Snowflake Data Cloud

Openflow · Snowpipe · Notebooks · Cortex AI

10 GB/s

Streaming throughput

5 sec

Ingest to query

7 GA

AI SQL functions

Platform overview

One platform across data, code, and AI

Data ingestion

Openflow and Snowpipe unify managed batch, streaming, and connector-based ingestion inside Snowflake.

Jupyter notebooks

Snowflake Notebooks bring native notebook development, SQL, Python, and governed ML directly to the platform.

Cortex AI

Agents, Search, Analyst, and AI functions deliver LLM, RAG, and analytics workflows without moving data out.

Unified governance, RBAC, lineage, and security policies apply across every layer.

Managed integration

Snowflake Openflow

GA September 2025 · Built on Apache NiFi · Managed natively inside Snowflake

What it does

Openflow provides managed connectors, routing, transformation, and orchestration for structured and unstructured data across databases, SaaS apps, object stores, event streams, and documents.

How teams use it

Teams define ingestion with SQL-style pipeline management, then combine it with schema evolution, monitoring, and Cortex enrichment to land analytics-ready data faster.

Batch + streaming

One managed fabric for both periodic and event-driven ingestion.

Schema evolution

Adapts to upstream change without heavy manual remapping.

AI-ready prep

Use Cortex functions in pipelines to classify, extract, summarize, and enrich content in flight.

External sources

Websites, PDFs, CSVs, and Swagger APIs

Website and HTTP ingestion

Openflow supports HTTP-based collection patterns, while Snowpark Python with network rules and external access integrations can call web endpoints, download HTML or JSON, and stage the result directly in Snowflake.

PDF and document extraction

Land documents in internal or external stages, then use AI_EXTRACT() to pull structured fields, tables, and text from forms, contracts, invoices, and scanned content.

CSV and file-based loads

Use Snowsight uploads, internal stages, or S3, ADLS, and GCS stages with COPY INTO and Snowpipe auto-ingest for CSV, JSON, Parquet, ORC, Avro, XML, and related formats.

Swagger / OpenAPI-driven APIs

For REST sources described by Swagger or OpenAPI, teams typically create Snowpark functions or procedures that authenticate, call endpoints, normalize JSON payloads, and insert results directly into Snowflake tables.

Typical pattern: source fetch → stage landing zone → COPY INTO / Snowpipe / Openflow → Cortex enrichment.

Security and governance

Authentication and Role-Based Access Control

Authentication options

Username/password + MFA — TOTP-based multi-factor (Duo, authenticator apps) enforced at the account or user level.
SSO / SAML 2.0 — Federate identity through Okta, Azure AD, OneLogin, or any SAML-compliant IdP. Snowflake acts as the service provider.
OAuth 2.0 — Delegate token-based access for BI tools (Tableau, Power BI), partner applications, and programmatic clients without sharing credentials.
Key-pair authentication — RSA 2048/4096-bit key pairs for service accounts, CI/CD pipelines, and automated ingestion connectors like Openflow and Snowpark.
Private Link — Route all traffic over AWS PrivateLink, Azure Private Endpoint, or GCP Private Service Connect — no public internet required.

Role-based access control (RBAC)

Discretionary ACL — Object owners grant privileges (SELECT, INSERT, USAGE) to roles. Roles are then granted to users or other roles.
Role hierarchy — Roles inherit permissions from lower roles, enabling clean separation of SYSADMIN, SECURITYADMIN, and custom functional roles.
Row-level security — Row Access Policies restrict which rows specific roles can see, enforced transparently at query time — no app-layer changes required.
Column masking — Dynamic Data Masking policies redact or hash sensitive columns (PII, PCI, PHI) based on role, without altering stored data.
Object tags and classification — Tag sensitive objects system-wide and apply governance policies uniformly at the classification level.

Network policies

Allowlist IP ranges and block all other connections at the account or user level.

Tri-Secret Secure

Customer-managed key (Snowflake + your KMS) so Snowflake cannot decrypt your data unilaterally.

Access history

Every query, object access, and data movement is logged and queryable via the ACCESS_HISTORY view.

Data science and ML

Snowflake Notebooks

Native notebooks in Snowflake Workspaces with governed SQL, Python, package management, and container-backed compute.

Jupyter-native workflow

Notebook cells support familiar SQL and Python authoring with direct access to Snowflake data and packages.

