# Snowflake Builders

Best practices, tips & tricks from Snowflake experts and community

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Do You Still Need Airflow for Snowflake Pipelines? An Honest Answer From Someone Who's Run Both

DevFeed: [Do You Still Need Airflow for Snowflake Pipelines? An Honest Answer From Someone Who's Run Both](<https://devfeed.tech/articles/do-you-still-need-airflow-for-snowflake-pipelines-an-honest-answer-from-someone-who-s-run-both-145278.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/do-you-still-need-airflow-for-snowflake-pipelines-an-honest-answer-from-someone-whos-run-both-0e8d4a5a9036?source=rss----34b6daafc07---4>)

Author: Sanket Prabhu

Published: 2026-11-20T01:56:59Z

Content type: opinion

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [Snowflake streaming data architecture](<https://devfeed.tech/topics/snowflake-streaming-data-architecture.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [airflow](<https://devfeed.tech/tags/airflow.md>), [apache-airflow](<https://devfeed.tech/tags/apache-airflow.md>), [dag](<https://devfeed.tech/tags/dag.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [task-graph](<https://devfeed.tech/tags/task-graph.md>)

### AI overview

Snowflake task graphs may now cover many pipelines that previously needed Airflow, but the new graph object, atomic deployments, advanced branching, and DCM support are described as Preview features. The article says native graphs suit end-to-end Snowflake workloads owned by one team, with data arriving in Snowflake and a preference for centralized lineage, logs, cost, and access. Airflow remains useful for cross-system dependencies, routine date-range backfills, and a company-wide orchestration control plane; Preview risk may also argue against moving critical production pipelines. A suggested middle ground is to let Airflow handle external readiness and triggers, then execute one Snowflake task graph for retries, branching, notifications, and transformations. Regular tasks with AFTER dependencies are identified as generally available.

### Source excerpt

Snowflake task graphs just grew up. That doesn't mean your DAGs are obsolete. Apache Airflow and Snowflake Graph Every time Snowflake ships a new orchestration feature, I see the same post on my feed: "Airflow is dead." I don't buy it. But I also don't buy the opposite reflex, where every Snowflake pipeline gets an Airflow DAG by default because "that's how we do orchestration here." I've run Airflow in production and I've spent about a decade building data platforms on Snowflake. Over that time, my answer to "do we need Airflow for this?" has changed more than once. With the task graph changes Snowflake released in late October, it changed again. If every step of your pipeline already runs inside Snowflake, Airflow is often a second system you're paying to babysit. That's the short version. The longer version has some important exceptions. Status note up front: the new TASK GRAPH object, atomic deployments, advanced branching and DCM DEFINE TASK GRAPH are all in Preview as of this writing (released Oct 23 and Oct 27, 2026). Regular tasks with AFTER dependencies have been GA for a long time. 1. Why Airflow ended up in so many Snowflake stacks It's worth being fair to Airflow here. Most teams didn't choose it by accident. For years, native Snowflake tasks were fine for simple chains but awkward for real pipelines: Configuration was scattered across every task in the graph Changing a running graph meant suspending the root, editing tasks one by one, and resuming Branching was limited. A child ran when its predecessors succeeded, and handling a failure path cleanly took workarounds Deployment wasn't atomic. A half-applied change could leave the graph in a state nobody intended So teams reached for an orchestrator that already handled all of that. My team was using Airflow because of the operator versatility. Is that still the right call today? 2. What actually changed in October Two releases matter here. Task graphs as a first-class object (Preview, Oct 23). You can no

## Snowflake Is Moving the Database Boundary: Why Postgres, OpenFlow, and Agents Matter More Together

DevFeed: [Snowflake Is Moving the Database Boundary: Why Postgres, OpenFlow, and Agents Matter More Together](<https://devfeed.tech/articles/snowflake-is-moving-the-database-boundary-why-postgres-openflow-and-agents-matter-more-together-143912.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/snowflake-is-moving-the-database-boundary-why-postgres-openflow-and-agents-matter-more-together-a2331da1a745?source=rss----34b6daafc07---4>)

Author: Akshay Sharma

Published: 2026-11-19T15:01:04Z

Content type: opinion

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Snowflake streaming data architecture](<https://devfeed.tech/topics/snowflake-streaming-data-architecture.md>), [Database](<https://devfeed.tech/topics/database.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [analytics-pipeline](<https://devfeed.tech/tags/analytics-pipeline.md>), [application](<https://devfeed.tech/tags/application.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [batch](<https://devfeed.tech/tags/batch.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [capture](<https://devfeed.tech/tags/capture.md>), [change-data-capture](<https://devfeed.tech/tags/change-data-capture.md>), [context](<https://devfeed.tech/tags/context.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-analytics](<https://devfeed.tech/tags/customer-analytics.md>), [dashboard](<https://devfeed.tech/tags/dashboard.md>), [data](<https://devfeed.tech/tags/data.md>), [database](<https://devfeed.tech/tags/database.md>), [decision](<https://devfeed.tech/tags/decision.md>), [postgres](<https://devfeed.tech/tags/postgres.md>), [postgresql](<https://devfeed.tech/tags/postgresql.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-cortex](<https://devfeed.tech/tags/snowflake-cortex.md>), [snowflake-openflow](<https://devfeed.tech/tags/snowflake-openflow.md>)

### AI overview

A fast warehouse does not ensure timely operational decisions when data must cross multiple boundaries. The author recounts a financial-services analytics pipeline whose dashboard showed customer status several hours late despite fast queries, due to friction among operational systems, ingestion, warehouse, semantic logic, and the application. The excerpt identifies change data capture, schema drift, batch coordination, reconciliation, and duplicate customer records as burdens. It argues that AI agents need reliable context because, unlike human analysts, they cannot readily interpret stale data or diagnose pipeline failures. The excerpt ends before explaining how Snowflake Postgres, OpenFlow, and agents address these issues.

### Source excerpt

These views are my own, not those of my employer. Continue reading on Snowflake Builders Blog: Data Engineers, App Developers, AI, & Data Science "

## AI Is Only as Powerful as the Data Behind It: Making Trading Data AI-Ready

DevFeed: [AI Is Only as Powerful as the Data Behind It: Making Trading Data AI-Ready](<https://devfeed.tech/articles/ai-is-only-as-powerful-as-the-data-behind-it-making-trading-data-ai-ready-142025.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/ai-is-only-as-powerful-as-the-data-behind-it-making-trading-data-ai-ready-9e17ab128829?source=rss----34b6daafc07---4>)

Author: oneZero

Published: 2026-11-18T18:01:05Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [AI-Ready Data](<https://devfeed.tech/topics/ai-ready-data.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Data Management](<https://devfeed.tech/topics/data-management.md>), [streaming data analytics](<https://devfeed.tech/topics/streaming-data-analytics.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [challenge](<https://devfeed.tech/tags/challenge.md>), [context](<https://devfeed.tech/tags/context.md>), [data](<https://devfeed.tech/tags/data.md>), [data-enablement](<https://devfeed.tech/tags/data-enablement.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [execution](<https://devfeed.tech/tags/execution.md>), [financial](<https://devfeed.tech/tags/financial.md>), [financial-markets](<https://devfeed.tech/tags/financial-markets.md>), [financial-services](<https://devfeed.tech/tags/financial-services.md>), [financial-services-industry](<https://devfeed.tech/tags/financial-services-industry.md>), [fintech](<https://devfeed.tech/tags/fintech.md>), [flow](<https://devfeed.tech/tags/flow.md>), [generative](<https://devfeed.tech/tags/generative.md>), [llm](<https://devfeed.tech/tags/llm.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [structured](<https://devfeed.tech/tags/structured.md>), [trading](<https://devfeed.tech/tags/trading.md>)

### AI overview

The excerpt argues that AI's usefulness in trading depends on high-quality, high-fidelity business data that is structured, understandable, and accessible in real time. Trading firms already generate abundant transactions, quotes, pricing, client-flow metrics, and other signals; the challenge is making them usable by AI. A generic model alone cannot answer firm-specific questions such as why execution quality changed for a client segment, which liquidity providers perform best under current conditions, or where trading-flow patterns are shifting. The author's takeaway is that AI enablement starts with data enablement.

### Source excerpt

By Andrew Ralich, CEO, oneZero Financial Systems There is a lot of discussion today about what artificial intelligence will mean for financial markets. In an industry where, in the modern era, written communications are mandated to conclude "past performance is not indicative of future results", the impact of LLM derived insights deserves some amount of scrutiny. At oneZero, we see AI as an accelerator in productivity that is exponentially more powerful when it has access to high quality, high fidelity data that can be used to make real-time decisions. The financial services industry has no shortage of data. Trading firms generate enormous volumes of transactions, quotes, pricing information, client-flow metrics and other signals every day. The challenge is making that data accessible, structured and usable enough for AI systems to do something meaningful with it. AI needs more than a model The excitement around generative and agentic AI can sometimes obscure a fundamental reality: an AI model is only as useful as the information it can access and understand. For a financial institution, that means AI needs access to the organization's own business data and context. More importantly, in the dynamic conditions generated in the recent period of capricious market volatility, that data needs to be accessible in real-time. A trading firm's most valuable questions are often highly specific: Why did execution quality change for a particular client segment? Which liquidity providers are performing best under current market conditions? Where are patterns emerging, or changing, in trading flow? What other factors may be influencing pricing, execution or risk? These are not questions that can be answered effectively by a generic AI model alone. They require high-quality proprietary data, properly structured and made accessible to the tools that a firm is using. That is why we believe AI enablement starts with data enablement. Lowering the barrier to valuable data Historically,

## From Prompts to Management Reports: Snowflake HTML Reports with CoCo

DevFeed: [From Prompts to Management Reports: Snowflake HTML Reports with CoCo](<https://devfeed.tech/articles/from-prompts-to-management-reports-snowflake-html-reports-with-coco-140791.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/from-prompts-to-management-reports-snowflake-html-reports-with-coco-2febecfc5e8c?source=rss----34b6daafc07---4>)

Author: Omid Karami

Published: 2026-11-18T15:01:04Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [finops](<https://devfeed.tech/topics/finops.md>), [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [data analyst](<https://devfeed.tech/topics/data-analyst.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-assisted](<https://devfeed.tech/tags/ai-assisted.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [attribution](<https://devfeed.tech/tags/attribution.md>), [coco](<https://devfeed.tech/tags/coco.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data](<https://devfeed.tech/tags/data.md>), [data-engineer](<https://devfeed.tech/tags/data-engineer.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-platform](<https://devfeed.tech/tags/data-platform.md>), [engineer](<https://devfeed.tech/tags/engineer.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [experiment](<https://devfeed.tech/tags/experiment.md>), [explore](<https://devfeed.tech/tags/explore.md>), [exploring](<https://devfeed.tech/tags/exploring.md>), [finops](<https://devfeed.tech/tags/finops.md>), [html](<https://devfeed.tech/tags/html.md>), [management](<https://devfeed.tech/tags/management.md>), [prompts](<https://devfeed.tech/tags/prompts.md>), [reports](<https://devfeed.tech/tags/reports.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-coco](<https://devfeed.tech/tags/snowflake-coco.md>)

### AI overview

Snowflake CoCo and HTML Reports can turn Account Usage queries into shareable cost reports, but the report's usefulness depends on validating its data and SQL. In this case study, the report covered service, warehouse, query, storage, and CoCo usage; observed consumption spanned only November 9-10 of the requested 30-day window, and database storage returned no rows. Service-level CoCo metering was 5.1573 credits versus 5.3893 at the model level; warehouse figures also differed, with causes unverified. The reviewed SQL aggregated CoCo credits and tokens separately by model before joining, addressing a potential join-fanout concern without proving full financial reconciliation. HTML Reports are published snapshots: refreshing workspace data does not update a published artifact, which must be republished. During Public Preview, scheduled refresh and delivery were unavailable. Because viewers may access embedded data without source-table privileges, authors should review results, sensitive identifiers, and audience permissions before publishing. The report supports investigation, not automatic cost optimization or verified savings.

