# Snowflake Agentic AI Platform - Architecture Diagrams ## 1. High-Level Layered Architecture ```mermaid graph TB subgraph "Layer 5: User Experience" UI1[Streamlit UI
7 Interactive Pages] UI2[AI Assistant
Natural Language] UI3[REST API
OpenAPI 3.0] UI4[MCP Interface
External LLM Apps] end subgraph "Layer 4: Orchestration & Communication" ORCH[Master Orchestrator
SP_ORCHESTRATE_ONBOARDING] STATE[Workflow State Machine
WORKFLOW_EXECUTIONS] LOGS[Execution Logging
AGENT_EXECUTION_LOG] end subgraph "Layer 3: Agent Execution" A1[Agent 1
Data Profiling
Schema + PII Detection] A2[Agent 2
Data Dictionary
DDL Generation] A4[Agent 4
Data Mapping
Transformation Logic] end subgraph "Layer 2: Intelligence" CORTEX[Snowflake Cortex AI] AI1[AI_COMPLETE
LLM Reasoning] AI2[AI_CLASSIFY
PII Detection] AI3[AI_EXTRACT
Entity Extraction] AI4[AI_FILTER
Conditions] AI5[AI_AGG
Aggregations] CORTEX --> AI1 CORTEX --> AI2 CORTEX --> AI3 CORTEX --> AI4 CORTEX --> AI5 end subgraph "Layer 1: Foundation" SF[Snowflake Data Platform] COMPUTE[Snowpark Compute
3 Warehouses] STORAGE[Stages & Tables
17 Core Tables] GOV[Governance
RBAC + Tagging + Masking] SF --> COMPUTE SF --> STORAGE SF --> GOV end subgraph "Non-AI Features" INFER[INFER_SCHEMA
Schema Detection] INFO[Information Schema
Metadata Queries] TAG[Object Tagging
Classification] MASK[Dynamic Masking
PII Protection] end UI1 --> ORCH UI2 --> ORCH UI3 --> ORCH UI4 --> ORCH ORCH --> A1 ORCH --> A2 ORCH --> A4 ORCH --> STATE ORCH --> LOGS A1 --> CORTEX A1 --> INFER A2 --> CORTEX A2 --> INFO A4 --> CORTEX A4 --> INFO A1 --> STORAGE A2 --> STORAGE A4 --> STORAGE STORAGE --> COMPUTE GOV --> TAG GOV --> MASK style UI1 fill:#29B5E8,color:#fff style UI2 fill:#29B5E8,color:#fff style UI3 fill:#29B5E8,color:#fff style UI4 fill:#29B5E8,color:#fff style ORCH fill:#00A3E0,color:#fff style A1 fill:#0073E6,color:#fff style A2 fill:#0073E6,color:#fff style A4 fill:#0073E6,color:#fff style CORTEX fill:#FF6B35,color:#fff style SF fill:#333,color:#fff ``` --- ## 2. Data Flow Architecture ```mermaid flowchart LR subgraph Input USER[User/System] FILE[Data Files
CSV/JSON/Parquet] end subgraph "Ingestion" STAGE[RAW_DATA_STAGE
Internal Stage] end subgraph "Agent 1: Profiling" SAMPLE[Sample Data
10K rows] INFER[INFER_SCHEMA
Auto-detect] STATS[Calculate Stats
Nulls/Cardinality] PII[AI_CLASSIFY
Detect PII/PHI] PROFILE_OUT[Profiling Results
Schema + PII + Quality] end subgraph "Agent 2: Dictionary" PARSE[Parse Schema] DDL_GEN[AI_COMPLETE
Generate DDL] OPTIMIZE[Optimize Types
AI-powered] ENRICH[Enrich Dictionary
Metadata Tables] DDL_OUT[DDL Proposals
Ready for Approval] end subgraph "Agent 4: Mapping" TARGET[Query Target
Information Schema] MAP_GEN[AI_COMPLETE
Field Mappings] TRANS[Generate SQL
Transformations] DBT_GEN[Create DBT Models
stg/int/fct] MAP_OUT[Transformation Code
Ready to Deploy] end subgraph Output LANDING[Landing Tables
STAGING Schema] CURATED[Curated Tables
CURATED Schema] DBT[DBT Project
