catalyst-zia

어떤 스킬인가요?

Catalyst Zia Services and QuickML — OCR, Face Analytics, Text Analytics, Object Detection, Barcode Reader, Content Moderation, and AutoML predictions. Trigger on 'Zia', 'QuickML', 'OCR', 'face detection', 'text analytics', 'AutoML', 'ML model', or 'train a model on Catalyst'. DC restrictions: Identity Scanner is IN DC only; AutoML/QuickML is not available in EU, AU, IN, JP, SA, CA data centers.

설치 방법

프로젝트 폴더에서 아래 명령어를 실행하고, 설치할 에이전트를 선택하세요.

설치 명령을 직접 실행해야 적용됩니다. 지원 에이전트와 필요한 권한·라이선스는 제작자의 안내를 확인하세요.

스킬 원문 파일

## Overview

QuickML is Catalyst's no-code AutoML platform. Upload a dataset, configure the problem type, train a model, and call it via SDK/API. No ML expertise required.

---

## Workflow

1. **Import Dataset** — CSV/Excel from Stratus, Data Store export, or direct upload
2. **Configure Training** — Select target column, problem type, algorithm
3. **Train** — QuickML runs feature engineering, model selection, cross-validation
4. **Deploy** — Deploy the best model as an endpoint
5. **Predict** — Call predictions from functions or external services

---

## Problem Types

| Type | Use Case | Example |
|------|----------|---------|
| `classification` | Classify into categories | Spam/Not Spam, Churn/No Churn |
| `regression` | Predict numerical values | Price prediction, Sales forecasting |
| `multi_label` | Multiple simultaneous labels | Tag assignment, Multi-category |

---

## SDK — Prediction

```javascript
const quickML = catalystApp.quickML();

// Get a model
const model = quickML.model(MODEL_ID);  // Model ID from console

// Single prediction
const result = await model.predict({
  feature1: 'value1',
  feature2: 42,
  feature3: 'category_a'
});
// { prediction: 'positive', confidence: 0.87 }

// Batch prediction
const batchResult = await model.batchPredict([
  { feature1: 'val1', feature2: 10 },
  { feature1: 'val2', feature2: 20 }
]);
```

---

## REST API

```
# Single prediction
POST /api/v1/ml/models/{model_id}/predict
Authorization: Zoho-oauthtoken {token}
{
  "feature1": "value",
  "feature2": 42
}
```

---

## Pricing

| Resource | Free Tier | Cost |
|----------|-----------|------|
| Training compute | 1 model/month | $0.10/model/hour |
| Predictions | 500/month | $0.001/prediction |
| Model storage | 1 model active | $5/model/month |

## Common Errors

| Error | Cause | Fix |
|-------|-------|-----|
| Model training stuck in `PROCESSING` | Dataset too small (< 50 rows) or all rows have the same target value | Add more varied data; QuickML requires at least 50 rows with distribution across classes |
| Prediction returns `null` | Feature columns in prediction request don't match training column names exactly | Match feature names case-sensitively to training dataset headers |
| `Model not deployed` error on predict | Model trained but deployment step skipped | Explicitly deploy model from Console → QuickML → Deploy before calling prediction API |
| Free tier prediction limit hit | 500 predictions/month free tier exhausted | Upgrade plan or wait for next calendar month reset |

## Overview

Catalyst Zia provides pre-trained AI/ML models as serverless APIs. No model training required — call the SDK methods and get results.

Pricing: Pay-per-call. Refer to the `catalyst-pricing` skill for current rates and free tier details.

> **All Zia methods are called directly on the `zia` object** — there are no sub-objects like `textAnalytics()`, `OCR()`, or `faceAnalytics()`.

---

## DC Restrictions

| Service | Availability |
|---------|-------------|
| Identity Scanner (Facial Comparison / Aadhaar) | **IN DC only** |
| AutoML / QuickML | **Not available** in EU, AU, IN, JP, SA, CA |
| All other Zia services | All data centers |

---

## Text Analytics

```javascript
const zia = catalystApp.zia();

// Sentiment Analysis
const sentimentResult = await zia.getSentimentAnalysis(
  ['I love this product!'],  // array of text strings
  ['optional', 'keywords']   // optional keyword hints
);

// Named Entity Recognition (NER)
const nerResult = await zia.getNERPrediction([
  'John works at Zoho in Chennai'
]);

// Keyword Extraction
const keywordsResult = await zia.getKeywordExtraction([
  'Catalyst is a cloud backend platform.'
]);

// All Text Analytics combined
const allResult = await zia.getTextAnalytics(
  ['Zoho Corporation is a multinational technology company.'],
  ['Zoho']  // optional keywords
);
```

---

## OCR (Optical Character Recognition)

```javascript
const zia = catalystApp.zia();

// Basic OCR
const result = await zia.extractOpticalCharacters(
  fs.createReadStream('./invoice.png')
);

// With options (model type, language)
const panResult = await zia.extractOpticalCharacters(
  fs.createReadStream('./pan.webp'),
  { modelType: 'PAN' }
);
```

