Snowflake published a technical blog post showing how Cortex Code can be used with Openflow to make routine data-integration work more conversational and interactive. The post demonstrates that engineers can use natural language to generate reviewable plans, inspect proposed changes, and approve steps before anything runs-speeding pipeline build and troubleshooting while reducing setup and maintenance friction. That matters because it aims to shorten the time from raw data to usable, AI-ready datasets by surfacing decisions and requiring explicit approval.
Openflow is presented as a native data connectivity service built on Apache NiFi that supports common ingestion patterns such as change — data-capture (CDC), Kafka ingestion, SaaS connectors, and file-based sources. Snowflake notes Openflow can run on Snowflake — managed infrastructure or in Bring Your Own Cloud (BYOC) deployments, and it connects directly into Snowflake without requiring additional pipeline tooling or staging layers — allowing flows to be orchestrated closer to the data platform.
Cortex Code appears inside Snowsight and is also available via CLI or a Desktop App, acting as an AI coding agent tailored to integration tasks. It includes a dedicated Openflow skill that understands connector behaviors, common configuration patterns, authentication models, and runtime signals. The agent operates with environment context so users do not need to re-explain setups on every interaction, bringing connector — specific knowledge and live signals into the same conversational flow.
The blog walks through a conversational build of a CDC pipeline that replicates MySQL data from AWS RDS into Snowflake. It shows Cortex Code laying out a reviewable plan and then executing discrete steps: configuring connector parameters, enabling controller services required for database connectivity, and starting the flow. are needed. Snowflake stresses the same conversational workflow applies across connector types, calling out PostgreSQL, Oracle, Apache Kafka and SaaS sources as examples. The combined approach is pitched as a way to avoid switching between different tools or hunting for context about who last modified a pipeline, because Cortex Code carries connector — specific guidance and runtime signals into a single interactive experience.
Overall, the stack is positioned to reduce the operational burden of stitching systems, managing credentials, and handling edge cases while keeping engineers in control. By showing exactly what will happen and requiring approval before changes are applied, the workflow intends to lower operational risk as pipelines are built, adjusted and monitored.
Sources
Replies (0)
No replies in this topic yet.
