examples

Examples

This directory contains example data, scripts, and usage examples for PATAS integration with Telegram.

📁 Files

Sample Data

• sample_telegram_logs.jsonl - Example JSONL file with 15 Telegram messages (mix of spam and ham)
  • Format: One JSON object per line (JSONL)
  • Contains: message_id, chat_id, user_id, created_at, text, language, message_type, has_media, labels
  • Use this for initial PoC and testing

Scripts

• run_poc_example.sh - Example bash script demonstrating a typical PoC workflow
  • Shows how to configure and run PoC
  • Includes error checking and output organization
  • Can be customized for your environment

🚀 Quick Start

1. Run PoC with Sample Data

# Using the example script
./examples/run_poc_example.sh

# Or manually
patas-tg poc \
    --config=config/config.yaml \
    --input=examples/sample_telegram_logs.jsonl \
    --out=artifacts/poc_report.md

2. Use Your Own Data

Prepare your Telegram logs in JSONL format matching the schema in docs/TELEGRAM_DATA_CONTRACT.md:

patas-tg poc \
    --config=config/config.yaml \
    --input=/path/to/your/telegram_logs.jsonl \
    --out=artifacts/my_poc_report.md

📋 Data Format

Each line in the JSONL file should be a JSON object with the following fields:

Required fields:

• message_id (string) - Unique message identifier
• text (string) - Message content
• created_at (ISO 8601 timestamp) - Message timestamp

Optional but recommended:

• chat_id (string) - Chat/channel identifier
• user_id (string) - User identifier
• language (string) - Language code (e.g., "ru", "en")
• message_type (string) - Message type (e.g., "text", "photo", "video")
• has_media (boolean) - Whether message contains media
• label_spam (boolean) - Spam label (for evaluation)
• label_not_spam (boolean) - Ham label (for evaluation)

Example:

{"message_id": "tg_msg_001", "chat_id": "chat_12345", "user_id": "user_999", "created_at": "2025-01-15T10:00:00Z", "text": "Example message text", "language": "ru", "message_type": "text", "has_media": false, "label_spam": true, "label_not_spam": false}

🔍 What to Expect

After running PoC, you'll get:

1. Patterns discovered - Semantic and deterministic patterns found in your data
2. Rules generated - SQL-like rules for each pattern
3. Metrics - Precision, recall, coverage, ham hit rate
4. Report - Human-readable Markdown report with all findings

📚 Related Documentation

• Data Contract - Detailed field specifications
• PoC Plan - Step-by-step PoC guide
• Overview - High-level integration overview

💡 Tips

• Start with the sample data to verify your setup
• Use small datasets (100-1000 messages) for initial PoC
• Ensure your data matches the expected schema
• Review the generated report carefully before scaling up
• Adjust configuration in config/config.yaml based on your needs