Supported formats: JPG, PNG, WEBP

---

## Face Analytics

```javascript
const zia = catalystApp.zia();

// Face detection with attributes (age, gender, emotion, smile)
const faceResult = await zia.analyseFace(
  fs.createReadStream('./face.png')
);

// Facial Comparison / E-KYC (⚠️ IN DC only — part of Identity Scanner)
const compareResult = await zia.compareFace(sourceImageStream, queryImageStream);
// { match: true/false, confidence: 0–1 }
```

---

## Object Detection

```javascript
const zia = catalystApp.zia();

const detections = await zia.detectObject(
  fs.createReadStream('./sample.webp')
);
// [{ label: 'car', confidence: 0.94, ... }]
```

---

## Barcode Scanner

```javascript
const zia = catalystApp.zia();

const barcodes = await zia.scanBarcode(
  fs.createReadStream('./barcode.png'),
  { format: 'code39' }  // optional; use 'ALL' for auto-detect
);
```

Supported formats: QR Code, Code 128, EAN, UPC, Data Matrix, PDF417, and more.

---

## Moderation

```javascript
const zia = catalystApp.zia();

const result = await zia.moderateImage(
  fs.createReadStream('./image.png'),
  { mode: 'moderate' }  // 'basic', 'moderate', or 'advanced' (default)
);
```

---

## Common Patterns

### OCR → Data Store pipeline

```javascript
module.exports = async (context, basicIO) => {
  const { image_row_id } = basicIO.getArgument();  // singular, not getArguments()
  
  // Get image from Stratus
  const bucket = catalystApp.stratus().bucket('myapp-files');
  const imageStream = await bucket.getObject('images/doc.png');
  
  // Run OCR
  const zia = catalystApp.zia();
  const ocrResult = await zia.extractOpticalCharacters(imageStream);
  
  // Store extracted text
  await catalystApp.datastore().table('OCRResults').insertRow({
    SourceRowID: image_row_id,
    ExtractedText: JSON.stringify(ocrResult),
    ProcessedAt: new Date().toISOString()
  });
  
  basicIO.write({ status: 'success' });  // write(), not setOutput()
  context.closeWithSuccess();
};
```

## Common Errors

| Error | Cause | Fix |
|-------|-------|-----|
| OCR returns empty string | Image resolution too low or file format not supported | Use JPG/PNG/WEBP at ≥ 150 DPI; avoid scanned PDFs without embedded text |
| `API limit exceeded` on Zia call | Free tier Zia API call quota exhausted | Upgrade plan; cache Zia results for identical inputs to reduce repeat calls |
| `Invalid file format` on image analysis | Unsupported MIME type sent | Supported formats: JPG, PNG, WEBP; convert before sending |
| Zia response latency > 5s | Large image or complex document sent synchronously | For bulk processing, use a Job function + async dispatch rather than a Basic I/O function |
| Identity Scanner or AutoML not available | DC restriction | Identity Scanner: IN DC only. AutoML/QuickML: not available in EU, AU, IN, JP, SA, CA |

---
name: catalyst-zia
description: "Catalyst Zia Services and QuickML — OCR, Face Analytics, Text Analytics, Object Detection, Barcode Reader, Content Moderation, and AutoML predictions. Trigger on 'Zia', 'QuickML', 'OCR', 'face detection', 'text analytics', 'AutoML', 'ML model', or 'train a model on Catalyst'. DC restrictions: Identity Scanner is IN DC only; AutoML/QuickML is not available in EU, AU, IN, JP, SA, CA data centers."
metadata:
  version: "2.0.0"
---

## How It Works

1. **Identify the capability** — OCR, Face Analytics, Text Analytics, Object Detection, Barcode Reader, Content Moderation (Zia Services), or AutoML/predictions (QuickML).
2. **Load `references/zia-services.md`** — for all Zia API calls with Node.js and Python examples.
3. **Load `references/quickml.md`** — for AutoML workflow (dataset upload → training → prediction) and pricing.
4. **Show both SDK examples** — Zia reference includes Node.js and Python; provide both or ask the user which platform they're using.

## Triggers

Use this skill for: "Zia", "QuickML", "OCR", "face detection", "text analytics", "object detection", "barcode reader", "content moderation", "AutoML", "ML model", "predict", "Zia Services", "image recognition", "sentiment analysis", "train a model on Catalyst", or "Zia API".

## References

| Reference | Load when the query is about… |
|-----------|-------------------------------|
| `references/zia-services.md` | All Zia APIs — Text Analytics, OCR, Face Analytics, Object Detection, Barcode Reader, Moderation — SDK examples in Node.js and Python |
| `references/quickml.md` | AutoML workflow, problem types (classification/regression), dataset upload, SDK prediction calls, pricing |