### Source excerpt

Exploring Snowflake's new HTML Reports capability through a practical cost consumption report -- and how AI-assisted reporting can support FinOps, cost visibility, and better platform decisions. AI-generated image created using ChatGPT by OpenAI.TL;DR Snowflake HTML Reports, introduced in Public Preview in November 2026, let practitioners use Snowflake CoCo to turn query results into shareable reports. This case study demonstrates a Snowflake Cost Consumption Report generated using the bundled cost-intelligence skill. It includes the original prompt, executed SQL, report screenshots, and validation findings. The experiment illustrates how AI-assisted reporting can support management decisions and FinOps, while highlighting the importance of data completeness, SQL verification, and transparent limitations. 1. From Data to Decisions: The Management Reporting Challenge "Where are our Snowflake credits going?" Answering this question often requires examining several Account Usage views. Service metering, warehouse consumption, query attribution, storage, and AI-assisted development activity each provide different perspectives. For Data Platform Teams, the challenge extends beyond retrieving metrics. Technical results must be organized into reports that stakeholders can understand, question, and use to prioritize investigations. I chose to explore whether CoCo and HTML Reports could bridge the gap between technical data exploration and management reporting. As a data engineer working with Snowflake, I see value in making platform consumption understandable beyond engineering teams. Cost consumption was a practical starting point because Snowflake already provides detailed Account Usage data. This allowed me to evaluate two things together: the AI-assisted reporting experience and the accuracy of the generated results. The objective was not to demonstrate automatic cost optimization or claim financial savings. It was to investigate whether an AI-assisted workflow could pro

## Building an Agent Factory for Data Migrations: a Small Test with Snowflake

DevFeed: [Building an Agent Factory for Data Migrations: a Small Test with Snowflake](<https://devfeed.tech/articles/building-an-agent-factory-for-data-migrations-a-small-test-with-snowflake-139492.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/building-an-agent-factory-for-data-migrations-a-small-test-with-snowflake-9d4622c8f2c7?source=rss----34b6daafc07---4>)

Author: Eylon Steiner

Published: 2026-11-17T18:01:03Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [supervisor agent pattern](<https://devfeed.tech/topics/supervisor-agent-pattern.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [ai-experiments](<https://devfeed.tech/topics/ai-experiments.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [audit](<https://devfeed.tech/topics/audit.md>)

Tags: [agent-factory](<https://devfeed.tech/tags/agent-factory.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-migration](<https://devfeed.tech/tags/data-migration.md>), [data-migrations](<https://devfeed.tech/tags/data-migrations.md>), [database-migration](<https://devfeed.tech/tags/database-migration.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [postgresql](<https://devfeed.tech/tags/postgresql.md>), [proof-of-concept](<https://devfeed.tech/tags/proof-of-concept.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [tests](<https://devfeed.tech/tags/tests.md>)

### AI overview

A small Pagila proof of concept migrated PostgreSQL views and functions to Snowflake using parallel AI translators, scripted comparisons, and an independent AI verifier. Scripts handled dependency waves, deployed translations only in throwaway schemas, and compared results on real data and edge cases; human decisions were logged for reuse. All 15 tables passed 352 profile checks after a timestamp-precision bug in the check was fixed. Among 12 proven logic objects, repeated runs and one run with a cheaper translator passed, while planted-bug trials showed data diffs missed 3 of 5 bugs and edge tests caught all 5 (a small sample tested once). The author stresses that matching real data can be accidental, empty results prove little, and harness errors caused many apparent failures--so checks themselves need scrutiny.

### Source excerpt

Building an agent factory for data migrations: a small test with Snowflake How I moved a PostgreSQL database's views and functions to Snowflake with parallel AI agents, scripted checks and an independent AI reviewer, and what broke along the way. The short version I built a small proof of concept (PoC) of an "agent factory" for SQL migration: AI translators working in parallel, scripts that test every translation against the real data, and a second, independent AI whose only job is to find what the tests missed. The test case: Pagila, a public PostgreSQL sample database, moved to Snowflake. The goal of this architecture is to migrate a database's logic (views, functions, procedures), not just its data, at a scale where "one AI plus one human reviewer" stops working, and without ever letting an AI grade its own work: Scale: many objects translated in parallel, in dependency order. Trust: scripts prove every translation against the real data and edge cases, then an independent reviewer looks for what the scripts missed. Human time only where it matters: people answer clear questions the factory can't decide, and each answer is reused. Why a PoC: before using this on a real migration, I wanted to learn on something small and public which parts work, where the agents fail, and what the scripts must catch. The most useful lesson: matching data is not the same as correct logic. Some translations "matched" PostgreSQL on the real data purely by luck. If you have never built anything with AI agents, this post is for you. I explain every part. 1. The problem: copying tables is easy, copying logic is not A database migration has two halves: Data: tables and rows. Boring, mechanical, solved by tools. Code: views, functions and procedures written in the source's SQL dialect. Every one must be rewritten for the target and checked. The code half is where projects stall. A common setup today is "one AI assistant plus one human": the AI translates an object, the human reviews it, re

## From Streamlit to Snowflake App Runtime: A Demo Migration

DevFeed: [From Streamlit to Snowflake App Runtime: A Demo Migration](<https://devfeed.tech/articles/from-streamlit-to-snowflake-app-runtime-a-demo-migration-136381.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/from-streamlit-to-snowflake-app-runtime-a-demo-migration-19e6e81d9873?source=rss----34b6daafc07---4>)

Author: Daan Bakboord

Published: 2026-11-17T01:03:18Z

Content type: tutorial

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [cloud-migration](<https://devfeed.tech/topics/cloud-migration.md>), [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>), [Streamlit](<https://devfeed.tech/topics/streamlit.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>)

Tags: [apache](<https://devfeed.tech/tags/apache.md>), [apache-ossie](<https://devfeed.tech/tags/apache-ossie.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [builds](<https://devfeed.tech/tags/builds.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [custom](<https://devfeed.tech/tags/custom.md>), [daanalytics](<https://devfeed.tech/tags/daanalytics.md>), [data-superhero](<https://devfeed.tech/tags/data-superhero.md>), [demo](<https://devfeed.tech/tags/demo.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [error](<https://devfeed.tech/tags/error.md>), [interface](<https://devfeed.tech/tags/interface.md>), [js](<https://devfeed.tech/tags/js.md>), [metric](<https://devfeed.tech/tags/metric.md>), [metrics](<https://devfeed.tech/tags/metrics.md>), [migration](<https://devfeed.tech/tags/migration.md>), [moving](<https://devfeed.tech/tags/moving.md>), [next](<https://devfeed.tech/tags/next.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [semantic-view](<https://devfeed.tech/tags/semantic-view.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [streamlit](<https://devfeed.tech/tags/streamlit.md>)

### AI overview

The Apache Ossie demo moved from Streamlit in Snowflake to a Next.js app on Snowflake App Runtime while keeping its SQL, data, and semantic view unchanged. The semantic view implements the metrics queried by the app, so replacing the interface did not require rewriting metric definitions; the demo showed different revenue totals with and without that view (115,550 and 103,200). The author valued App Runtime's control over layout, ordinary links, and server-rendered pages. Streamlit remains a Python-first option when its widgets meet the need; a custom web interface entails owning a separate interface. This was a five-question demo migration, not a performance benchmark or a complete production migration. The excerpt begins describing implementation changes but cuts off before giving further migration or deployment steps.

### Source excerpt

Moving a data app should not change what its numbers mean. I moved the Apache Ossie demo from Streamlit in Snowflake to Snowflake App Runtime to try a different interface around the same metric definitions. This walkthrough covers the Next.js rewrite, deployment and the authentication error I encountered. In the same table, different revenue the demo answered one question twice: 115,550 without a semantic view, 103,200 with it. That demo ran as a Streamlit in Snowflake app. I have now moved it to Snowflake App Runtime, as a Next.js app that Snowflake builds and runs next to the data. The numbers did not move. That was the point of the exercise. The Ossie YAML describes the intended definitions; the Snowflake semantic view implements the metrics queried by the app. Keeping those definitions outside the interface meant I could replace the interface without rewriting them. This is a migration of a small demo, with five comparison questions. It is not a performance benchmark or a complete production migration. The code is in the Apache Ossie demo repository. Why try a different runtime? For this demo, the useful change was control over the interface: ordinary links, server-rendered pages and a custom layout. Streamlit had already made it easy to build the original data app. Moving to Next.js let me shape how readers navigate the explanation and share a particular comparison. I would make that choice based on the application you need to build. If Python widgets already meet the requirement, keeping Streamlit avoids owning a separate web interface. Snowflake describes the distinction in its App Runtime overview: Streamlit offers a Python-first approach, while App Runtime supports custom web applications. For me, the architectural question was whether the metric definitions would survive that change intact. What changed The SQL, the data and the semantic view stayed exactly as they were. Everything between the user and the query changed: one Python app file became a small

## I benchmarked Snowflake Postgres mirroring for two days. Here are the numbers.

DevFeed: [I benchmarked Snowflake Postgres mirroring for two days. Here are the numbers.](<https://devfeed.tech/articles/i-benchmarked-snowflake-postgres-mirroring-for-two-days-here-are-the-numbers-136382.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/i-benchmarked-snowflake-postgres-mirroring-for-two-days-here-are-the-numbers-b50eba22b8de?source=rss----34b6daafc07---4>)

Author: Ondřej Blaha

Published: 2026-11-17T01:00:34Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Snowflake streaming data architecture](<https://devfeed.tech/topics/snowflake-streaming-data-architecture.md>), [elt](<https://devfeed.tech/topics/elt.md>), [Transactions](<https://devfeed.tech/topics/transactions.md>), [data type](<https://devfeed.tech/topics/data-type.md>), [Apache Iceberg](<https://devfeed.tech/topics/apache-iceberg.md>)

Tags: [analytics](<https://devfeed.tech/tags/analytics.md>), [arrays](<https://devfeed.tech/tags/arrays.md>), [batch](<https://devfeed.tech/tags/batch.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [gotchas](<https://devfeed.tech/tags/gotchas.md>), [impact](<https://devfeed.tech/tags/impact.md>), [latency](<https://devfeed.tech/tags/latency.md>), [load](<https://devfeed.tech/tags/load.md>), [mirroring](<https://devfeed.tech/tags/mirroring.md>), [numbers](<https://devfeed.tech/tags/numbers.md>), [oltp](<https://devfeed.tech/tags/oltp.md>), [operations](<https://devfeed.tech/tags/operations.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [postgres](<https://devfeed.tech/tags/postgres.md>), [postgresql](<https://devfeed.tech/tags/postgresql.md>), [postgresql-replication](<https://devfeed.tech/tags/postgresql-replication.md>), [refresh](<https://devfeed.tech/tags/refresh.md>), [replication](<https://devfeed.tech/tags/replication.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [source](<https://devfeed.tech/tags/source.md>), [two-days](<https://devfeed.tech/tags/two-days.md>)

### AI overview

On a Snowflake Postgres STANDARD_M instance in one AWS region, mirrored rows reached the near-real-time `TABLE$live` view in about 26 seconds at the median and 37 seconds at p90 across refresh intervals; the regular table could lag by up to one interval. Under the tested write load, moving from a 30-second to a 2-minute interval cut apply-task cost 73%, while small OLTP transactions showed little source impact and batch inserts into mirrored tables were 20-30% slower. The author recommends starting at 2 minutes, using `$live` for operational freshness, adding primary keys before mirroring, and committing bulk loads in chunks of up to 100,000 rows. Important limits include lossy mappings for some PostgreSQL types, `$changes` history being cleared by column changes, and `RESTART_MIRROR` triggering a full table copy. Results are from one instance size and region with limited repetitions, and are not an SLA.

