ChatGPT Plugin Idea for OLTP & Postgres: Choosing a Citus Shard Key
This short video shows the promise of chat completion models. This prototype shows how you could use OpenAI’s ChatGPT to create a GPT-based shard-key selection specialist to help Azure Cosmos DB for PostgreSQL customers (aka Citus on Azure customers) with OLTP workflows (such as multi-tenant SaaS) choose their Citus sharding key and distribute their Postgres database. How? This video shows how one could carefully craft prompts and feed the chat completion model with curated, real-time data.
Imagine if you were to have access to an OpenAI plugin—as part of the OpenAI plugin ecosystem—and that you could feed real-time information from your Azure Cosmos DB for PostgreSQL clusters to GPT-4. Specifically, when you install and enable the plugin, what if it could fetch your Azure subscription, Azure resource group, and PostgreSQL cluster information—as well as certain schema and table data, given a connection string in your Azure Key Vault. Combining the curated, real-time information coming from the ChatGPT plugin with carefully crafted prompts, you would be able to improve GPT-4’s reasoning and completion skills in the context of Citus shard key selection for your distributed PostgreSQL cluster.
The video shows that, with the help of prompt design and a ChatGPT plugin, GPT-4 could understand the write-heavy nature of the online transaction processing queries and suggest the right distribution strategy. It could also infer that no repartition joins (i.e., resource-intensive data shuffling) would occur given its strategy.
⭐ Stay connected with the Citus open source project on GitHub: https://github.com/citusdata/citus
💡Take the Azure Cosmos DB for PostgreSQL training at https://aka.ms/pg-azurecosmosdb-training
🔔 Subscribe to the Citus monthly technical newsletter: https://aka.ms/citus-newsletter
#PostgreSQL #Citus #AzureCosmosDB