CPU and GPU execution

Container runtime enables governed ML development, training, and experimentation without exporting data elsewhere.

SQL + Python together

Analysts and engineers can move from query to model to chart inside one notebook artifact.

Governed by Snowflake

RBAC, policies, and data controls stay intact because notebook execution happens in-platform.

Operationalization

Pair notebooks with Tasks, pipelines, model development, and Git-based workflows for production delivery.

Built-in exploration

Use charts, dataframes, and notebook-native visuals to validate ingestion, feature engineering, and AI outputs.

Pipeline path: ingest → transform in notebooks → train and evaluate → deploy with Cortex and Snowpark.

AI services

Snowflake Cortex AI

Cortex Agents

Agentic workflows that combine tools, structured data, unstructured context, and LLM reasoning.

Cortex Analyst

Natural-language-to-SQL analytics through semantic models for governed business questions.

Cortex Search

Hybrid search and RAG retrieval over documents, tables, and embeddings with semantic ranking.

AI SQL functions

AI_CLASSIFYAI_EXTRACT AI_EMBEDAI_SENTIMENT AI_TRANSCRIBEAI_TRANSLATE AI_SIMILARITY

Model access and guardrails

Support spans Snowflake-hosted and partner LLM options for generation, embeddings, and reasoning.

Security controls keep prompts, policies, and governed data flows inside the Snowflake environment.

Flagship AI experience

Snowflake Intelligence

A conversational layer that unifies Analyst, Search, Agents, and governed enterprise context.

Natural-language analytics

Users ask business questions in plain language and receive structured answers backed by governed semantic models.

RAG across enterprise content

Documents, PDFs, notes, and structured facts can be retrieved and synthesized in a single response path.

Actionable workflows

Agents can trigger tools, procedures, and orchestrated downstream actions instead of only generating text.

Model abstraction

Teams focus on outcomes while Snowflake manages model access patterns, routing, and governance layers.

Inline charting and explanation

Answers can include charts, summaries, and analytic framing generated from live Snowflake results.

Threaded context

Follow-up questions preserve context, making iterative analysis more natural for business and technical users.

Visualization and analytics

Dashboard Creation

Snowsight native dashboards

Tile-based layout — Dashboards are flexible collections of tiles, each backed by a SQL query. Create from scratch or promote any worksheet to a dashboard in one step.
Chart types — Bar, line, scatter, heat grid, and scorecard visualizations, all driven by query results and fully customizable in the UI.
Dynamic filters — Add cross-tile filters, date range pickers, and custom parameters to make dashboards interactive without writing filter logic into every query.
Real-time collaboration — Share dashboards with other Snowsight users via role-based access. Multiple users can view and interact simultaneously.
Role and warehouse context — Choose the execution role and compute warehouse per dashboard, controlling both data access and query cost.

AI-assisted and embedded options

Cortex Intelligence charts — Ask a natural-language question in Snowflake Intelligence and get an auto-generated Vega-Lite chart over live data — no dashboard setup required.
Notebook-based dashboards — Build rich, code-driven dashboards with Plotly, Altair, Streamlit, or Matplotlib inside Snowflake Notebooks, then share the notebook as the delivery artifact.
Streamlit in Snowflake — Deploy Python Streamlit apps directly in Snowflake Workspaces — interactive dashboard apps with custom UI components, no separate host needed.
BI tool integration — Tableau, Power BI, Looker, and Sigma connect natively via Snowflake's JDBC/ODBC drivers and partner connectors, using Snowflake as the live query engine.
Governance inherited — All dashboards — Snowsight, Notebooks, and BI tools — respect Snowflake RBAC, row-level policies, and column masking automatically.

Note: Legacy Dashboards are being removed from Snowflake on June 22, 2026. Migrate existing legacy dashboards to Snowsight Dashboards or Notebooks before that date.

️

One Platform. End to End.

Snowflake unifies data ingestion, development, AI, and visualization — all governed, all in one place.

Ingest

Openflow brings any data in — structured, unstructured, batch or streaming

Develop

Notebooks and Streamlit apps run natively on your data — no exports needed

AI

Cortex Agents, Analyst, Search, and SQL AI functions — all inside the platform

Visualize

Snowsight dashboards or connect Tableau, Power BI, Sigma, and more

Governed · Secure · Scalable