### Source excerpt

Latency distributions, cost per refresh interval, impact on the source database, bulk loads and operations. Measured on the smallest instance that supports mirroring. Snowflake Postgres, managed PostgreSQL running inside a Snowflake account, has been generally available since 24 February 2026, and data mirroring into Snowflake followed on 15 October 2026. I didn't just want to see that it works. I wanted real numbers under a simulated load, so that I can tell clients how it actually behaves and compare the cost with what we had expected. The marketing says "seconds" and "sub-minute", and nobody had published a latency distribution, a cost per interval or the impact on the source database, so I spent two days measuring them. TL;DR Freshness: new rows show up in TABLE$live after ~26 s (median), ~37 s (p90), no matter whether the refresh interval is 30 seconds or 10 minutes. The table itself lags behind $live by up to one refresh interval. Cost: the serverless apply task cost 0.063 credits/h idle at a 30-second interval and 0.25 credits/h under a steady write load. Switching to 2 minutes cut the loaded cost by 73 %, to 10 minutes by 88 %. Source impact: small OLTP transactions barely notice the mirror. Batch inserts into a mirrored table were 20-30 % slower than into a non-mirrored table on the same instance. Operations: SUSPEND_MIRROR took 2.5-4.7 s under load, RESTART_MIRROR re-copies every table despite saying "soft", adding an 8M-row table to a running mirror took under a minute, and the replication slot survived every suspend/resume of the instance. Gotchas: numeric without precision becomes FLOAT, json/jsonb become TEXT, 2-D arrays become NULL, PostgreSQL refuses ADD COLUMN ... DEFAULT <volatile> on mirrored tables, any ADD, DROP or RENAME COLUMN wipes that table's $changes history, and tables without a primary key turn insert-only. Why this matters Every team that runs an application on Postgres and analytics in Snowflake owns a pipeline between the two. It is usu

## Treat LLM Output as Evidence, Not Truth: A Production Pattern for Sentiment and NER

DevFeed: [Treat LLM Output as Evidence, Not Truth: A Production Pattern for Sentiment and NER](<https://devfeed.tech/articles/treat-llm-output-as-evidence-not-truth-a-production-pattern-for-sentiment-and-ner-136383.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/treat-llm-output-as-evidence-not-truth-a-production-pattern-for-sentiment-and-ner-43bdbd4bed6e?source=rss----34b6daafc07---4>)

Author: Akshay Sharma

Published: 2026-11-17T00:58:11Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [ner](<https://devfeed.tech/topics/ner.md>), [Data Quality](<https://devfeed.tech/topics/data-quality.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>)

Tags: [access-controls](<https://devfeed.tech/tags/access-controls.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [cost-management](<https://devfeed.tech/tags/cost-management.md>), [customer-feedback](<https://devfeed.tech/tags/customer-feedback.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [decision](<https://devfeed.tech/tags/decision.md>), [enterprise-scale](<https://devfeed.tech/tags/enterprise-scale.md>), [graph](<https://devfeed.tech/tags/graph.md>), [llm](<https://devfeed.tech/tags/llm.md>), [ner](<https://devfeed.tech/tags/ner.md>), [production](<https://devfeed.tech/tags/production.md>), [score](<https://devfeed.tech/tags/score.md>), [sentiment-analysis](<https://devfeed.tech/tags/sentiment-analysis.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-cortex](<https://devfeed.tech/tags/snowflake-cortex.md>), [sql](<https://devfeed.tech/tags/sql.md>)

### AI overview

A sentiment model can read a customer's polite tone correctly while still misrepresenting the business issue, as in a fraud dispute phrased as thanks. The excerpt argues that LLM sentiment and named-entity extraction should be treated as evidence requiring validation, not as truth: production use also needs source lineage, replayability, access controls, quality thresholds, cost management, and alignment with the business decision. It notes that Snowflake Cortex is the article's context, but this excerpt does not show its implementation pattern or the demos' limitations.

### Source excerpt

How to turn customer comments into validated, graph-ready data with Snowflake Cortex, and what the demos leave out Continue reading on Snowflake Builders Blog: Data Engineers, App Developers, AI, & Data Science "

## Operational Ontology on Snowflake: From Business Meaning to Governed AI Action

DevFeed: [Operational Ontology on Snowflake: From Business Meaning to Governed AI Action](<https://devfeed.tech/articles/operational-ontology-on-snowflake-from-business-meaning-to-governed-ai-action-131048.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/operational-ontology-on-snowflake-from-business-meaning-to-governed-ai-action-6f3ef270d3f2?source=rss----34b6daafc07---4>)

Author: Tianxia Jia

Published: 2026-11-12T21:36:38Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [knowledge-engineering](<https://devfeed.tech/topics/knowledge-engineering.md>), [business logic](<https://devfeed.tech/topics/business-logic.md>), [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Authorization](<https://devfeed.tech/topics/authorization.md>), [function calling](<https://devfeed.tech/topics/function-calling.md>), [schema design](<https://devfeed.tech/topics/schema-design.md>), [AI-Ready Data](<https://devfeed.tech/topics/ai-ready-data.md>)

Tags: [acting](<https://devfeed.tech/tags/acting.md>), [actions](<https://devfeed.tech/tags/actions.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-reasoning](<https://devfeed.tech/tags/ai-reasoning.md>), [also](<https://devfeed.tech/tags/also.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [article](<https://devfeed.tech/tags/article.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [best](<https://devfeed.tech/tags/best.md>), [bottleneck](<https://devfeed.tech/tags/bottleneck.md>), [business](<https://devfeed.tech/tags/business.md>), [governed-ai](<https://devfeed.tech/tags/governed-ai.md>), [meaning](<https://devfeed.tech/tags/meaning.md>), [ontology](<https://devfeed.tech/tags/ontology.md>), [operational](<https://devfeed.tech/tags/operational.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

The proposed Snowflake Operational Ontology architecture links governed business definitions to live operational state, domain rules, decision services, AI reasoning, and controlled execution. Its lifecycle is Observe -> Understand -> Evaluate -> Decide -> Execute -> Verify: Cortex Agents investigate and explain options, while specialized services validate or simulate plans; an authorized human approves actions before procedures dispatch them to external systems. Versioned snapshots, isolated simulation results, idempotency, and observed outcomes support traceability and verification. The article stresses that a system's acknowledgement does not prove an action completed. Snowflake Business Ontology governs concepts and relationships but does not execute rules or actions. Standard tables are the default for history and events; hybrid tables can serve selected indexed transactional workloads, with regional availability and quotas, but do not support streams. Applications therefore need another supported event source or dispatcher for hybrid-table state changes. The examples are illustrative synthetic reference implementations, not production integrations.

### Source excerpt

A reference architecture for AI agents to reason over operational state, simulate decisions, execute governed actions, and verify real-world outcomes. From Enterprise Knowledge to Operational Decisions At 1:30 PM, packing station PB-04 goes offline in a distribution center. Orders begin accumulating, and a carrier scheduled to depart at 5:00 PM may leave before its shipments are ready. An AI agent connected to warehouse data can identify the failed station, trace affected orders, and explain the shipping risk. But the warehouse manager needs to decide what to do. Can qualified workers move to another station? Can orders be rerouted? Which recovery plan has the best chance of meeting the deadline without creating another bottleneck? Answering these questions requires current operational state, resource capabilities, business constraints, and a way to evaluate proposed changes. Acting on a decision also requires authorization and confirmation that the intended changes occurred. In the earlier Ontology on Snowflake series, I explored how ontology definitions, enterprise data, semantic models, and AI reasoning establish a shared understanding of business information. Snowflake now has a product of Business Ontology for governed business definiton (Public Preview soon). This article builds on that foundation by introducing a Snowflake Operational Ontology Reference Architecture. It connects governed business meaning to operational state, constraints, simulation, AI reasoning, and controlled actions. The goal is to establish a reusable pattern. A warehouse, supply chain network, and data center require different operational models, but they can share architectural contracts and a common decision lifecycle. 1. What Makes an Ontology Operational? An ontology defines the concepts, relationships, and semantics of a domain. In our warehouse example, the ontology defines Packing Station, Worker, Order, and Carrier Commitment, along with relationships such as workers operating s

## Streamlit's Floor, React's Ceiling, Inside Snowflake

DevFeed: [Streamlit's Floor, React's Ceiling, Inside Snowflake](<https://devfeed.tech/articles/streamlit-s-floor-react-s-ceiling-inside-snowflake-131049.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/streamlits-floor-react-s-ceiling-inside-snowflake-c78f90e39dc8?source=rss----34b6daafc07---4>)

Author: Vinod S

Published: 2026-11-12T18:01:04Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Streamlit](<https://devfeed.tech/topics/streamlit.md>), [React](<https://devfeed.tech/topics/react.md>), [Front end](<https://devfeed.tech/topics/frontend.md>), [Python](<https://devfeed.tech/topics/python.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>), [Web Accessibility (a11y)](<https://devfeed.tech/topics/web-accessibility.md>)

Tags: [apache-2-0](<https://devfeed.tech/tags/apache-2-0.md>), [dashboard](<https://devfeed.tech/tags/dashboard.md>), [data-visualization](<https://devfeed.tech/tags/data-visualization.md>), [emea](<https://devfeed.tech/tags/emea.md>), [html](<https://devfeed.tech/tags/html.md>), [island](<https://devfeed.tech/tags/island.md>), [kpis](<https://devfeed.tech/tags/kpis.md>), [pandas](<https://devfeed.tech/tags/pandas.md>), [permissions](<https://devfeed.tech/tags/permissions.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [python](<https://devfeed.tech/tags/python.md>), [queries](<https://devfeed.tech/tags/queries.md>), [react](<https://devfeed.tech/tags/react.md>), [simplicity](<https://devfeed.tech/tags/simplicity.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [streamlit](<https://devfeed.tech/tags/streamlit.md>)

### AI overview

RiSiS adds a React island to a Streamlit in Snowflake page so the browser can update linked charts and KPIs during hover and drag, while Python and SiS continue to handle sessions, queries, permissions and deployment. The example sends a 49,000-row deal cube (about 1 MB) to the browser once per session, where it handles filtering; its author says speed is secondary to the interactions and measurements cover only one dashboard and data size. A native Streamlit twin gives the same answers and may be sufficient. Islands require custom styling and accessibility work, add code and a 790 KB per-session bundle, and rely on a pinned Streamlit sanitizer. The article says warehouse runtime supports Streamlit only through 1.52.2 and lacks Components v2, while container runtime is needed for newer features. For a full web stack, Snowflake App Runtime is presented as an option, with server-side APIs, queries and input checks left to the developer.