Git Repository] end USER -->|Upload| FILE FILE -->|PUT| STAGE STAGE --> SAMPLE SAMPLE --> INFER INFER --> STATS STATS --> PII PII --> PROFILE_OUT PROFILE_OUT --> PARSE PARSE --> DDL_GEN DDL_GEN --> OPTIMIZE OPTIMIZE --> ENRICH ENRICH --> DDL_OUT DDL_OUT --> TARGET TARGET --> MAP_GEN MAP_GEN --> TRANS TRANS --> DBT_GEN DBT_GEN --> MAP_OUT DDL_OUT -->|Approved| LANDING MAP_OUT -->|DBT Run| CURATED MAP_OUT --> DBT style SAMPLE fill:#E8F4F8 style INFER fill:#C8E6C9 style PII fill:#FF6B35,color:#fff style DDL_GEN fill:#FF6B35,color:#fff style MAP_GEN fill:#FF6B35,color:#fff style LANDING fill:#FFE082 style CURATED fill:#81C784,color:#fff ``` --- ## 3. Agent Workflow Orchestration ```mermaid sequenceDiagram participant User as User/System participant UI as Streamlit UI participant Orch as Orchestrator participant WF as Workflow State participant A1 as Agent 1
Profiling participant A2 as Agent 2
Dictionary participant A4 as Agent 4
Mapping participant DB as Snowflake Tables participant AI as Cortex AI User->>UI: Upload file.csv UI->>Orch: CALL SP_ORCHESTRATE_ONBOARDING(...) Orch->>WF: Create workflow instance WF->>DB: INSERT INTO WORKFLOW_EXECUTIONS Note over Orch,A1: Step 1: Profiling Orch->>A1: Execute profiling A1->>DB: Load sample data A1->>AI: AI_CLASSIFY (PII detection) A1->>AI: AI_COMPLETE (synonyms) AI-->>A1: Results A1->>DB: Save profiling results A1-->>Orch: Return profile_id Note over Orch,A2: Step 2: Dictionary Orch->>A2: Execute dictionary(profile_id) A2->>DB: Retrieve profiling results A2->>AI: AI_COMPLETE (DDL generation) AI-->>A2: Generated DDL A2->>DB: Save DDL proposals A2->>DB: Enrich data dictionary A2-->>Orch: Return dictionary_id Note over Orch,A4: Step 3: Mapping Orch->>A4: Execute mapping(dictionary_id) A4->>DB: Query Information Schema A4->>AI: AI_COMPLETE (field mappings) AI-->>A4: Mappings + SQL A4->>DB: Save field mappings A4->>DB: Save DBT models A4-->>Orch: Return mapping_id Orch->>WF: Mark workflow COMPLETED WF->>DB: UPDATE WORKFLOW_EXECUTIONS Orch-->>UI: Return results UI-->>User: Show success + results ``` --- ## 4. Multi-Interface Architecture ```mermaid graph TB subgraph "External Users & Systems" USER1[Data Engineers
via Streamlit] USER2[AI Assistants
Claude/VS Code] USER3[External Apps
via REST API] USER4[Conversational
Natural Language] end subgraph "Interface Layer" STREAMLIT[Streamlit App
7 Pages] MCP[MCP Server
Tool Definitions] REST[REST API
SQL API Gateway] INTEL[Snowflake Intelligence
Cortex Analyst] end subgraph "Authentication" AUTH[Auth Layer] OAUTH[OAuth 2.0] KEYPAIR[Key Pair JWT] RBAC[Role-Based Access
AGENT_USER/ADMIN/VIEWER] end subgraph "Orchestration Core" ORCH[Master Orchestrator] AGENTS[Agent Execution Layer] end USER1 --> STREAMLIT USER2 --> MCP USER3 --> REST USER4 --> INTEL STREAMLIT --> AUTH MCP --> AUTH REST --> AUTH INTEL --> AUTH AUTH --> OAUTH AUTH --> KEYPAIR AUTH --> RBAC OAUTH --> ORCH KEYPAIR --> ORCH RBAC --> ORCH ORCH --> AGENTS style USER1 fill:#29B5E8,color:#fff style USER2 fill:#29B5E8,color:#fff style USER3 fill:#29B5E8,color:#fff style USER4 fill:#29B5E8,color:#fff style AUTH fill:#FF6B35,color:#fff style ORCH fill:#0073E6,color:#fff style AGENTS fill:#00A3E0,color:#fff ``` --- ## 5. Security & Governance Architecture ```mermaid graph TB subgraph "Data Classification" DETECT[AI_CLASSIFY
PII/PHI Detection] TAG[Object Tagging
Metadata Classification] CLASS[Data Classification
PUBLIC/INTERNAL/CONFIDENTIAL] end subgraph "Access Control" RBAC[Role-Based Access Control] ADMIN[AGENT_ADMIN
Full Access] USER[AGENT_USER
Execute & View] VIEWER[AGENT_VIEWER
Read-Only] end subgraph "Data Protection" MASK[Dynamic Data Masking
PII Fields] REDACT[Log Redaction
