Key points while writting prompt for AI

 

1️⃣ Be Clear & Specific

Don’t be vague. Say exactly what you want.

❌ “Extract details from this text”
✅ “Extract vehicle registration number, make, and damage details from the text and return JSON”

2️⃣ Define the Role of the AI

Tell the AI who it should act as.

✅ “You are a police incident data extractor”
✅ “Act as a senior Python developer reviewing extraction logic”

This sets context and improves accuracy.

3️⃣ Specify the Output Format

Always mention how the response should look.

  • JSON / Table / Bullet points

  • Field names

  • Yes/No instead of true/false

e.g: Return the output in strict JSON with no extra text.

4️⃣ Add Rules & Constraints

Very important for consistency.

  • What must be included

  • What must not be done

  • Edge cases

Example:

  • Do NOT merge vehicles

  • Capture vehicles even if repeated in different sections

5️⃣ Handle Edge Cases Explicitly

AI guesses if you don’t guide it.

✅ “If registration number is missing, set it to null”
✅ “If vehicle is mentioned indirectly, still extract it”

6️⃣ Use Step-by-Step Instructions

Helps reasoning and reduces mistakes.

Example:

Step 1: Identify all vehicle mentions  
Step 2: Validate registration number  
Step 3: Create separate entries for each occurrence

7️⃣ Provide Examples (Very Powerful 🔥)
Few-shot prompting improves accuracy a lot.
  • Input text
  • Expected output
Even one example can dramatically improve results.
8️⃣ Avoid Ambiguous Language
Words like maybe, probably, relevant confuse AI.

❌ “Extract important info”

✅ “Extract registration_number, make, and vehicle_description”




9️⃣ Control Verbosity
Tell the AI how much to talk.
  • “No explanation”
  • “Only output JSON”
  • “Explain reasoning briefly”
🔟 Test & Iterate
Prompt writing is iterative.
  • Test with real edge cases
  • Refine rules
  • Tighten wording

Comments

Post a Comment

Popular posts from this blog

How Create API in python using flask library

 from flask import Flask, request, jsonify app = Flask(__name__) # In-memory database items = {} # Get all items @app.route('/items', methods=['GET']) def get_items():     return jsonify(list(items.values())), 200 # Get a single item by ID @app.route('/items/<int:item_id>', methods=['GET']) def get_item(item_id):     item = items.get(item_id)     if item:         return jsonify(item), 200     return jsonify({'error': 'Item not found'}), 404 # Create a new item @app.route('/items', methods=['POST']) def create_item():     data = request.json     item_id = len(items) + 1     items[item_id] = {'id': item_id, 'name': data.get('name'), 'description': data.get('description')}     return jsonify(items[item_id]), 201 # Update an existing item @app.route('/items/<int:item_id>', methods=['PUT']) def update_item(item_id):     if item_id not in items:       ...
Read more

Book teams appointment with email notification

 using System; using System.Net; using System.Net.Mail; using System.Text; class Program {     static void Main()     {         string teamsLink = "https://teams.microsoft.com/l/meetup-join/xxxxxxxx";         DateTime startUtc = new DateTime(2026, 1, 10, 10, 0, 0, DateTimeKind.Utc);         DateTime endUtc = startUtc.AddMinutes(30);         string ics = $@" BEGIN:VCALENDAR PRODID:-//YourApp//EN VERSION:2.0 METHOD:REQUEST BEGIN:VEVENT UID:{Guid.NewGuid()} DTSTAMP:{DateTime.UtcNow:yyyyMMddTHHmmssZ} DTSTART:{startUtc:yyyyMMddTHHmmssZ} DTEND:{endUtc:yyyyMMddTHHmmssZ} SUMMARY:Project Discussion DESCRIPTION:Join the Teams meeting:\n\n{teamsLink} LOCATION:Microsoft Teams Meeting ORGANIZER;CN=Organizer:mailto:organizer@example.com ATTENDEE;ROLE=REQ-PARTICIPANT:mailto:user@example.com STATUS:CONFIRMED URL:{teamsLink} END:VEVENT END:VCALENDAR";         var mail = new MailMe...
Read more