### Source excerpt

Keep the simplicity of Streamlit in Snowflake, and move the interactions that a rerun cannot do to the front end Picture a revenue review. An analyst opens the pipeline dashboard and points at EMEA on the regional bar chart. A tooltip appears. The other charts and the KPIs do not change. She wants more. She wants every chart and KPI to show the EMEA numbers while she points. Then she wants to drag across two quarters on the trend line and watch bookings and the commit share change during the drag. That is how she finds the quarter where the commit share fell. Without these interactions, she must click, wait for the rerun, read the result and click again, for every guess. I build many Streamlit in Snowflake (SiS) apps. I use SiS because it is the simplest way that I know to show data to people, not to mention that you can have an app up and running with a SINGLE PROMPT. I did not want to move my apps out of SiS. I wanted SiS to support these interactions. I call the result RiSiS (React in SiS). This post shows how RiSiS works, what it costs, and one open question. TL;DR: react-in-streamlit-in-snowflake is an open example. It adds a React island to a SiS page with st.html. An island is one React component at one position on a Streamlit page. Python and SiS continue to control the session, the queries, the permissions and the deploy. The island adds three interactions that a rerun cannot do: hover previews across all charts KPIs that change during a drag linked charts that change together The example is free under Apache-2.0 and runs in your own account. Why I use SiS Streamlit has a low floor. A chart is three lines of Python. SiS adds the parts that I would otherwise build myself: One language for the data. Snowpark, pandas and the queries stay in Python. No hosting work. Snowflake runs the app. Users open the app in Snowsight with their usual sign-in. Known permissions. The app queries with the privileges of its owner (security overview). A role controls what the ap

## Same Table, Different Revenue: Apache Ossie and Snowflake to the Rescue

DevFeed: [Same Table, Different Revenue: Apache Ossie and Snowflake to the Rescue](<https://devfeed.tech/articles/same-table-different-revenue-apache-ossie-and-snowflake-to-the-rescue-129821.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/same-table-different-revenue-apache-ossie-and-snowflake-to-the-rescue-a26e5c99f5c5?source=rss----34b6daafc07---4>)

Author: Daan Bakboord

Published: 2026-11-11T22:01:02Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [apache](<https://devfeed.tech/tags/apache.md>), [apache-ossie](<https://devfeed.tech/tags/apache-ossie.md>), [apache-software-foundation](<https://devfeed.tech/tags/apache-software-foundation.md>), [context](<https://devfeed.tech/tags/context.md>), [count](<https://devfeed.tech/tags/count.md>), [customer](<https://devfeed.tech/tags/customer.md>), [daanalytics](<https://devfeed.tech/tags/daanalytics.md>), [data-governance](<https://devfeed.tech/tags/data-governance.md>), [metric](<https://devfeed.tech/tags/metric.md>), [orders](<https://devfeed.tech/tags/orders.md>), [revenue](<https://devfeed.tech/tags/revenue.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [semantic-layer](<https://devfeed.tech/tags/semantic-layer.md>), [semantic-view](<https://devfeed.tech/tags/semantic-view.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [table](<https://devfeed.tech/tags/table.md>)

### AI overview

The demo shows how shared metric definitions prevent different tools from producing defensible but inconsistent revenue figures. On the sample orders, total_revenue sums amounts only for completed orders (103,200); other metrics use different rules, such as active_customers including completed, shipped, and processing orders. Apache Ossie (Incubating) expresses sources, relationships, dimensions, metrics, filters, and descriptions in a portable YAML contract; it does not execute queries. A Snowflake semantic view implements that contract with metric expressions and relationships so queries can request named metrics without restating their rules. Modeling join paths matters: grouping order revenue by customer region is safe, while joining orders to line items can duplicate order totals, so category analysis should use line-level revenue. The article presents this shared context as a way to keep dashboards, apps, and AI agents consistent, while noting that rules such as excluding open orders from category revenue must be built into the relevant metric.

### Source excerpt

Ask two tools for revenue from the same orders table and both can return a different, defensible number. Here is what that looks like in the demo. The difference is 12,350. Both queries read the same orders and do the arithmetic correctly. What changes is which orders qualify as revenue. In this demo, revenue counts only when an order is completed. Apache Ossie and a Snowflake semantic view make that rule explicit. That definition is context. It is the contract that tells a person, a dashboard, or an AI agent which reading of the table is allowed to be called revenue. Context is the contract A table stores values. Context says what those values mean when someone asks a business question. For a metric, the contract has a few parts: the expression, such as a sum or a count the rows that qualify, such as a status filter the entities it can be sliced by, and how those entities join the other metrics it is built from a description that a human and a model can both read Leave any of those out and the column name does the rest of the work. total_amount looks like revenue. It also looks like revenue before discounts, and like revenue that includes orders that never completed. An analyst picks one reading in a worksheet. A dashboard picks another. An agent, given the same column list, picks a third and sounds sure about it. The sample model shows the distinction with three metrics on the same orders table. total_revenue sums total_amount where order_status = 'completed'. On this data that is 103,200. gross_revenue uses the same completed orders and adds discount_amount back. Same orders, different question: value before the discount rather than after it. active_customers counts distinct customers with an order in completed, shipped, or processing. It is a wider rule than revenue. The two metrics can disagree. In this sample, the four open orders belong to customers who also have a completed order, so the customer count does not move. The next customer who has only a shipped

## Clean-Looking Addresses Still Fail Delivery. Here Is How to Catch Them in Snowflake

DevFeed: [Clean-Looking Addresses Still Fail Delivery. Here Is How to Catch Them in Snowflake](<https://devfeed.tech/articles/clean-looking-addresses-still-fail-delivery-here-is-how-to-catch-them-in-snowflake-129820.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/clean-looking-addresses-still-fail-delivery-here-is-how-to-catch-them-in-snowflake-32cd51d5f4c4?source=rss----34b6daafc07---4>)

Author: Jan Rehorik

Published: 2026-11-11T20:01:01Z

Content type: tutorial

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Remote Procedure Call (RPC)](<https://devfeed.tech/topics/rpc.md>)

Tags: [address-validation](<https://devfeed.tech/tags/address-validation.md>), [api](<https://devfeed.tech/tags/api.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [batch](<https://devfeed.tech/tags/batch.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [carrier](<https://devfeed.tech/tags/carrier.md>), [data](<https://devfeed.tech/tags/data.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-quality](<https://devfeed.tech/tags/data-quality.md>), [dedupe](<https://devfeed.tech/tags/dedupe.md>), [encrypted](<https://devfeed.tech/tags/encrypted.md>), [external](<https://devfeed.tech/tags/external.md>), [fields](<https://devfeed.tech/tags/fields.md>), [formatting](<https://devfeed.tech/tags/formatting.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [install](<https://devfeed.tech/tags/install.md>), [integration](<https://devfeed.tech/tags/integration.md>), [join](<https://devfeed.tech/tags/join.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mailing](<https://devfeed.tech/tags/mailing.md>), [marketplace](<https://devfeed.tech/tags/marketplace.md>), [route](<https://devfeed.tech/tags/route.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

The Snowflake app validates US addresses against postal reference data through a stored procedure that reads a supplied SELECT, sends selected address fields to the Service Objects API, and writes results to a named table. The excerpt's workflow maps seven required input columns, runs a batch with one CALL, then uses DPV codes and errors to route records--for example, accept DPV 1, flag DPV 3 or 4 for unit review, and investigate other outcomes. Original source values can remain intact for review and rollback. Data leaves the account through an approved external access integration, is sent over HTTPS, and the app says it retains nothing. The 90-day or 1,000-record trial requires external access; Snowflake trial accounts may need that capability enabled. Validation confirms deliverability, not resident identity or whether a customer has moved.

### Source excerpt

Clean-Looking Addresses Still Fail Delivery. Here Is How to Catch ThemHow to validate and standardize US addresses in Snowflake using SQL One source writes "26 S Chestnut St." Another writes "26 s chstnt, vntura, californa" with the wrong ZIP code. Both are the same building in Ventura, and both will break a join, a dedupe rule, or a mailing run. Formatting rules in SQL can make strings look alike. They cannot tell you whether an address is a real delivery point, or whether it is missing the suite number the carrier needs. That takes validation against postal reference data, and in most stacks it means an export, an API client, and a re-import to maintain. In this tutorial, you'll use SQL to submit addresses for validation, review the results, and decide what happens to each record. You will: Install Service Objects' US Address Validation and Verification for Snowflake app from the Marketplace. Shape your address columns into the input the app expects, without changing your source table. Validate a batch of records with one CALL. Read the results and route each record with SQL. Every screenshot and result below comes from a real run in a Snowsight worksheet. Architecture DiagramThe app runs in your account. Only the address fields you select cross the boundary, through one integration you approve. The app is installed into your Snowflake account and exposes a stored procedure. You pass it a SELECT statement and an output table. It reads those rows, sends the address fields to the Service Objects API for validation, and writes the results into the table you named. From there, the results are ordinary Snowflake data. The external access integration is the one point where data leaves your account. The app sends address data to the Service Objects API through a single external access integration you grant and revoke, and retains nothing. Data is encrypted in transit over HTTPS. That boundary is worth showing your security reviewer up front. Why formatting rules are not

## Agents Are Code Too: What I Learned Shipping a Cortex Agent Through Azure DevOps

DevFeed: [Agents Are Code Too: What I Learned Shipping a Cortex Agent Through Azure DevOps](<https://devfeed.tech/articles/agents-are-code-too-what-i-learned-shipping-a-cortex-agent-through-azure-devops-129094.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/agents-are-code-too-what-i-learned-shipping-a-cortex-agent-through-azure-devops-fab3adae29e8?source=rss----34b6daafc07---4>)

Author: Sanket Prabhu

Published: 2026-11-11T15:01:07Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Azure DevOps](<https://devfeed.tech/topics/azure-devops.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [CI/CD Pipeline](<https://devfeed.tech/topics/ci-cd-pipeline.md>), [DevOps](<https://devfeed.tech/topics/devops.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [analyst](<https://devfeed.tech/tags/analyst.md>), [azure-devops](<https://devfeed.tech/tags/azure-devops.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [ci-cd-pipeline](<https://devfeed.tech/tags/ci-cd-pipeline.md>), [code](<https://devfeed.tech/tags/code.md>), [config](<https://devfeed.tech/tags/config.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [cortex-agent](<https://devfeed.tech/tags/cortex-agent.md>), [data](<https://devfeed.tech/tags/data.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-platforms](<https://devfeed.tech/tags/data-platforms.md>), [dev](<https://devfeed.tech/tags/dev.md>), [domain](<https://devfeed.tech/tags/domain.md>), [environment](<https://devfeed.tech/tags/environment.md>), [finance](<https://devfeed.tech/tags/finance.md>), [git](<https://devfeed.tech/tags/git.md>), [instructions](<https://devfeed.tech/tags/instructions.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [orchestrator](<https://devfeed.tech/tags/orchestrator.md>), [platform](<https://devfeed.tech/tags/platform.md>), [snowflake-cortex](<https://devfeed.tech/tags/snowflake-cortex.md>)

### AI overview

The main production challenge was safely promoting a finance Cortex Agent across sbx, dev, uat and prod; the author treats the agent as code managed through the existing Azure DevOps path. Store its YAML specification, semantic view, Cortex Search service, grants and evaluation questions in one repo. Use CREATE AGENT ... FROM SPECIFICATION with one environment variable for the target database; COPY GRANTS with OR REPLACE to preserve grants. Deploy the semantic view first, then search, agent and grants in the same release, with least-privilege deployment credentials and approval gates for uat and prod. A successful CREATE AGENT does not ensure runtime behavior: run fixed domain questions after each deployment, checking expected answers and whether questions route to Analyst or Search. The article also describes a Snowflake developer guide that evaluates semantic-view SQL against verified queries and agent answer correctness and tool-selection accuracy; check account release status before relying on agent evaluations and versioning. Grants must cover the agent, search service and objects behind the semantic view.

