Chatgpt retrieval plugin

ServerOpenAI offers plugins that allow ChatGPT to interface with API endpoints to retrieve external information. There is a variety of examples and instructions given in Introduction - OpenAI ChatGPT Plugins. ChatGPT plugins are currently available in beta.The ChatGPT retrieval plugin provides an easy way to read data from a vector database using semantic similarity, and upsert new data into the database. For example, user conversations with ChatGPT, and their embedding representations can be stored for later retrieval. With PostgreSQL as a supported datastore, the retrieval plugin can store and interact with a PostgreSQL database.The diagram below depicts the interactions between the retrieval plugin, database and OpenAI/ChatGPT service. The retrieval plugin exposes endpoints such as /query and /upsert, and further details of functionality can be found in the retrieval plugin API Endpoints documentation.Figure 1: Retrieval plugin internals. Data upsert flow (1), (2) & (3). ChatGPT query flow (a), (b), (c) & (d).Initially, you may want to input data into the database that ChatGPT can later recall, as you can see in the diagram above:(1) To input data into the database, /upsert endpoint or /upsert-file can be called.(2) Within the retrieval plugin this will trigger calls to get embedding from the OpenAI API endpoint.(3) After obtaining the embedding, the retrieval plugin stores the input data and embeddings into the backend datastore.In the context of this post, PostgreSQL is used as a datastore which is managed by Azure Database for PostgreSQL - Flexible Server. The embeddings are stored as vector types offered by the pgvector extension.OK, now let's explore the ChatGPT portion of the workflow (indicated by letters) in Figure 1's diagram above:(a) language queries posed by the user to ChatGPT(b) the OpenAI / ChatGPT service will reach out to the /query endpoint of the plugin(c) the plugin then calls OpenAI / ChatGPT

ChatGPT appends the knowledge it gains using retrieval plugin.At Build, Azure OpenAI is announcing a private preview for plugins. Below you can see a sneak peek of using PostgreSQL as retrieval plugin datastore with Azure OpenAI. On posting the question the retrieval plugin was triggered. The plugin queries the PostgreSQL database and returns relevant results to the service. The model used the results from the query to give a response grounded in data entered earlier in the database. Figure 4: Example Azure OpenAI chat session asking about PG extension support in Azure Database for PostgreSQL - Flexible Server.We are just at the beginning of a quantum shift in AI and Data. Some features and functionality mentioned in this post are currently in beta/private preview mode. As we go through this journey with OpenAI / ChatGPT and Azure, there are interesting challenges of data privacy, performance, and enterprise security which will have to be addressed. So... there is a lot more to come. Until next time, goodbye, and happy ChatGPT-ing. !Here are a few links where you can learn more on the topics relevant to this post:Azure Database for PostgreSQL – Flexible ServerAzure Cosmos DB for PostgreSQLGitHub - pgvector/pgvector: Open-source vector similarity search for PostgresChatGPT plugins (openai.com)GitHub - openai/chatgpt-retrieval-plugin: The ChatGPT Retrieval Plugin lets you easily find personal or work documents by asking questions in natural language.Azure OpenAI Service – Advanced Language Models | Microsoft Azure

To add your bearer token).Testing in ChatGPTTo test a locally hosted plugin in ChatGPT, follow these steps:Run the API on localhost: poetry run devFollow the instructions in the Testing a Localhost Plugin in ChatGPT section of the README.For more detailed information on setting up, developing, and deploying the ChatGPT Retrieval Plugin, refer to the full Development section below.AboutPluginsPlugins are chat extensions designed specifically for language models like ChatGPT, enabling them to access up-to-date information, run computations, or interact with third-party services in response to a user's request. They unlock a wide range of potential use cases and enhance the capabilities of language models.Developers can create a plugin by exposing an API through their website and providing a standardized manifest file that describes the API. ChatGPT consumes these files and allows the AI models to make calls to the API defined by the developer.A plugin consists of:An APIAn API schema (OpenAPI JSON or YAML format)A manifest (JSON file) that defines relevant metadata for the pluginThe Retrieval Plugin already contains all of these components. Read the Chat Plugins blogpost here, and find the docs here.Retrieval PluginThis is a plugin for ChatGPT that enables semantic search and retrieval of personal or organizational documents. It allows users to obtain the most relevant document snippets from their data sources, such as files, notes, or emails, by asking questions or expressing needs in natural language. Enterprises can make their internal documents available to their employees through ChatGPT using this plugin.The plugin uses OpenAI's text-embedding-ada-002 embeddings model to generate embeddings of document chunks, and then stores and queries them using a vector database on the backend. As an open-source and self-hosted solution, developers can deploy their own Retrieval Plugin and register it with ChatGPT. The Retrieval Plugin supports several vector database providers, allowing developers to choose their preferred one from a list.A FastAPI server exposes the plugin's endpoints for upserting, querying, and deleting documents. Users can refine their search results by using metadata filters by source, date, author, or other criteria. The plugin can be hosted on any cloud platform that supports Docker containers, such as Fly.io, Heroku or Azure Container Apps. To keep the vector database updated with the latest documents, the plugin can process and store documents from various data sources continuously, using incoming webhooks to the upsert and delete endpoints. Tools like Zapier or Make can help configure the webhooks based on events or