Snowflake Cortex AI

Cursor-Regel für Snowflake Cortex AI — korrekte Nutzung der LLM-Funktionen direkt in SQL.

Original-Beschreibung der Autoren: Cursor rules for Snowflake Cortex AI Functions (AI_COMPLETE, AI_CLASSIFY, AI_EXTRACT, etc.) and Cortex Search for RAG applications.

Die Regel

---
description: "Cursor rules for Snowflake Cortex AI Functions (AI_COMPLETE, AI_CLASSIFY, AI_EXTRACT, etc.) and Cortex Search for RAG applications."
globs: **/*
alwaysApply: false
---
// Snowflake Cortex AI
// Expert guidance for Cortex AI Functions and Cortex Search (hybrid vector+keyword search)

You are an expert in Snowflake Cortex — the AI layer of Snowflake including Cortex AI Functions (SQL-callable LLM/ML functions) and Cortex Search (managed hybrid search for RAG applications). All processing runs inside Snowflake with no data leaving the platform.

// ═══════════════════════════════════════════
// CORTEX AI FUNCTIONS
// ═══════════════════════════════════════════

// Available Functions (use these names — they are the current versions):
// AI_COMPLETE       — General-purpose LLM completion (text, images, documents).
// AI_CLASSIFY       — Classify text/images into user-defined categories (multi-label supported).
// AI_FILTER         — Returns TRUE/FALSE for text/image input. Use in WHERE clauses.
// AI_AGG            — Aggregate insights across rows of text (no context window limit).
// AI_EMBED          — Generate embedding vectors (similarity search, clustering).
// AI_EXTRACT        — Extract structured info from text, images, or documents.
// AI_SENTIMENT      — Sentiment score from text (-1 to 1).
// AI_SUMMARIZE_AGG  — Summarize across rows (no context window limit).
// AI_SIMILARITY     — Embedding similarity between two inputs.
// AI_TRANSCRIBE     — Transcribe audio/video from stages.
// AI_PARSE_DOCUMENT — OCR or text+layout extraction from documents in stages.
// AI_REDACT         — Redact PII from text.
// AI_TRANSLATE      — Translate between supported languages.

// Helper Functions:
// TO_FILE('@stage', 'filename')  — File reference for document processing.
// AI_COUNT_TOKENS(model, text)   — Check token count before calling a model.
// PROMPT('template {0}', arg)    — Build prompt objects for AI_COMPLETE.
// TRY_COMPLETE                   — Returns NULL on failure instead of error.

// AI_COMPLETE — The Primary Function
// Models: claude-4-opus, claude-4-sonnet, claude-sonnet-4-5, claude-opus-4-5, claude-haiku-4-5,
//         gemini-3-pro, llama3.1-70b, llama3.1-8b, llama3.3-70b, mistral-large2, mistral-small2, deepseek-r1

// Text completion:
SELECT AI_COMPLETE(MODEL => 'claude-4-sonnet', PROMPT => 'Summarize: ' || review_text) FROM reviews;

// Document processing:
SELECT AI_COMPLETE(
  MODEL => 'claude-4-sonnet',
  PROMPT => PROMPT('Extract the invoice total from {0}', TO_FILE('@docs', 'invoice.pdf'))
);

// Structured JSON output:
SELECT AI_COMPLETE(MODEL => 'claude-4-sonnet',
  PROMPT => 'Extract name, email, company as JSON: ' || raw_text)::VARIANT AS extracted FROM contacts;

// AI_CLASSIFY:
SELECT AI_CLASSIFY(ticket_text, ['billing', 'technical', 'account', 'other']) AS category FROM tickets;
// Multi-label: AI_CLASSIFY(input, categories, {'output_mode': 'multi'})

// AI_FILTER (natural-language WHERE):
SELECT * FROM reviews WHERE AI_FILTER(review_text, 'mentions product quality issues');

// AI_AGG (cross-row aggregation):
SELECT AI_AGG(feedback_text, 'What are the top 3 themes?') FROM customer_feedback;

// AI_EXTRACT (entity extraction):
SELECT AI_EXTRACT(email_body, 'meeting date', 'attendees', 'action items') FROM emails;

// AI_SENTIMENT: SELECT review_text, AI_SENTIMENT(review_text) AS sentiment FROM product_reviews;
// AI_EMBED:     SELECT AI_EMBED(description) AS embedding FROM products;
// AI_PARSE_DOCUMENT: SELECT AI_PARSE_DOCUMENT(TO_FILE('@docs', 'contract.pdf'), MODE => 'LAYOUT');
// AI_TRANSCRIBE:     SELECT AI_TRANSCRIBE(TO_FILE('@media', 'recording.mp3')) AS transcript;
// AI_REDACT:         SELECT AI_REDACT(customer_notes) AS redacted FROM support_cases;

// Privileges: USE AI FUNCTIONS account privilege + SNOWFLAKE.CORTEX_USER database role (both granted to PUBLIC by default).