### Source excerpt

Building the agent took a week. Getting it safely from sandbox to production took longer, and taught me more. Most Cortex Agent content I read ends at the same place: a working demo in Snowsight, a clever question, a correct answer. That is where my real work started. In my recent project, I built a Cortex Agent for the finance domain. The agent itself came together quickly. What took real thought was the part nobody demos: moving it through sbx, dev, uat and prod with the same discipline we already applied to every other Snowflake object on the platform. After roughly a decade of building data platforms, I've learned to be suspicious of anything that only exists in one environment. So here is what I learned treating an agent like code, and what I'd do the same way again. An agent you can't rebuild from Git is a prototype, no matter how good its answers are.1. The myth: "An agent is just a config in the UI" It's easy to see a Cortex Agent as a few settings in a form: pick a semantic view, add a search service, write some instructions, done. But look at what actually decides whether the agent gives a good answer: Tool descriptions that tell the orchestrator when to use which tool Orchestration and response instructions The semantic view behind Cortex Analyst, including its verified queries The Cortex Search service and the columns it exposes Grants that decide who can call the agent and what it can reach Every one of these is a behavior change. If someone edits a tool description in prod by hand, the agent can start routing questions differently, and you have no diff, no review and no way back. On our platform, warehouses, roles, schemas and grants were already managed through Terraform and Azure DevOps. It didn't make sense for the agent to be the one thing people clicked together. So where does an agent actually live, if not in the UI? 2. Everything the agent needs goes in the repo The good news is that Snowflake gives you a clean way to express an agent as code. C

## Is your Iceberg catalog healthy?

DevFeed: [Is your Iceberg catalog healthy?](<https://devfeed.tech/articles/is-your-iceberg-catalog-healthy-129096.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/is-your-iceberg-catalog-healthy-939dc7817154?source=rss----34b6daafc07---4>)

Author: Daniel Hunt

Published: 2026-11-10T22:38:46Z

Content type: comparison

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Apache Iceberg](<https://devfeed.tech/topics/apache-iceberg.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [vendor lock-in](<https://devfeed.tech/topics/vendor-lock-in.md>), [Data analysis](<https://devfeed.tech/topics/data-analysis.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [apache](<https://devfeed.tech/tags/apache.md>), [apache-iceberg](<https://devfeed.tech/tags/apache-iceberg.md>), [apache-polaris](<https://devfeed.tech/tags/apache-polaris.md>), [architect](<https://devfeed.tech/tags/architect.md>), [blog](<https://devfeed.tech/tags/blog.md>), [catalog](<https://devfeed.tech/tags/catalog.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [completeness](<https://devfeed.tech/tags/completeness.md>), [contributing](<https://devfeed.tech/tags/contributing.md>), [databricks](<https://devfeed.tech/tags/databricks.md>), [development](<https://devfeed.tech/tags/development.md>), [disclosure](<https://devfeed.tech/tags/disclosure.md>), [experience](<https://devfeed.tech/tags/experience.md>), [extension](<https://devfeed.tech/tags/extension.md>), [focus](<https://devfeed.tech/tags/focus.md>), [functional](<https://devfeed.tech/tags/functional.md>), [functionality](<https://devfeed.tech/tags/functionality.md>), [governance](<https://devfeed.tech/tags/governance.md>), [iceberg](<https://devfeed.tech/tags/iceberg.md>), [open-source](<https://devfeed.tech/tags/open-source.md>)

### AI overview

The comparison rates Apache Polaris highest overall (10/10 for project health), with Apache Gravitino close behind (9/10); Lakekeeper scores 5/10, Unity Catalog OSS 3/10, and Nessie 0/10 under the author's criteria. Polaris and Gravitino lead on commit velocity and contributing-organisation counts, though commits outside their lead contributors account for 32% and 15%, respectively. Lakekeeper has eight contributing organisations but remains single-vendor. On capability, Polaris and Lakekeeper implement the full Iceberg REST catalog specification, including multi-table commits, authorization and scoped credential vending. Gravitino supports federation through connectors but lacks multi-table transactions; Unity Catalog OSS lacks named-role RBAC and documented migration from the managed service; Nessie labels REST catalog support experimental. The author presents the results as a reproducible, time-bounded assessment based on public project data and documentation, with unattributed contributors making organisation shares lower bounds.

### Source excerpt

A comparison of five open-source Iceberg catalogs on functionality and OSS project health. Apache Iceberg™ allows an organisation to own their data layer in an open format in their own storage, and has become the de facto open table format. Still, every engine must reach those Iceberg tables through an Iceberg catalog. A natural extension is an open-source Apache Iceberg catalog to complete the open lakehouse stack, acting as an open governance layer, so that access to your tables isn't dependent on any one vendor's managed service. In this blog, I aim to simplify the decision process when it comes to choosing an Iceberg catalog and therefore building an open lakehouse on Iceberg. I focus on the functional completeness of each catalog for Iceberg data and dive deep on OSS project health. After all, an OSS project with one vendor driving development carries similar lock-in risk to a proprietary product. My goal here is to provide transparency on the relative health of common OSS catalogs, specifically Apache Polaris™, Apache Gravitino™, Lakekeeper, Unity Catalog OSS and Project Nessie. The heart of the blog is a collation of public information from each project (which you can review in the public Git repo), but the criteria and analysis are informed by my experience as a specialist architect working with 50+ organisations implementing Apache Iceberg over the past couple of years. This is a companion post to Iceberg Catalogs: Below the Surface (updated November 2026), which scores six vendor-hosted catalogs and shares the same underlying repo. I split the two posts since the first one was becoming a dissertation! Disclosure: This is my personal opinion, not representative of the views of my employer. I'm a specialist architect at Snowflake, one of the vendors contributing towards the Apache Polaris project. Every claim below is sourced and reproducible. Don't take my word for it: test the claims, read the linked project docs, and please comment if you spot any inconsi

## Native Asyncio Comes to the Snowflake Python Connector

DevFeed: [Native Asyncio Comes to the Snowflake Python Connector](<https://devfeed.tech/articles/native-asyncio-comes-to-the-snowflake-python-connector-127038.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/native-asyncio-comes-to-the-snowflake-python-connector-5941385ac80e?source=rss----34b6daafc07---4>)

Author: Patryk Czajka

Published: 2026-11-10T17:24:52Z

Content type: release

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [async](<https://devfeed.tech/topics/async.md>), [await](<https://devfeed.tech/topics/await.md>), [Python](<https://devfeed.tech/topics/python.md>), [FastAPI](<https://devfeed.tech/topics/fastapi.md>)

Tags: [2017](<https://devfeed.tech/tags/2017.md>), [3](<https://devfeed.tech/tags/3.md>), [addition](<https://devfeed.tech/tags/addition.md>), [aio](<https://devfeed.tech/tags/aio.md>), [api](<https://devfeed.tech/tags/api.md>), [applications](<https://devfeed.tech/tags/applications.md>), [arrow](<https://devfeed.tech/tags/arrow.md>), [async](<https://devfeed.tech/tags/async.md>), [asyncio](<https://devfeed.tech/tags/asyncio.md>), [await](<https://devfeed.tech/tags/await.md>), [blocking](<https://devfeed.tech/tags/blocking.md>), [connect](<https://devfeed.tech/tags/connect.md>), [core](<https://devfeed.tech/tags/core.md>), [dashboard](<https://devfeed.tech/tags/dashboard.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-quality](<https://devfeed.tech/tags/data-quality.md>), [event](<https://devfeed.tech/tags/event.md>), [event-loop](<https://devfeed.tech/tags/event-loop.md>), [fan](<https://devfeed.tech/tags/fan.md>), [python](<https://devfeed.tech/tags/python.md>), [rust](<https://devfeed.tech/tags/rust.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

Snowflake Connector for Python 5.0.0 adds `snowflake.connector.aio`, an awaitable API for Python 3.11+ built on the same Rust Universal Core as the existing synchronous connector. Async applications can await connection, query, fetch, transaction, Arrow, and pandas operations, keeping the event loop free while Snowflake work waits on the network; independent queries can run concurrently on separate cursors, though warehouse capacity and queuing still affect total time. The synchronous API remains available for scripts and ordinary functions, and both APIs share the implementation. Version 5.0.0 also adds pandas results for JSON responses; Core Arrow conversion is opt-in and in public preview. The excerpt recommends checking the upgrade guide and comparing results, types, and errors when moving from 4.x.

### Source excerpt

Version 5.0.0 is the same package you already install, now with an awaitable API built on Universal Core. If your Python service runs on an event loop, calling Snowflake has meant handing the call to a thread pool so it doesn't block everything else. In our public preview post we introduced Universal Core; this is the Python feature it made room for. Snowflake Connector for Python 5.0.0 became generally available on October 15. It is still snowflake-connector-python, with the same synchronous API. Check the upgrade guide for behavior changes before switching. The package is a thin Python wrapper around Universal Core, the Rust core shared with the ODBC driver. The big addition is snowflake.connector.aio: native asyncio support, the feature you've asked us for most, going back to a 2017 GitHub issue. TL;DR If you run async Python against Snowflake, install snowflake-connector-python 5.0.0 on Python 3.11 or later. Async applications import snowflake.connector.aio and await the calls that talk to Snowflake: connect, execute, fetch, transactions, Arrow, and pandas. One event loop can now hold many Snowflake queries in flight next to the rest of your I/O, with no thread pool around the connector. Asyncio benefits Most of what an application does with Snowflake is wait. It sends SQL, the warehouse runs it, and the result comes back over HTTPS. Asyncio is built for that: while one coroutine waits on the network, the event loop runs the others. With a native async connector, your application can await Snowflake calls directly, without moving them onto a worker thread. Web services and APIs. Applications built with FastAPI and other async frameworks can now query Snowflake without blocking the server while they wait for results. Snowflake SQLAlchemy supports SQLAlchemy's asyncio extension on this connector, so an AsyncSession can run those queries directly. Fan-out queries. A dashboard tile, a data-quality check, or a metadata crawl often runs a few dozen small, independent

## Who is actually using your Snowflake MCP server?

DevFeed: [Who is actually using your Snowflake MCP server?](<https://devfeed.tech/articles/who-is-actually-using-your-snowflake-mcp-server-124191.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/who-is-actually-using-your-snowflake-mcp-server-683a6cb543d7?source=rss----34b6daafc07---4>)

