Docs/Precision vs Performance

Precision vs Performance

Maximum accuracy rendering engine with >99.5% element detection rate. Perfect for visual QA, dataset generation, and mission-critical automation.

Accuracy Rate
>99.5%
element detection
Compute Cost
10 Credits
per request
Latency
~1-2s
average response

Core Capabilities

Precision Mode is our most accurate rendering engine, designed for use cases where pixel-perfect element detection is critical. It uses advanced rendering techniques to achieve >99.5% accuracy, ensuring virtually no interactive elements are missed.

Key Features

  • Maximum Accuracy: >99.5% element detection rate vs 97% in Performance Mode
  • Pixel-Perfect Coordinates: Highly accurate bounding boxes for visual AI models
  • Complex Site Handling: Excels at sites with dynamic content, overlays, and shadows
  • Visual Mode Ready: Required for Visual Mode (screenshot + coordinates)
  • Advanced Rendering: Handles JavaScript-heavy apps, animations, and edge cases

Performance vs Precision

FeaturePerformance ModePrecision Mode
Accuracy Rate97%>99.5%
Response Time<400ms~1-2s
Credit Cost2 credits10 credits
Complex SitesGoodExcellent
Dynamic ContentGoodExcellent
Visual Mode SupportNot availableRequired
Best ForHigh-volume automation, RAG, general useVisual QA, datasets, critical automation

How Precision Mode Works

Advanced techniques for maximum accuracy

Precision Mode employs sophisticated rendering techniques to achieve industry-leading accuracy without exposing implementation details:

Full Page Rendering

Waits for all assets, animations, and dynamic content to fully load before analyzing the page, ensuring nothing is missed.

Enhanced Element Detection

Uses multiple detection strategies to identify interactive elements, including those hidden by overlays, shadows, or complex layouts.

Coordinate Validation

Validates bounding box coordinates against multiple reference points to ensure pixel-perfect accuracy for visual AI models.

State Synchronization

Captures accurate element states (disabled, checked, focused) even for dynamically updating components.

Request Format

Enable Precision Mode by setting render_quality

Precision Mode Request

1curl -X POST https://api.sentienceapi.com/v1/observe \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{
5    "url": "https://app.example.com/dashboard",
6    "mode": "map",
7    "options": {
8      "render_quality": "precision"
9    }
10  }'

Precision Mode with Smart Filtering

Combine Precision Mode's accuracy with smart filtering to focus on the most important elements while maintaining maximum accuracy.

1curl -X POST https://api.sentienceapi.com/v1/observe \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{
5    "url": "https://app.example.com/dashboard",
6    "mode": "map",
7    "options": {
8      "render_quality": "precision",
9      "limit": 100,
10      "filter": {
11        "allowed_roles": ["button", "textbox", "link"],
12        "min_area": 400
13      }
14    }
15  }'

Key Parameter

options.render_qualityrequiredstring

Set to "precision" to enable Precision Mode. Defaults to "performance" if not specified.

Response Format

Same structure as Map Mode, with enhanced accuracy

Example Response

1{
2  "engine": "precision",
3  "status": "success",
4  "url": "https://app.example.com/dashboard",
5  "layout_viewport": {
6    "width": 1024,
7    "height": 768,
8    "device_pixel_ratio": 1.0
9  },
10  "interactable_elements": [
11    {
12      "id": 1,
13      "uid": "search-input",
14      "tag": "input",
15      "role": "textbox",
16      "text": "",
17      "selector": "input#search-input",
18      "bbox": {
19        "x": 850,
20        "y": 25,
21        "w": 220,
22        "h": 36
23      },
24      "is_visible": true,
25      "z_index": 10,
26      "attributes": {
27        "type": "search",
28        "placeholder": "Search...",
29        "aria_label": "Search dashboard"
30      }
31    },
32    {
33      "id": 2,
34      "uid": "create-new-btn",
35      "tag": "button",
36      "role": "button",
37      "text": "Create New",
38      "selector": "button.btn-primary",
39      "bbox": {
40        "x": 1100,
41        "y": 20,
42        "w": 120,
43        "h": 44
44      },
45      "is_visible": true,
46      "z_index": 10,
47      "attributes": {
48        "aria_label": "Create new project"
49      },
50      "states": {
51        "is_disabled": false
52      }
53    }
54  ],
55  "timestamp": "2025-12-11T10:30:00.456Z",
56  "total_elements_extracted": 2,
57  "filters_applied": {
58    "limit": null,
59    "filter": null
60  }
61}
Note: The response format is identical to Map Mode. The key difference is the "engine": "precision" field and the enhanced accuracy of coordinates and element detection.

When to Use Precision Mode

Visual QA & Testing

Verify pixel-perfect layouts, detect visual regressions, and validate UI elements with maximum accuracy.

Dataset Generation

Create high-quality training datasets for computer vision models with accurate element annotations.

Visual AI Models

Provide pixel-perfect coordinates to vision-language models for screenshot analysis and interaction.

Critical Automation

Mission-critical workflows where missing a single element could cause failures or data loss.

💡 When to Use Performance Mode Instead

For high-volume automation, RAG systems, or general use cases where 97% accuracy is sufficient, Performance Mode offers 5x cost savings (2 credits vs 10) and faster response times (<400ms vs ~1-2s).

Visual Mode (Precision Required)

Screenshot capture with element coordinates

Visual Mode automatically uses Precision Mode to capture screenshots with accurate element coordinates. This is perfect for visual AI models that need both visual and structural information.

Visual Mode Request

1curl -X POST https://api.sentienceapi.com/v1/observe \
2  -H "Authorization: Bearer YOUR_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{
5    "url": "https://app.example.com/dashboard",
6    "mode": "visual"
7  }'
Cost: Visual Mode always uses Precision Mode and costs 10 credits per request. You don't need to specify render_quality - it's automatic.

Cost Optimization

Get the most value from Precision Mode

✓ Use Smart Filtering

Combine Precision Mode with limit and filter options to reduce token costs while maintaining accuracy.

✓ Reserve for Critical Workflows

Use Performance Mode for exploratory work and switch to Precision Mode only when accuracy is mission-critical.

✓ Batch Similar Requests

If processing multiple similar pages, use Precision Mode on one representative page to validate, then switch to Performance Mode for the rest.

Next Steps

Ready to explore other modes or dive deeper into the API?