LANGCHAIN — Meet Connery an Open Source Plugin Infrastructure for OpenGPTs and LLM Apps

Privacy is not something that we’re merely entitled to, it’s an absolute prerequisite. — Marlon Brando

Meet Connery is an open-source framework designed to solve the challenges of integrating LLM-based apps with real-world scenarios. It provides a plugin infrastructure tailored for LLM applications, simplifying integration with third-party services and ensuring safety. In this article, we will explore the components of Connery and provide an example of how to run Connery actions from OpenGPTs.

Plugin Ecosystem

The Connery plugin ecosystem consists of actions and plugins. Actions represent basic tasks, while plugins are collections of related actions. Each plugin is an open-source GitHub repository with TypeScript code of a specific structure.

Plugin Infrastructure

The Runner

The Runner is the core of Connery. It integrates plugins from GitHub, providing a user interface and features for connection management, personalization, and safety. Users can set up their own isolated Runner, configured with a set of plugins and a standardized API for clients.

Clients

Clients serve as the interface through which end-users trigger actions. For example, OpenGPTs from LangChain allows end-users to customize and personalize their GPTs by connecting them to the real world with Connery actions.

Example: Running Connery Actions from OpenGPTs

Let’s consider an example of using Connery actions from OpenGPTs. Suppose we want to summarize a webpage and send it by email. This involves two actions from different plugins: “Summarize public webpage” from the Summarization plugin and “Send email” from the Gmail plugin.

Try the Demo

A demo version of OpenGPTs hosted by LangChain comes with a preconfigured Connery Runner and necessary actions. Users can summarize any article and send it to their email.

Behind the Scenes

Here’s a simplified process of what happens behind the scenes in the demo:

1. The user sends a request to the OpenGPT by submitting a prompt.
2. OpenGPT pulls actions from Connery Runner through the LangChain Toolkit.
3. The Runner prepares actions by downloading source code from GitHub repositories and sending their info back to the OpenGPT.
4. OpenGPT calls action, and the Runner executes it.
5. OpenGPT uses the results to finish its task.

Set up Your Own OpenGPT with Connery Actions

To configure your own OpenGPT and actions:

1. Set up the Connery Runner using the Quickstart guide.
2. Install plugins with the actions you want to use in your agent.
3. Fork the OpenGPTs repo and configure it as specified in the README.
4. Specify the CONNERY_RUNNER_URL and CONNERY_RUNNER_API_KEY environment variables in the .env file of the OpenGPTs to connect it to your Connery Runner.

In conclusion, Connery provides an open-source plugin infrastructure for integrating LLM-based apps with real-world scenarios. By leveraging the plugin ecosystem and infrastructure, developers and end-users can personalize and use Connery actions to enhance their LLM applications.