Author: Ondřej Blaha

Published: 2026-11-09T15:01:04Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [MSP MCP](<https://devfeed.tech/topics/msp-mcp.md>), [JOIN](<https://devfeed.tech/topics/join.md>), [Access Control](<https://devfeed.tech/topics/access-control.md>), [ai observability](<https://devfeed.tech/topics/ai-observability.md>), [Single sign-on (SSO)](<https://devfeed.tech/topics/sso.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>)

Tags: [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [integration](<https://devfeed.tech/tags/integration.md>), [join](<https://devfeed.tech/tags/join.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-server](<https://devfeed.tech/tags/mcp-server.md>), [oauth](<https://devfeed.tech/tags/oauth.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-cortex](<https://devfeed.tech/tags/snowflake-cortex.md>)

### AI overview

Snowflake does not record the MCP server name in query history, but it records the OAuth security integration used at login. Assigning a separate integration to each MCP server and joining QUERY_HISTORY.AUTHN_EVENT_ID to LOGIN_HISTORY.EVENT_ID lets teams attribute queries by integration, user, role, and time; a shared integration only identifies a group. The article explains how to configure OAuth integrations and build usage reports, while noting that login rows are not calls and that query counts can include procedure or dbt work and require filtering out Snowflake's automatic skills-list queries. PAT and key-pair sign-ins lack an integration name, so attribution is weaker. Cortex Analyst leaves tagged validation queries but does not execute returned SQL; Cortex Search activity generally cannot be tied to a user or channel in account usage. Query and login history are delayed, and some queries may be missing or sampled. The reported method is a workaround based on the author's internal October 2026 testing.

### Source excerpt

How to see who sends queries through MCP, how many and how often, using one join in account usageSnowflake has no column that tells you which MCP server a query came through. It does store the name of the OAuth security integration that every MCP client signs in with. Give each MCP server its own integration, join one column from query history to login history, and you have usage per server, user and day. This article shows how, and where the approach stops working. Setting up a Snowflake-managed MCP server takes a few minutes. A few weeks later, the same client usually comes back with a different question: is anyone using it? Which users, how often, and is it worth it? I get this question from clients regularly. The last one put it as a requirement: a model that contains only queries run by their reporting team through one specific MCP server, without the same people's Snowsight work, and with user, role, timestamp and volume. Two identical rows in query history My first thought was that a filter on query history would do it. It won't, because query history has no column that says "this came from MCP server X", and I did not find one anywhere else in account usage either. That client's reporting team works in Snowsight every day and has recently started asking Claude as well. When an analyst writes select region, sum(revenue) from sales in a worksheet and later asks Claude the same question, Claude sends the same SQL through the MCP server. QUERY_HISTORY then holds two rows with the same text, the same user, the same role and, if the warehouse is shared, the same warehouse.The two queries differ only in how the analyst signed in.The sign-in tells them apart An MCP client such as Claude Desktop, Claude Code or Cursor signs in through an OAuth security integration, while Snowsight uses SSO or a password. Snowflake stores the name of the integration with every login, and every query points to the login that opened its session, so all it takes is one join: select q.que

## Missed out on badges from Snowflake Expedition?

DevFeed: [Missed out on badges from Snowflake Expedition?](<https://devfeed.tech/articles/missed-out-on-badges-from-snowflake-expedition-122706.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/missed-out-on-badges-from-snowflake-expedition-9fc8d4292e95?source=rss----34b6daafc07---4>)

Author: Kevin Nguyen

Published: 2026-11-08T18:01:03Z

Content type: tutorial

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Learning](<https://devfeed.tech/topics/learning.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Finance](<https://devfeed.tech/topics/finance.md>)

Tags: [account](<https://devfeed.tech/tags/account.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-features](<https://devfeed.tech/tags/ai-features.md>), [audit](<https://devfeed.tech/tags/audit.md>), [badge](<https://devfeed.tech/tags/badge.md>), [balance](<https://devfeed.tech/tags/balance.md>), [bank](<https://devfeed.tech/tags/bank.md>), [books](<https://devfeed.tech/tags/books.md>), [build](<https://devfeed.tech/tags/build.md>), [campaign](<https://devfeed.tech/tags/campaign.md>), [cowork](<https://devfeed.tech/tags/cowork.md>), [education](<https://devfeed.tech/tags/education.md>), [english](<https://devfeed.tech/tags/english.md>), [exports](<https://devfeed.tech/tags/exports.md>), [finance](<https://devfeed.tech/tags/finance.md>), [free](<https://devfeed.tech/tags/free.md>), [general](<https://devfeed.tech/tags/general.md>), [guide](<https://devfeed.tech/tags/guide.md>), [hands-on-lab](<https://devfeed.tech/tags/hands-on-lab.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [learning](<https://devfeed.tech/tags/learning.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

You can earn either Snowflake Expedition badge by following its on-demand recording and guide, building the lab in a trial account, and passing the auto-grader. The marketing lab uses Snowflake CoCo to identify overlapping campaigns for the same audience, then Snowflake CoWork for plain-English planning questions. The finance lab reconciles payroll, general-ledger, and bank data into one transaction model, then uses CoWork to find imbalances. Sign up through the relevant lab page's regional trial link to enable the required AI features; a regular trial lacks them and requires starting over. Run the unmodified auto-grader in your account, and use the same email for the trial and grader. It checks for missed steps, which you can fix before rerunning it. The article also advertises a limited-time free Snowflake course pass through DataCamp.

### Source excerpt

The Expedition lab recordings are on demand. Build along in a free trial account, pass the auto-grader, and earn a badge to share on social media. Missed Snowflake Expedition, or started a lab and didn't finish? You can still earn your badges. The hands on lab at Snowflake Expedition provided an opportunity to earn a badge to be shared on social media. Register for free to watch the recording, build the lab in your own trial account, and run the auto-grader. If your build passes, you earn the badge, which shows you actually built it, not just watched. Badge 1: Campaign Planning, Simplified (Marketing) Build a campaign planning model with Snowflake CoCo. You take raw campaign exports, find campaigns that target the same audience at the same time, and then use Snowflake CoWork to ask planning questions in plain English. Start here (trial sign-up, auto-grader, badge status): https://northstar.streamlit.app/Expedition_Day2 Watch the recording: https://www.snowflake.com/en/expedition/ams/agenda?agendaPath=session/1991984 Badge 2: Close Faster, Audit Smarter (Finance) Reconcile payroll, the general ledger, and bank statements into one transaction model with Snowflake CoCo. Find the gaps, then use Snowflake CoWork to ask where the books are out of balance. Start here (trial sign-up, auto-grader, badge status): https://northstar.streamlit.app/Expedition_Day3 Watch the recording: https://www.snowflake.com/en/expedition/ams/agenda?agendaPath=session/1991999 How to earn your badge Follow the recording and the guide to build the lab in that trial account. Make sure you sign up for a trial from the lab page above, using the link for your region. The trial links on that page turn on the AI features the lab needs. A regular trial sign-up won't have them, and you'd have to start over. Run the auto-grader. You generate it on the lab page and run it in your account. It checks your build and tells you if you missed a step, so you can fix it and run it again. Don't edit the auto-grader

## Shipping a Data Catalog App on Snowflake App Runtime: Notes From a Real Build

DevFeed: [Shipping a Data Catalog App on Snowflake App Runtime: Notes From a Real Build](<https://devfeed.tech/articles/shipping-a-data-catalog-app-on-snowflake-app-runtime-notes-from-a-real-build-121770.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/shipping-a-data-catalog-app-on-snowflake-app-runtime-notes-from-a-real-build-2fcc74d4e5c0?source=rss----34b6daafc07---4>)

Author: Augusto Rosa

Published: 2026-11-06T22:01:02Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>), [Single sign-on (SSO)](<https://devfeed.tech/topics/sso.md>), [Streamlit](<https://devfeed.tech/topics/streamlit.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>)

Tags: [app](<https://devfeed.tech/tags/app.md>), [application-development](<https://devfeed.tech/tags/application-development.md>), [catalog](<https://devfeed.tech/tags/catalog.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [data-catalog](<https://devfeed.tech/tags/data-catalog.md>), [data-governance](<https://devfeed.tech/tags/data-governance.md>), [data-superhero](<https://devfeed.tech/tags/data-superhero.md>), [dbt](<https://devfeed.tech/tags/dbt.md>), [definition](<https://devfeed.tech/tags/definition.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [next](<https://devfeed.tech/tags/next.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [shipping](<https://devfeed.tech/tags/shipping.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

A five-day rebuild moved a Streamlit business glossary and approval app to a Next.js app on Snowflake App Runtime, deployed with `snow app deploy`. The account keeps data in Snowflake and uses existing SSO; identity-aware writes live in the app because Sigma connects through a shared service role. The main constraint is access design: caller's-rights queries require both user privileges and matching CALLER grants held by the app owner, while shared SNOWFLAKE objects such as Cortex functions and GET_LINEAGE require owner's rights. The deploying role becomes the app owner and sets the privilege ceiling, so the article recommends a dedicated, narrowly granted role and setting up grants before building pages. It also describes limits and tradeoffs: lineage shows only metadata visible to the owner, dynamic tables need MONITOR rather than REFERENCES, and steward workflows require real users for acceptance testing. The build sequence provisions Snowflake objects and grants, seeds draft terms from dbt without overwriting human edits, then deploys and grants app usage.

### Source excerpt

Caller's rights, CALLER grants, and the shared objects that refuse to cooperatePhoto by Boitumelo on Unsplash Most data platforms hit a governance problem right when they start working. The dbt models ship, the semantic views exist, the dashboards render. Then someone in finance asks what Net Revenue means in this warehouse and who signed off on it, and nobody has a good answer. Plus, wow, Snowflake App Runtime is pretty amazing: an app very close to your data. tl;dr Over five days I rebuilt a business glossary and definition-approval app on Snowflake App Runtime and deployed it into a client account. It is a Next.js app. It went live with snow app deploy, and nobody had to touch Docker, a container registry, or a separate auth layer. The data stays in the account, and users sign in with the SSO they already have. To be fair, the catalog companion was not my idea. A co-worker built the first version in Streamlit, and I extended it and moved it to App Runtime. The UI was the quick part. Three things that took me the most time to figure out: Caller's rights is two permission checks. A query runs as the signed-in user only if the user's own role allows it and the app's owner role holds a matching CALLER grant. Miss the CALLER grant and the page fails with a "does not exist or not authorized" error that looks like a missing object. Shared objects in the SNOWFLAKE database cannot take CALLER grants. Cortex functions and GET_LINEAGE come back as Unknown user-defined function under caller's rights. Those calls run with owner's rights, and you scope what the owner can see. The owner role sets the ceiling for every user. The role that deploys the app becomes its owner. Deploy as a dedicated role with a short list of grants, and never as SYSADMIN or ACCOUNTADMIN. Get the grants in place first, then write the pages. In that order the build is boring, which is what you want. What I built and why it runs on App Runtime The client is a retailer partway through moving off a legacy

