Setup Guide – Crawleye Prompt-Based AI Video Analysis

Welcome to Crawleye — the prompt-driven AI video analysis platform.
Whether you want to analyze live RTSP camera feeds or uploaded recordings, Crawleye lets you define exactly what the AI should detect, track, or analyze.

In this guide you’ll learn how to:

Create your Crawleye account
Add videos (upload or connect RTSP feeds)
Write effective AI prompts
Choose analysis models and features
Retrieve and use your results
Manage billing & usage

1️⃣ Create Your Account

Go to https://crawleye.com.
Click Sign Up and fill in your details:
- Email & Password OR social login
- Accept Terms of Service and Privacy Policy
Confirm your email address (check spam if you don’t see the verification link).

💡 Why register?
An account lets you save prompts, track processed minutes, and securely access results.

2️⃣ Add Your First Video

Crawleye supports two input methods:

A) Upload Recorded Video

Navigate to Videos → Upload
Click New Upload — system returns a presigned URL
Drag & drop your file or use the generated upload link
Supported formats: MP4, AVI, MOV, MKV
Max file size per upload: Based on your plan (see pricing)

B) Connect an RTSP Camera Feed

Navigate to Cameras → Add New
Enter camera name, RTSP URL, and optional description
Test connection — if successful, camera is added
Streams can be analyzed continuously or at scheduled intervals

📌 Tip: For security, use a dedicated network/VPN for camera access.

3️⃣ Write an Effective Prompt

Crawleye’s prompt-based detection means you can ask for anything you can describe.

Examples:

“Count the number of forklifts entering the loading dock every 5 minutes”
“Detect all people without safety helmets and mark timestamps”
“Track the number of occupied parking spots over the course of the video”
“Extract all text from signs and labels using OCR”
“Identify moments where two workers are within 1 meter of each other for more than 5 seconds”

Best practices:

Be specific (“count red trucks” vs. “detect trucks”)
Include time intervals if relevant
State conditions (“only if in restricted area”)
Combine features (object detection + OCR + action recognition)

4️⃣ Choose Models & Features

When starting an analysis, choose:

Model Size:
- S (7B) – Fastest, lowest cost
- M (13B) – Balanced
- L (32B) – Most accurate & detailed
Features:
- Object / Action detection
- OCR (text extraction)
- Re-ID / tracking (follow same object over time)
- LLM summary (natural language report)

💡 Higher model sizes and more features use more GPU minutes.

5️⃣ Run the Analysis

Go to Videos or Cameras
Click Analyze
Fill in:
- Prompt text
- Model & features
- Optional webhook URL for automated results delivery
Click Start Analysis

The job will queue and run on Crawleye’s GPU infrastructure.
You’ll see:

Queued → Processing → Completed status
Estimated and final cost in minutes & USD

6️⃣ Retrieve & Use Results

When complete:

View results in dashboard
Download:
- JSON detections
- CSV timelines
- Annotated video previews
- LLM-generated summaries
If webhook provided, results are sent automatically

Use results for:

Compliance audits
Operational reports
Security incident logs
Research data analysis

7️⃣ Monitor Usage & Billing

GET /api/usage (or view in dashboard) for processed minutes & storage
Purchase extra credits or upgrade plan in Billing
Review detailed charges & invoices in Payment History

8️⃣ API Quickstart

If integrating programmatically:

# Upload video
curl -X POST "https://api.crawleye.com/v1/videos/upload" \
  -H "X-API-Key: YOUR_KEY"

# Analyze with prompt
curl -X POST "https://api.crawleye.com/v1/videos/analyze" \
  -H "X-API-Key: YOUR_KEY" \
  -d '{
        "file_id": "abc123",
        "prompt": "Count people without safety helmets",
        "model_class": "M_13B",
        "features": ["detect_only"]
      }'

# Get results
curl -X GET "https://api.crawleye.com/v1/videos/{video_id}/results" \
  -H "X-API-Key: YOUR_KEY"

  Full API reference: Video Analysis API