// ═══════════════════════════════════════════
// CORTEX SEARCH — Hybrid Vector + Keyword Search
// ═══════════════════════════════════════════

// Fully managed search combining vector (semantic) and keyword (lexical) search.
// Use cases: RAG for LLM chatbots, enterprise search, AI-powered Q&A.

// Single-index (simplest):
CREATE OR REPLACE CORTEX SEARCH SERVICE my_search
  ON transcript_text
  ATTRIBUTES region, agent_id
  WAREHOUSE = my_wh
  TARGET_LAG = '1 day'
  EMBEDDING_MODEL = 'snowflake-arctic-embed-l-v2.0'
  AS (SELECT transcript_text, region, agent_id FROM support_transcripts);

// Multi-index (text + vector on multiple columns):
CREATE OR REPLACE CORTEX SEARCH SERVICE my_multi_search
  TEXT INDEXES transcript_text, summary
  VECTOR INDEXES transcript_text (model='snowflake-arctic-embed-l-v2.0')
  ATTRIBUTES region
  WAREHOUSE = my_wh
  TARGET_LAG = '1 hour'
  AS (SELECT transcript_text, summary, region FROM support_transcripts);

// Key Parameters: ON (single-index column), TEXT INDEXES, VECTOR INDEXES, ATTRIBUTES (filter columns),
// TARGET_LAG (freshness), EMBEDDING_MODEL, PRIMARY KEY (optimized incremental refresh).

// Query — Python API (recommended for apps):
from snowflake.core import Root
root = Root(session)
service = root.databases["db"].schemas["schema"].cortex_search_services["my_search"]
resp = service.search(
    query="internet connection issues",
    columns=["transcript_text", "region"],
    filter={"@eq": {"region": "North America"}},
    limit=5
)

// Query — REST API:
// POST /api/v2/databases/<db>/schemas/<schema>/cortex-search-services/<service>:query
// Body: {"query": "...", "columns": [...], "filter": {...}, "limit": N}

// Filter syntax:
// {"@eq": {"region": "NA"}}, {"@contains": {"tags": "urgent"}}, {"@gte": {"score": 0.8}}
// {"@and": [f1, f2]}, {"@or": [f1, f2]}, {"@not": f}

// Scoring config — adjust text vs vector vs reranker weights:
resp = service.search(query="billing dispute", columns=["transcript_text"],
    scoring_config={"weights": {"texts": 0.3, "vectors": 0.5, "reranker": 0.2}}, limit=10)

// RAG Pattern:
… (hier gekürzt — Kopieren/Download liefert die vollständige Regel)

So nutzt du sie

Die Regel kopieren (Button oben) oder als Datei herunterladen und im Projekt unter .cursor/rules/ ablegen — Cursor lädt sie beim nächsten Start automatisch. Ältere Cursor-Versionen lesen alternativ eine einzelne .cursorrules-Datei im Projektstamm; dort einfach den Regel-Text ohne den Kopfblock zwischen den ----Zeilen einfügen.

Der Regel-Text ist englisch — Cursor versteht ihn unabhängig von der Sprache, in der Sie mit dem Editor chatten.

Im Detail

Regel für die Arbeit mit Snowflake Cortex AI, den in Snowflake integrierten LLM- und ML-Funktionen (z. B. COMPLETE, SUMMARIZE, EXTRACT_ANSWER direkt in SQL). Sie gibt Cursor Konventionen vor, wie Cortex-Funktionen korrekt in SQL- bzw. Snowpark-Code eingebunden werden, inklusive Umgang mit Kosten- und Modellwahl-Aspekten. Interessant für Data-Teams, die LLM-Funktionalität direkt im Data Warehouse nutzen wollen, ohne Daten in externe APIs zu exportieren — etwa für Textklassifikation oder Zusammenfassungen auf Snowflake-Tabellen. Wer noch keine Cortex-Funktionen nutzt oder auf externe LLM-APIs setzt, hat wenig Nutzen davon, da die Regel sich auf die Snowflake-spezifische SQL-Syntax konzentriert.

Praxis-Tipp

Beispiel-Prompt: „Schreibe eine SQL-Query, die SNOWFLAKE.CORTEX.SUMMARIZE auf die Spalte review_text anwendet” — die Regel sorgt für korrekte Funktionssignatur und Fehlerbehandlung.

Lizenz & Quelle

• Lizenz: CC0 1.0
• Quelle: PatrickJS/awesome-cursorrules (GitHub)

Lade …