## Query Databricks Lakeflow Pipeline Tables from Snowflake, Without Copying Data

DevFeed: [Query Databricks Lakeflow Pipeline Tables from Snowflake, Without Copying Data](<https://devfeed.tech/articles/query-databricks-lakeflow-pipeline-tables-from-snowflake-without-copying-data-121423.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/query-databricks-lakeflow-pipeline-tables-from-snowflake-without-copying-data-dde301ede202?source=rss----34b6daafc07---4>)

Author: Vikneshwara Balakrishnan

Published: 2026-11-06T15:01:02Z

Content type: tutorial

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Apache Iceberg](<https://devfeed.tech/topics/apache-iceberg.md>), [databricks](<https://devfeed.tech/topics/databricks.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [OpenID connect (OIDC)](<https://devfeed.tech/topics/oidc.md>)

Tags: [apache-iceberg](<https://devfeed.tech/tags/apache-iceberg.md>), [apache-spark](<https://devfeed.tech/tags/apache-spark.md>), [aws](<https://devfeed.tech/tags/aws.md>), [data](<https://devfeed.tech/tags/data.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [databricks](<https://devfeed.tech/tags/databricks.md>), [lakeflow](<https://devfeed.tech/tags/lakeflow.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

The excerpt reports that all 13 selected Lakeflow pipeline tables were read in place from Snowflake on AWS after enabling `pipelines.externalMetadata.enabled`, Iceberg v3 compatibility, and Unity Catalog external access. Individual Iceberg tables served raw and curated data: they allow a precise table list and refresh only when the pipeline event log indicates changes, but require explicit refreshes. A catalog-linked database served the gold schema with automatic discovery and refresh; it is read-only here, exposes whole namespaces, and can propagate drops to Databricks if writes are allowed. The setup requires OAuth catalog integrations, vended credentials, and grants for the Snowflake service principal. Reported pitfalls include disabled metastore external access, missing pipeline metadata opt-in, and Iceberg UUID mismatches after full refreshes; the author handled the latter by recreating affected Snowflake Iceberg objects. The author describes pipeline-table external access as Public Preview and reports limitations including unsupported external writes, row-level security, and masking. Results come from one environment, and some explanations and fixes are explicitly unverified.

### Source excerpt

Streaming tables and materialized views, read in place as Iceberg: the settings that matter, what the logs show, and the Snowflake errors I hit Most guides to reading Databricks from Snowflake start with a plain Delta table. Mine are different: every table here is built by Lakeflow Declarative Pipelines (formerly Delta Live Tables), so they are streaming tables and materialized views, which an outside engine cannot read out of the box. I got Snowflake to read all 13 of them in place, as Apache Iceberg, with no copy: three layers on AWS, each in its own Unity Catalog catalog. I tried two Snowflake features on them: Iceberg tables over Unity Catalog's REST endpoint, and a catalog-linked database. This article covers how each works on Lakeflow tables, the Snowflake checks that confirm it, and the four Snowflake errors that cost me time. In short: Lakeflow tables need one opt-in (pipelines.externalMetadata.enabled) to be readable from Snowflake; with it, both Iceberg tables over the REST endpoint (raw, curated) and a catalog-linked database (gold) read all 13 in place. Disclaimer: This is my own work, not affiliated with or endorsed by Snowflake, Databricks or AWS. Parts were established by trial and error in one environment, so verify against the official documentation.Why Lakeflow tables are different A Lakeflow pipeline publishes streaming tables (incremental) and materialized views (recomputed queries). Both are Unity Catalog managed tables, but neither is a plain Delta table you can point another engine at. Three things follow, and they shaped everything below. They need an opt-in. Per Databricks' documentation, they appear externally "as managed tables" with the same name, but only with pipelines.externalMetadata.enabled = 'true', UniForm for Iceberg and Iceberg v3 (Databricks Runtime 17.3 or later). Skip the property and Snowflake rejects the views (Error 2). The pipeline owns them. Its run-as identity owns every table, so everyone else, including Snowflake's ser

## Using Life-Science Ontologies from a Snowflake AI Agent: Gene Ontology and the OLS4 MCP

DevFeed: [Using Life-Science Ontologies from a Snowflake AI Agent: Gene Ontology and the OLS4 MCP](<https://devfeed.tech/articles/using-life-science-ontologies-from-a-snowflake-ai-agent-gene-ontology-and-the-ols4-mcp-119798.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/using-life-science-ontologies-from-a-snowflake-ai-agent-gene-ontology-and-the-ols4-mcp-c04fd5edc4ab?source=rss----34b6daafc07---4>)

Author: Tatsuya Koreeda

Published: 2026-11-05T22:01:01Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [The Gene Ontology Consortium](<https://devfeed.tech/topics/geneontology.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-sdk](<https://devfeed.tech/tags/agent-sdk.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [apis](<https://devfeed.tech/tags/apis.md>), [context-layer](<https://devfeed.tech/tags/context-layer.md>), [data](<https://devfeed.tech/tags/data.md>), [enrichment](<https://devfeed.tech/tags/enrichment.md>), [enterprise-data](<https://devfeed.tech/tags/enterprise-data.md>), [healthcare](<https://devfeed.tech/tags/healthcare.md>), [life-sciences](<https://devfeed.tech/tags/life-sciences.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-server](<https://devfeed.tech/tags/mcp-server.md>), [ontologies](<https://devfeed.tech/tags/ontologies.md>), [ontology](<https://devfeed.tech/tags/ontology.md>), [palantir](<https://devfeed.tech/tags/palantir.md>), [rna](<https://devfeed.tech/tags/rna.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

The article demonstrates a Snowflake workflow that computes Gene Ontology enrichment for dexamethasone-responsive genes and uses OLS4 to supply term definitions and hierarchies for an AI agent's interpretation. In the airway RNA-seq example, DESeq2 identifies 999 significant genes (519 up, 480 down); a Snowflake Python stored procedure calculates enrichment, matching clusterProfiler across 4,540 terms under the stated settings. OLS4's unauthenticated MCP server can be queried directly, but the Cortex Agent MCP connector requires OAuth, so the workflow instead uses the Cortex Code Agent SDK or a custom tool calling OLS4's REST API through an external access integration. Looking up hierarchies helps the agent recognize related enriched terms that reflect the same 28 genes. The agent is limited to Snowflake SQL and OLS4 tools, with non-tool knowledge marked separately as "Outside knowledge."

### Source excerpt

Introduction In enterprise data work, the word "ontology" comes up more and more. The idea is to give an AI agent an ontology as the context layer it refers to when it works with company data. Life science has had ontologies such as the Gene Ontology (GO) for more than 20 years, and anyone can query them through public APIs. EMBL-EBI's Ontology Lookup Service (OLS4) also offers an MCP server, so an AI agent can look up GO definitions and hierarchies directly. In this article, I run a GO enrichment in Snowflake on the genes that changed in an RNA-seq experiment (differentially expressed genes, DEGs), and then have an AI agent interpret the result while it looks terms up in OLS4. Life-science ontologies In the data-platform world, the usual example is the Palantir Ontology. Its documentation calls it "an operational layer for the organization." It holds the organization's objects and links, plus actions and functions. It is an ontology for modelling and running your own business. Healthcare and life-science ontologies such as GO, the Human Phenotype Ontology (HPO), Mondo and ChEBI are different. They are reference ontologies that define scientific concepts independent of any organization. They give shared identifiers and hierarchies to gene functions, phenotypes, diseases and chemicals, and researchers use them as the standard when they annotate data. Purpose: Palantir Ontology represents and runs the business; GO / HPO / Mondo standardize scientific concepts. Core: Palantir has objects, links and actions; GO / HPO / Mondo have classes, relations and definitions. Relation to data: Palantir holds the actual data; GO / HPO / Mondo are the standard used to annotate data. The timeline below shows how life-science ontologies and the infrastructure for distributing and querying them developed. How life-science ontologies and their infrastructure developed Of these, GO, which appeared in 2000, is widely used in enrichment analysis downstream of RNA-seq. Life-science research

## Choosing How ML Models Move to Production on Snowflake: Code, Model, or Hybrid Promotion

DevFeed: [Choosing How ML Models Move to Production on Snowflake: Code, Model, or Hybrid Promotion](<https://devfeed.tech/articles/choosing-how-ml-models-move-to-production-on-snowflake-code-model-or-hybrid-promotion-119796.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/choosing-how-ml-models-move-to-production-on-snowflake-code-model-or-hybrid-promotion-aca2ae322bb6?source=rss----34b6daafc07---4>)

Author: Luis Villavicencio

Published: 2026-11-05T20:01:01Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [ML model versioning](<https://devfeed.tech/topics/ml-model-versioning.md>), [Deployment-Driven Development](<https://devfeed.tech/topics/deployment-driven-development.md>), [staging](<https://devfeed.tech/topics/staging.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ci-cd](<https://devfeed.tech/tags/ci-cd.md>), [code](<https://devfeed.tech/tags/code.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data](<https://devfeed.tech/tags/data.md>), [data-ingestion](<https://devfeed.tech/tags/data-ingestion.md>), [dev](<https://devfeed.tech/tags/dev.md>), [development](<https://devfeed.tech/tags/development.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [environment](<https://devfeed.tech/tags/environment.md>), [environments](<https://devfeed.tech/tags/environments.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-development](<https://devfeed.tech/tags/ml-development.md>), [mlops](<https://devfeed.tech/tags/mlops.md>), [model](<https://devfeed.tech/tags/model.md>), [model-registry](<https://devfeed.tech/tags/model-registry.md>), [production](<https://devfeed.tech/tags/production.md>), [promotion](<https://devfeed.tech/tags/promotion.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

The promotion pattern determines whether code, a trained model, or both move through Snowflake environments. Code Promotion deploys training code to each environment and retrains there, so PROD uses production data but needs production compute. Model Promotion trains once in DEV and copies the artifact through each environment's Model Registry; this suits expensive or infrequent training but depends on DEV data being representative. Hybrid trains in STAGING, which must access production data, then promotes the artifact to PROD without training there. Choose based on data access, training cost and frequency, CI/CD readiness, and automation needs, and validate the choice with data and platform stakeholders. Teams can begin manually; from semi-automation onward, CI/CD moves objects between environments. The article also recommends least-privilege promotion roles and metric-gated versions with the prior version archived for rollback.

