We’re thrilled to announce Inductor’s partnership with MongoDB and release our latest open-source LLM application starter template, designed for a documentation Q&A bot leveraging MongoDB vector search. (GitHub repo here). This template is designed to streamline the development process for RAG-based (Retrieval-Augmented Generation based) LLM applications by leveraging the powerful vector search functionality of MongoDB, along with a seamless Inductor integration for rapid prototyping, testing, experimentation, and monitoring.

This starter template not only provides the foundational scaffolding for a RAG-based LLM application but also incorporates an end-to-end developer workflow optimized for rapid iterative development and delivery. With MongoDB vector search and Inductor, you can efficiently implement and optimize data retrieval, and ensure the quality and performance of your LLM application.

Key components of this integrated workflow include:

• MongoDB Vector Search: MongoDB vector search enables fast, scalable data retrieval in order to bring your unique data into your LLM application via retrieval-augmented generation (RAG).
• Advanced Test Suites: Systematically test your LLM application to ensure quality and reliability.
• Hyperparameter Optimization: Automate experimentation to rapidly find the optimal design for your LLM app, considering factors like model choice, prompt configuration, and retrieval augmentation.
• Auto-Generated Playground: Instantly and securely share a prototyping environment that integrates with test suites and hyperparameters for collaborative development.
• Integrated Logging: Monitor live traffic to understand usage, resolve issues, facilitate A/B testing, and continually improve your application.

This new template for developing an LLM-powered documentation Q&A bot takes minutes to set up. With minimal effort, you can integrate and configure the application to work with your own documentation, and benefit from MongoDB's efficient and scalable vector search capabilities.

‍

Benefits of using MongoDB vector search for RAG-based LLM applications

Integrating MongoDB vector search into your RAG-based (Retrieval-Augmented Generation based) LLM application offers a number of advantages - enabling efficient, scalable, production-grade data retrieval in order to easily and rapidly enable your LLM application to operate on your unique data.  Here are the key benefits:

Efficient large-scale data retrieval

MongoDB's vector search capabilities enable the efficient handling of large datasets. By leveraging high-dimensional vectors to represent data, MongoDB allows for fast and accurate retrieval of relevant information, improving the performance of your LLM applications. This is particularly beneficial for RAG systems that require quick access to vast amounts of context to generate accurate responses.

Improved search accuracy

With vector search, MongoDB can perform similarity searches that go beyond traditional keyword-based methods. This means your RAG LLM application can retrieve contextually relevant documents even if the exact keywords aren’t present. This leads to more accurate and meaningful responses, enhancing the overall user experience.

Scalability and flexibility

MongoDB's architecture is designed for scalability, allowing your application to grow seamlessly as your data and user base expand. Whether you're dealing with a few thousand documents or millions, MongoDB can scale to meet your needs without compromising performance. Additionally, its flexible schema supports a variety of data types, making it easier to integrate diverse data sources.

‍

Turbocharge development speed with a seamless integration with Inductor

You will see in this starter template that integrating Inductor ensures that MongoDB’s vector search capabilities are seamlessly combined with the ability to rapidly prototype, test, experiment, and monitor your LLM application.  This enables a rapid, streamlined progression from prototype to production, significantly speeding up time to market for your LLM applications.

Enhanced experimentation and optimization

As seen in this starter template, MongoDB vector search can be easily combined with Inductor's hyperparameter optimization tools. This enables you to rapidly and systematically  experiment with different retrieval configurations, model parameters, and data representations to find the optimal setup for your application. Such iterative development ensures that you can rapidly and continually improve your LLM application’s accuracy and efficiency.

Inductor Custom Playgrounds enable you to auto-generate a powerful, instantly shareable playground for your LLM app with a single CLI command - and run it within your environment. Playgrounds provide a developer-first approach to prototype and iterate on LLM programs fast, as well as loop collaborators (including less-technical collaborators) into your development process, and get their feedback early and often.

Rigorous and continuous evaluations

An Inductor test suite is included with the documentation Q&A bot application to evaluate its performance and enable you to systematically test and improve. The included test suite consists of a set of test cases, each containing a set of input (i.e., argument) values for the LLM application and an example of an output value that should be considered high-quality or correct. The test suite also includes a set of quality measures specifying how to evaluate the output of the LLM application. Quality measures can be programmatic, human, or LLM-powered. Using Inductor test suites you can:

• Rapidly customize quality evaluation for your use case
• Auto-generate shareable UIs for human evals, and automate with rigorous LLM-powered evals
• Construct, evolve, and share test cases
• Automatically orchestrate test suite execution

Robust monitoring and logging

With integrated logging capabilities, you can monitor search queries and retrieval performance in real time. This helps in identifying bottlenecks, understanding user behavior, and resolving issues quickly. The detailed logs also facilitate A/B testing, enabling data-driven decisions to further enhance your application.

Continuous and cost-efficient improvement

By leveraging Inductor’s LLM app development platform and MongoDB vector search within your RAG LLM application, you can achieve a higher level of performance, accuracy, and scalability. This powerful combination ensures that your LLM applications are well-equipped to handle complex queries and provide users with precise, contextually relevant responses.

The documentation Q&A bot application

The LLM-powered documentation Q&A bot leveraging MongoDB vector search (GitHub repo here) is a RAG-based LLM application that answers questions using one or more Markdown documents as a data source to provide context. This starter template is intended for use cases like Q&A on developer documentation that have one or more Markdown documents on which you would like to provide a question-answering (Q&A) capability.

Get started

To get started in minutes, visit the GitHub repo, clone the documentation Q&A starter template leveraging MongoDB vector search, and follow the simple steps provided to start systematically developing your LLM application.

‍