Sensitive Data] ENCRYPT[Column Encryption
At Rest] end subgraph "Audit & Compliance" LOG[Execution Logs
AGENT_EXECUTION_LOG] QUERY[Query History
Account Usage] METRICS[Cost Tracking
Token Usage] ALERT[Alerts
Anomalies/Thresholds] end subgraph "Guardrails" VALIDATE[Input Validation
Schema Checks] SYNTAX[SQL Validation
Parser Checks] APPROVE[Approval Gates
DDL/Mapping Review] LIMIT[Cost Limits
Resource Monitors] end DETECT --> TAG TAG --> CLASS CLASS --> MASK RBAC --> ADMIN RBAC --> USER RBAC --> VIEWER ADMIN --> LOG USER --> LOG VIEWER --> LOG MASK --> REDACT LOG --> QUERY LOG --> METRICS METRICS --> ALERT VALIDATE --> SYNTAX SYNTAX --> APPROVE APPROVE --> LIMIT style DETECT fill:#FF6B35,color:#fff style MASK fill:#E53935,color:#fff style RBAC fill:#0073E6,color:#fff style LOG fill:#00A3E0,color:#fff ``` --- ## 6. Technology Stack Diagram ```mermaid graph LR subgraph "Frontend" ST[Streamlit
Python UI] PLOT[Plotly
Visualizations] end subgraph "API Layer" OPENAPI[OpenAPI 3.0
Specification] SQLAPI[Snowflake SQL API
Native Endpoints] MCP_S[MCP Server
Tool Protocol] end subgraph "Orchestration" PYTHON[Python 3.10
Stored Procedures] SNOWPARK[Snowpark
DataFrame API] end subgraph "AI/ML" CORTEX[Cortex AI
8 Functions] MODELS[LLM Models
Mistral/Llama/Claude] FINETUNE[Fine-Tuning
Phase 2] end subgraph "Data Platform" COMPUTE[Virtual Warehouses
Auto-Scale] STORAGE[Stages & Tables
Time Travel] STREAM[Streams
Change Tracking] TASK[Tasks
Scheduling] end subgraph "Transformation" DBT_CORE[DBT Core
SQL Templates] JINJA[Jinja2
Templating] GIT[Git Integration
Version Control] end subgraph "Governance" RBAC_G[RBAC
3 Roles] TAG_G[Object Tags
Classification] MASK_G[Data Masking
Policies] INFO[Information Schema
Metadata] end ST --> PYTHON PLOT --> ST OPENAPI --> SQLAPI MCP_S --> SQLAPI PYTHON --> SNOWPARK SNOWPARK --> CORTEX CORTEX --> MODELS MODELS --> FINETUNE SNOWPARK --> COMPUTE COMPUTE --> STORAGE STORAGE --> STREAM STREAM --> TASK PYTHON --> DBT_CORE DBT_CORE --> JINJA JINJA --> GIT RBAC_G --> TAG_G TAG_G --> MASK_G MASK_G --> INFO style ST fill:#29B5E8,color:#fff style CORTEX fill:#FF6B35,color:#fff style COMPUTE fill:#0073E6,color:#fff style DBT_CORE fill:#FF6B35,color:#fff style RBAC_G fill:#00A3E0,color:#fff ``` --- ## 7. ASCII Art Architecture (Universal Compatibility) ``` ┌─────────────────────────────────────────────────────────────────────────────┐ │ LAYER 5: USER EXPERIENCE │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ │ │ Streamlit │ │ AI │ │ REST API │ │ MCP │ │ │ │ UI │ │ Assistant │ │ (OpenAPI) │ │ Interface │ │ │ │ 7 Pages │ │ NL Query │ │ Endpoints │ │ External │ │ │ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ │ └─────────┼──────────────────┼──────────────────┼──────────────────┼──────────┘ │ │ │ │ └──────────────────┴──────────────────┴──────────────────┘ │ ┌─────────────────────────────────────────────────────────────────────────────┐ │ LAYER 4: ORCHESTRATION & COMMUNICATION │ │ │ │ ┌───────────────────────────────────────────────────────────────────────┐ │ │ │ Master Orchestrator (SP_ORCHESTRATE_ONBOARDING) │ │ │ │ │ │ │ │ ┌──────────────────┐ ┌──────────────────┐ ┌───────────────┐ │ │ │ │ │ Workflow State │ │ Execution Log │ │ Metrics │ │ │ │ │ │ Machine │ │ & Tracking │ │ Collection │ │ │ │ │ └──────────────────┘ └──────────────────┘ └───────────────┘ │ │ │ └───────────────────────────────────────────────────────────────────────┘ │ └──────────────────┬───────────────────┬───────────────────┬──────────────────┘ │ │ │ ▼ ▼ ▼ ┌─────────────────────────────────────────────────────────────────────────────┐ │ LAYER 3: AGENT EXECUTION │ │ │ │ ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐ │ │ │ AGENT 1 │ │ AGENT 2 │ │ AGENT 4 │ │ │ │ ┌───────────┐ │ │ ┌───────────┐ │ │ ┌───────────┐ │ │ │ │ │ Profiling │ │ │ │Dictionary │ │ │ │ Mapping │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ • Schema │ │ │ │ • DDL Gen │ │ │ │ • Field │ │ │ │ │ │ • PII │──┼───┼─▶│ • Optimize│──┼───┼─▶│ Maps │ │ │ │ │ │ • Stats │ │ │ │ • Enrich │ │ │ │ • Transform│ │ │ │ │ │ • Quality │ │ │ │ • Validate│ │ │ │ • DBT Gen │ │ │ │ │ └─────┬─────┘ │ │ └─────┬─────┘ │ │ └─────┬─────┘ │ │ │ └────────┼────────┘ └────────┼────────┘ └────────┼────────┘ │ │ │ AI │ AI │ AI │ │ ▼ ▼ ▼ │ └─────────────────────────────────────────────────────────────────────────────┘ │ │ │ └──────────────────────┴─────────────────────┘ │ ┌─────────────────────────────────────────────────────────────────────────────┐ │ LAYER 2: INTELLIGENCE │ │ │ │ ┌───────────────────────────────────────────────────────────────────────┐ │ │ │ Snowflake Cortex AI │ │ │ │ Model Garden │ │ │ │ │ │ │ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────┐ │ │ │ │ │AI_COMPLETE│ │AI_CLASSIFY│ │AI_EXTRACT │ │AI_FILTER │ │AI_AGG │ │ │ │ │ │LLM Reason│ │PII/PHI Det│ │Entity Ext │ │Conditions│ │Insights│ │ │ │ │ └──────────┘ └──────────┘ └──────────┘ └──────────┘ └────────┘ │ │ │ │ │ │ │ │ Models: Mistral Large, Llama 3.1, Claude 3.5, DeepSeek │ │ │ └───────────────────────────────────────────────────────────────────────┘ │ │ │ │ Non-AI Features: │ │ • INFER_SCHEMA (Schema Detection) • Information Schema (Metadata) │ │ • Object Tagging (Classification) • Dynamic Masking (PII Protection) │ └─────────────────────────────────────────────────────────────────────────────┘ │ ┌─────────────────────────────────────────────────────────────────────────────┐ │ LAYER 1: FOUNDATIONAL PLATFORM │ │ │ │ ┌───────────────────────────────────────────────────────────────────────┐ │ │ │ Snowflake Data Cloud │ │ │ │ │ │ │ │ ┌────────────────┐ ┌────────────────┐ ┌────────────────────────┐ │ │ │ │ │ Compute │ │ Storage │ │ Governance │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ • 3 Warehouses │ │ • 17 Tables │ │ • RBAC (3 roles) │ │ │ │ │ │ • Auto-Suspend │ │ • Stages │ │ • Object Tagging │ │ │ │ │ │ • Auto-Scale │ │ • Time Travel │ │ • Data Masking │ │ │ │ │ │ • Snowpark │ │ • File Formats │ │ • Audit Logs │ │ │ │ │ └────────────────┘ └────────────────┘ └────────────────────────┘ │ │ │ └───────────────────────────────────────────────────────────────────────┘ │ └─────────────────────────────────────────────────────────────────────────────┘ Data Flow: User → Agent 1 → Agent 2 → Agent 4 State Management: All in Snowflake Tables Security: RBAC + PII Detection + Masking + Audit Logs ``` --- ## 8. Agent Interaction Flow (Simplified) ``` FILE UPLOAD │ ▼ ┌─────────────────────────────────────────────────────────────────┐ │ ORCHESTRATOR │ │ (Workflow Manager) │ └────┬──────────────────┬──────────────────┬─────────────────────┘ │ │ │ │ 1. Profile │ 2. Dictionary │ 3. Mapping ▼ ▼ ▼ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ AGENT 1 │ │ AGENT 2 │ │ AGENT 4 │ │ │ │ │ │ │ │ Schema │──────▶│ DDL Gen │──────▶│ Field │ │ PII │ JSON │ Metadata │ JSON │ Mappings │ │ Stats │ │ Enrich │ │ DBT Code │ └────┬─────┘ └────┬─────┘ └────┬─────┘ │ │ │ ▼ ▼ ▼ ┌──────────────────────────────────────────────────┐ │ SNOWFLAKE TABLES │ │ │ │ • PROFILING_RESULTS • DDL_PROPOSALS │ │ • AGENT_HISTORY • FIELD_MAPPINGS │ │ • WORKFLOW_EXECUTIONS • MONITORING │ └──────────────────────────────────────────────────┘ ``` --- ## 9. Deployment Architecture ``` DEPLOYMENT ENVIRONMENTS ┌────────────────────┐ ┌────────────────────┐ ┌────────────────────┐ │ DEV ENVIRONMENT │ │ TEST ENVIRONMENT │ │ PROD ENVIRONMENT │ │ │ │ │ │ │ │ • Development │ │ • Integration │ │ • Production │ │ • Testing │ │ • UAT │ │ • Live Data │ │ • Debugging │ │ • Performance │ │ • SLA Monitoring │ │ │ │ • Security Tests │ │ • DR/HA │ │ AGENTIC_PLATFORM_ │ │ AGENTIC_PLATFORM_ │ │ AGENTIC_PLATFORM_ │ │ DEV │ │ TEST │ │ PROD │ │ │ │ │ │ │ │ ┌──────────────┐ │ │ ┌──────────────┐ │ │ ┌──────────────┐ │ │ │ 6 Schemas │ │ │ │ 6 Schemas │ │ │ │ 6 Schemas │ │ │ │ • AGENTS │ │ │ │ • AGENTS │ │ │ │ • AGENTS │ │ │ │ • METADATA │ │ │ │ • METADATA │ │ │ │ • METADATA │ │ │ │ • WORKFLOWS │ │ │ │ • WORKFLOWS │ │ │ │ • WORKFLOWS │ │ │ │ • STAGING │ │ │ │ • STAGING │ │ │ │ • STAGING │ │ │ │ • CURATED │ │ │ │ • CURATED │ │ │ │ • CURATED │ │ │ │ • MONITORING │ │ │ │ • MONITORING │ │ │ │ • MONITORING │ │ │ └──────────────┘ │ │ └──────────────┘ │ │ └──────────────┘ │ └────────────────────┘ └────────────────────┘ └────────────────────┘ │ │ │ └─────────────────────────┴─────────────────────────┘ │ ┌──────────────▼──────────────┐ │ GIT REPOSITORY │ │ • SQL Scripts │ │ • Agent Code │ │ • DBT Projects │ │ • Documentation │ └─────────────────────────────┘ ``` --- ## Legend ### Colors (in Mermaid diagrams) - 🔵 **Blue (#29B5E8)**: User interfaces and interaction points - 🟠 **Orange (#FF6B35)**: AI-powered components (Cortex AI) - 🔷 **Dark Blue (#0073E6)**: Agent execution layer - 🟢 **Green (#C8E6C9)**: Non-AI Snowflake features - ⚫ **Black (#333)**: Foundational platform components ### Symbols - **→** : Data flow - **▼** : Layer connection - **├─** : Component relationship - **[Box]** : System component - **{Curly}** : Process/function --- ## How to Use These Diagrams 1. **For Executive Presentations**: Use diagrams 1, 2, and 7 (high-level overview) 2. **For Technical Reviews**: Use diagrams 3, 4, and 5 (detailed workflows) 3. **For Security Audits**: Use diagram 5 (security architecture) 4. **For Developer Onboarding**: Use diagrams 2, 3, and 6 (data flow and tech stack) 5. **For Deployment Planning**: Use diagram 9 (deployment architecture) --- ## Diagram Tools These diagrams use: - **Mermaid.js** for interactive diagrams (supported by GitHub, GitLab, VS Code) - **ASCII Art** for universal compatibility (view in any text editor) To render Mermaid diagrams: - **GitHub/GitLab**: Automatically rendered in markdown - **VS Code**: Install "Markdown Preview Mermaid Support" extension - **Online**: Use [mermaid.live](https://mermaid.live) editor --- **Last Updated**: October 17, 2025 **Version**: 1.0 **Part of**: Snowflake Agentic AI Platform MVP Documentation

Design & Creative

Snowflake Agentic AI Platform - Architecture Diagrams