurate, meaningful responses. Using retrieval-augmented generation (RAG), this system seamlessly integrates with an LLM, such as GPT-4o, to provide users with answers and deeper insights from the documents.

Whether you're navigating technical manuals, financial reports, or academic papers, this capability turns static PDFs into a rich source of interactive knowledge.

Supercharge your LLM app development with Inductor

A key feature that sets the Chat With Your PDFs LLM starter template apart is its seamless integration with Inductor, a powerful platform built to streamline the full development and delivery lifecycle of your LLM application. From rapid prototyping to systematic testing and real-time monitoring, Inductor equips developers with the tools needed to build, refine, and deliver applications at every stage. Let’s dive into some of the standout features this integration provides.

Instantly generate a playground UI for experimentation

Experimentation is at the heart of LLM application development, and with Inductor, you can spin up a fully interactive playground UI with a single command. This environment enables you to interactively test different queries, iterate on your application’s behavior, and share your Chat With Your PDFs app with teammates and subject matter experts – all without writing additional code. With your playground, you’ll have a flexible, no-code UI to rapidly iterate on and improve your LLM app. Just run:

inductor playground app:chat_with_pdf

Systematically evaluate your app using test suites

The Chat With Your PDFs template includes a built-in set of test suites designed to test your application’s performance across various scenarios, providing actionable feedback so that you can quickly identify and address any issues. Each test suite includes pre-configured test cases, quality measures, and hyperparameters for evaluation of your app. An example of the results of running an included test suite can be found here.

Whether you're validating how your LLM app handles complex queries or testing its response accuracy across different PDFs, Inductor enables you to rapidly do the evaluations that you need to improve your app efficiently.

Rapidly refine your app with Inductor hyperparameters

Inductor hyperparameters enable you to rapidly test and compare different configurations of your LLM application. Hyperparameters are fully integrated in the Inductor platform where they can be leveraged in playgrounds, test suites, and live A/B testing. Using hyperparameters in playgrounds allows for controlled and collaborative interactive experimentation. Using hyperparameters in test suites allows for systematic selection of optimized configurations.

Here are some examples of hyperparameters included in the Chat With Your PDFs template:

  • query_num_chat_messages: Adjusts the number of previous chat messages used in the query for retrieving relevant information from your vector database.
  • query_result_num: Controls how many results are retrieved from the vector database for each query.
  • query_filter_out_program_messages: Determines whether or not chatbot-generated messages are filtered out from the query sent to the vector database.

By leveraging hyperparameters, you can rapidly run in-depth experiments on multiple different configurations and identify the optimal setup for your application.

Monitor live executions in real time

Understanding how your LLM app performs in real-world conditions is crucial. With Inductor’s live execution logging, every detail of your app’s execution is automatically recorded, including inputs, outputs, the specific text snippets retrieved by the RAG system, and more. This enables you to monitor your app’s performance in real time, giving you invaluable insights into user behavior, system efficiency, and areas for improvement.

🚀 What’s Next?

Ready to build your own AI-powered PDF chatbot? Get started in minutes! Simply visit the GitHub repo, clone the Chat With Your PDFs starter template, and follow the easy steps to start developing your LLM application. It’s that simple! 💻✨

Want to dive deeper? Explore more about Inductor in our documentation or book a demo to see how we can help supercharge your AI projects! 🛠️🚀

Newsletter
Stay in the loop
Receive product announcements,  insights, and best practices for shipping production-ready LLM apps.
Thank you for subscribing!
Oops! Something went wrong while submitting the form.