### Source excerpt

Data access, training cost, CI/CD readiness, and business needs decide how your models reach production -- and shape your environments, your pipelines, and your risk. Moving a model to production Once a model works in development, it has to reach production. What moves between environments, where training happens, and which data training needs differ by promotion pattern -- and so do the risks. For example, when a model is trained in DEV and promoted as-is, models trained on dev data may not fully represent production distributions. There is no single correct answer; the right pattern depends on business needs and must be validated with the stakeholders. This article walks through the options. Key concepts first Before comparing options, a few terms used throughout: Environment -- an isolated set of objects for one stage of the lifecycle. Let's assume DEV, STAGING, and PROD. On Snowflake, an environment is generally a separate database in one account, or an entirely separate account. Promotion -- moving work from one environment to the next, e.g., DEV -> STAGING -> PROD. Model artifact -- the trained model itself, stored and versioned in the Model Registry. Model code -- the pipeline that produces the model: data ingestion, feature engineering, training, evaluation, and registration. Promotion pattern -- the answer to what moves between environments: the code, the artifact, or both. There are three promotion patterns: Code Promotion -- ship the code. The training code moves from DEV to STAGING to PROD, and the model is retrained in each environment. The production model is trained on production data, in production. No model is copied between environments. Model Promotion -- ship the model. The model is trained once, in DEV, and the trained artifact is promoted to STAGING and PROD. It trains only once, which suits expensive or infrequent training -- but the model only knows the data it was trained on in the lower environment (DEV). Hybrid Promotion -- ship the code, then the mode

## Best Practices for Scoping and Organising Snowflake Semantic Views

DevFeed: [Best Practices for Scoping and Organising Snowflake Semantic Views](<https://devfeed.tech/articles/best-practices-for-scoping-and-organising-snowflake-semantic-views-119795.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/best-practices-for-scoping-and-organising-snowflake-semantic-views-6e2e2758b538?source=rss----34b6daafc07---4>)

Author: Nithya Rangarajan

Published: 2026-11-05T15:01:08Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [semantic-layer](<https://devfeed.tech/topics/semantic-layer.md>), [Data Management](<https://devfeed.tech/topics/data-management.md>), [Data Quality](<https://devfeed.tech/topics/data-quality.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [best-practices](<https://devfeed.tech/tags/best-practices.md>), [composition](<https://devfeed.tech/tags/composition.md>), [context](<https://devfeed.tech/tags/context.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [governance](<https://devfeed.tech/tags/governance.md>), [name](<https://devfeed.tech/tags/name.md>), [ownership](<https://devfeed.tech/tags/ownership.md>), [semantic-layer](<https://devfeed.tech/tags/semantic-layer.md>), [semantics](<https://devfeed.tech/tags/semantics.md>), [size](<https://devfeed.tech/tags/size.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-semantic-views](<https://devfeed.tech/tags/snowflake-semantic-views.md>), [sql](<https://devfeed.tech/tags/sql.md>), [topology](<https://devfeed.tech/tags/topology.md>)

### AI overview

The evaluation found that semantic view topology and size did not change answer accuracy in the tested setup, while duplicate dimension names caused a hard-to-spot wrong answer. The article recommends maintaining domain views under team ownership and composing them for use cases; tested topologies were effectively equivalent, so governance and shareability should guide the choice. Accuracy stayed stable up to 118K tokens across 104 tables and 20 questions, though the test used only 1-3 joins and distinct names in added tables. Prefix entities and relationships by domain, and give ambiguous dimensions unique business-role names with clear descriptions. For access, direct semantic view queries need SELECT on the view, while agents in default mode generate SQL against physical tables and require table grants. Explicit strict SQL generation routes through the view under its owner's rights, avoiding those grants but limiting calculations to defined metrics, dimensions, and facts. Strict mode must be enabled; it is private preview at the time described.

### Source excerpt

Snowflake Semantic Views provide a governed, reusable layer that defines how tables relate, what metrics mean, and who can access what. This article describes a set of best practices and lessons learned for scoping, naming, sizing, and securing semantic views in an enterprise setting. Each recommendation is grounded in evaluation data from a 20-question test suite run across multiple architectures on the Tasty Bytes dataset. At the time of writing Composable semantic views and strict mode are still in private preview. Introduction Does one big semantic view hurt accuracy? Should you organise by team or by use case? I tested both. Neither topology nor size changed the answers, but a single duplicated column name did. Semantic views are quickly becoming the standard way to expose governed data to Cortex Agents, Snowflake CoWork, BI tools, and downstream SQL consumers. Over the past several months I have worked with multiple customers on their introduction of semantic views as part of their AI and analytics strategy. Every organisation's data architecture, team structure and governance requirements are different, and those differences shape how semantic views should be scoped. A semantic view is also highly visible: it defines the metrics and dimensions that business users, analysts and AI agents all rely on. Decisions about what goes into it, who owns it, and how it is secured have broad consequences. This article covers: ownership and composition patterns semantic view sizing naming conventions for disambiguation access control, including the new strict SQL generation mode Each recommendation comes with results from controlled experiments. Ownership and Composition The most common scoping question enterprise architects face is: should we organise semantic views around who owns the data, or around how it gets queried? The answer is both, at different layers. Ownership and Scope A two-layer model separates the concerns cleanly: Base layer -- ownership. Each team maintai

## Cortex Agent Versioning and Multi-Environment Deployment

DevFeed: [Cortex Agent Versioning and Multi-Environment Deployment](<https://devfeed.tech/articles/cortex-agent-versioning-and-multi-environment-deployment-119000.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/cortex-agent-versioning-and-multi-environment-deployment-18cb43e8097e?source=rss----34b6daafc07---4>)

Author: Jayashree Suresh

Published: 2026-11-04T18:28:41Z

Content type: tutorial

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Deployment Strategies](<https://devfeed.tech/topics/deployment-strategies.md>), [release engineering](<https://devfeed.tech/topics/release-engineering.md>), [versioning](<https://devfeed.tech/topics/versioning.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [cortex](<https://devfeed.tech/tags/cortex.md>), [cortex-agent](<https://devfeed.tech/tags/cortex-agent.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [environment](<https://devfeed.tech/tags/environment.md>), [multi-environment](<https://devfeed.tech/tags/multi-environment.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowflake-cortex](<https://devfeed.tech/tags/snowflake-cortex.md>), [versioning](<https://devfeed.tech/tags/versioning.md>)

### AI overview

For Snowflake Cortex Agents, use one agent when all environments share the same tools and data; when each environment has separate databases and services, separate agents keep version history, drafts, defaults, and access independent. The proposed Dev -> Test -> Prod pipeline detects data and agent changes, validates specifications, builds changed dbt objects, creates a committed agent version when its specification changes, evaluates it, and promotes it only after checks and approvals pass. Data-only changes are evaluated after the rebuilt data is already live, so evaluation can detect a problem but cannot gate that change. For comparable evaluations, use committed versions, pin metric versions, and calibrate thresholds across repeated runs. Rollback moves the alias and default version to the prior agent version, restoring its configuration but not the data or tools it uses. Recreated objects may need COPY GRANTS to retain explicit privileges; future grants are not copied. These are the author's recommendations, not an official Snowflake reference architecture.

### Source excerpt

Introduction Organisations are moving rapidly toward agentic workflows, enabling business users to query data in natural language without requiring deep technical expertise. As these agents are now directly involved in critical business decisions, updating them requires strict engineering rigour. Managing agent versions, automated evaluation, structured approvals, and rapid rollbacks are non-negotiable. This article explores a practical approach to managing Snowflake Cortex Agent versioning and multi-environment deployment. Most data teams already have a release process for their data. This article uses one such setup as an example: dbt transforms and loads the silver and gold tables, and builds the views and semantic views on top of them. These dbt models are built in a development environment, validated in test and promoted to production, often with a Git branch for each environment. Cortex Agents add one more layer to this process. An agent has its own configuration, made up of instructions, models, and tools, and this configuration also needs to move from one environment to the next. Before building the pipeline, it is critical to decide whether a single agent should serve all environments, or whether each environment should have its own agent. One agent for all environments, or one agent per environment? Snowflake's versioning model is built around a single agent. Changes are made in the live version, committed, tested through an alias and, once approved, set as the default version for CoWork users. Rolling back means moving the alias and the default version back to the previous version. When a single agent is a good fit A single agent fits when the versions at each stage use the same tools and data. Changes are then limited to the agent configuration, such as instructions, sample questions or model, and promoting a version releases exactly what was tested. Consider a FinOps assistant built by the platform team. It answers questions such as "Which warehouses us

## Enterprise ML Architecture: Making Every Boundary Earn Its Place

DevFeed: [Enterprise ML Architecture: Making Every Boundary Earn Its Place](<https://devfeed.tech/articles/enterprise-ml-architecture-making-every-boundary-earn-its-place-119001.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/enterprise-ml-architecture-making-every-boundary-earn-its-place-c0ffaac36fd4?source=rss----34b6daafc07---4>)

Author: Anupama Gangadhar

Published: 2026-11-04T18:21:34Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Inference](<https://devfeed.tech/topics/inference.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [data-governance](<https://devfeed.tech/topics/data-governance.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [observability](<https://devfeed.tech/topics/observability.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [also](<https://devfeed.tech/tags/also.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [automatic](<https://devfeed.tech/tags/automatic.md>), [based](<https://devfeed.tech/tags/based.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [campaign](<https://devfeed.tech/tags/campaign.md>), [can](<https://devfeed.tech/tags/can.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [churn](<https://devfeed.tech/tags/churn.md>), [compute](<https://devfeed.tech/tags/compute.md>), [consider](<https://devfeed.tech/tags/consider.md>), [container](<https://devfeed.tech/tags/container.md>), [cost](<https://devfeed.tech/tags/cost.md>), [created](<https://devfeed.tech/tags/created.md>), [customer](<https://devfeed.tech/tags/customer.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [enterprise-architecture](<https://devfeed.tech/tags/enterprise-architecture.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [making](<https://devfeed.tech/tags/making.md>), [ml](<https://devfeed.tech/tags/ml.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

ML workload placement should follow measured boundary costs and operating needs, not a platform's feature list. The excerpt proposes seven tests: assess data movement; compare total cost, including operations and change; distinguish batch, online and specialised workloads; place reusable features; decide training, governance, execution and monitoring separately; follow where predictions are consumed; and check whether agentic workflows change the workload requirements. Snowflake may suit batch scoring near governed data, reusable features and Snowflake based consumption. Mature low latency endpoints or workloads needing specialised runtimes may remain elsewhere. Hybrid designs can combine Snowflake features and governance with external training or online serving, then return predictions and outcomes for observation. Choose to keep, integrate, consolidate or build new based on workload evidence such as volume, latency, utilisation, recovery, controls and cost; recommendations should remain conditional when evidence is incomplete.

### Source excerpt

Seven Workload Tests for Better Architecture Decisions Enterprise ML architectures carry history. Many were built when the data platform handled storage, transformation and analytics, while a separate ML platform supplied notebooks, training infrastructure, registries, deployment pipelines, online endpoints, monitoring and specialist compute. Those choices created skills, controls and production services that may have improved over years. Snowflake now provides capabilities across more stages of the ML lifecycle. These capabilities now include feature management, development and job execution, model governance, warehouse and container-based inference, observability, explainability and lineage. The platform can also work with models trained elsewhere and with inference performed outside Snowflake. These broader capabilities merit reassessment, not automatic migration. A platform gaining a capability is not, by itself, an architecture justification for moving a workload. Placement should follow workload evidence: data movement, execution pattern, latency, feature reuse, operating maturity, prediction consumption, governance and total cost. Here is a set of seven tests to structure that discussion and decision making. Test 1: Does data movement materially affect the workload? "The data is already in Snowflake" is not a sufficient reason to run a machine learning workload there. Locality matters when the volume, frequency and operational consequences of crossing the platform boundary are substantial. As an example, consider a churn model that scores 50 million customer records each night. Source data and features reside in Snowflake, and the probabilities return there for campaign segmentation. External scoring creates a round trip; extract a large feature population, transfer it to another service, score it, move the predictions back, reconcile the result and recover from failures across two platforms. Running inference closer to the governed data may remove meaningful

[Next page](<https://devfeed.tech/sources/snowflake-builders.md?cursor=WyIyMDI2LTExLTA0VDE4OjIxOjM0KzAwOjAwIiwgImE3ODllZTQzLWJlOGEtNDg5Mi05MzlkLTE3NGExMDRhOGJlYyJd>)