Mohammed Razi Kallai Logo
Back to articles

Razi.pro Agentic Unified Chatbot: RAG‑Powered Automation + Tool Combos

How to use Razi.pro’s agentic, RAG‑powered unified chatbot to orchestrate tools, automate workflows, and get intent‑aligned results—with lots of real examples.

14 min read
#Agentic AI#RAG#Chatbot#Automation#Developer Tools#Next.js

Razi.pro Agentic Unified Chatbot: RAG‑Powered Automation + Tool Combos

Why this matters

Most chatbots answer questions. The Razi.pro chatbot executes. It’s an agentic, RAG‑powered system that can understand intent, fetch your context, and actively use on‑site tools to deliver tangible outcomes — not just text. That means one chat can: compress your files, generate images, draft emails, format JSON, create event assets, analyze data, and even post feedback — end‑to‑end.

This guide shows how to get the most out of the unified chatbot, how Retrieval‑Augmented Generation (RAG) keeps answers grounded in real project knowledge, and how tool chaining turns a single request into a finished result.

What “Agentic Unified Chatbot” means

An agentic chatbot doesn’t stop at answers — it runs actions. On Razi.pro, the chatbot can call built‑in tools, pass outputs between them, and continue until your goal is met.

Core capabilities:

  • Tool orchestration: The bot can pick and run tools (e.g., image compression → upload → short link).
  • Multi‑step reasoning: It plans steps to achieve your outcome, not just reply once.
  • Context awareness: It uses your chat history, tool outputs, and site data.
  • Safe execution: Tools run client‑side when possible for speed and privacy.

RAG: Answers grounded in real context

Retrieval‑Augmented Generation (RAG) adds trustworthy context to responses. Instead of guessing, the bot retrieves relevant docs and project notes (like feature guides, tool references, or FAQs) and uses them to craft accurate, source‑linked answers.

Benefits:

  • Up‑to‑date knowledge tied to this site
  • Lower hallucinations and clearer citations
  • Answers that reflect real features and constraints

What the chatbot can do right now

Here are popular tools the agent can invoke from chat. You can also open them directly if you prefer a UI:

All of these are callable from the unified chat — the agent chooses and chains them for you.

One‑chat, multi‑tool examples (agentic runs)

Try these in the chatbot — it will pick the right tools and stitch results together.

  1. "Compress these 4 images to WebP under 200KB, upload them, and give me short links."
  • Tools: Image Compress → File Share → URL Shortener
  • Outcome: Optimized images + shareable short URLs
  1. "Extract text from this PDF, summarize it, and draft a polite email reply in a friendly tone."
  • Tools: OCR → Text Generation → Email Drafter
  • Outcome: Clean summary + ready‑to‑send email
  1. "Generate 3 hero images for a fintech landing page: minimal, blue palette, high contrast."
  • Tools: AI Image Generator
  • Outcome: Downloadable images in‑chat
  1. "Create an event called ‘AI Dev Meetup’, add agenda (9:00 Check‑in, 10:00 Keynote, 12:00 Lunch), create 5 setup tasks, and a poll about frameworks."
  • Tools: Event Manager (create event → bulk agenda → bulk tasks → poll)
  • Outcome: A ready event workspace with engagement tools
  1. "Humanize this marketing copy to sound natural, then format the JSON payload and generate a QR code for the signup link."
  • Tools: AI Humanizer → JSON Formatter → QR Code
  • Outcome: Polished copy + valid JSON + QR image
  1. "Decode this JWT, check expiration, and produce a clear summary table."
  • Tools: JWT Decoder
  • Outcome: Human‑readable breakdown + validity check
  1. "Create a SQL query to list paid users by month for the last 6 months for PostgreSQL."
  • Tools: SQL Builder
  • Outcome: Ready‑to‑run SQL with comments
  1. "Generate 20 fake users, export as CSV, and upload with a short link."
  • Tools: Fake Data → File Share → URL Shortener
  • Outcome: Downloadable CSV + short link

Creative tool combos you’ll love

  • Content sprint kit: Blog Outline → AI Humanizer → Email Drafter (newsletter) → QR for article
  • Design delivery: Image Generator → Image Compressor → File Share → Short Link for client handoff
  • Dev QA pack: Diff Checker → JSON Formatter → JWT Decoder → Structured bug report draft
  • Event engagement: Event Manager (Q&A + Polls + Feedback + Resources) → Analytics overview
  • Support workflow: OCR receipts → Summarize → Percent calculations → Email client with formatted results

Pro tip: Ask in goals, not tools — e.g., “Make these images portfolio‑ready and shareable under 1MB each” — and the agent will pick steps.

Prompt templates that work

Use these as starting points:

  • "Turn this AI‑sounding paragraph into warm, human copy (light rewrite)."
  • "Draft a professional reply apologizing for delay, friendly tone, 120–150 words."
  • "Create a SQL query in Postgres to get top 10 customers by revenue since July."
  • "Compare these two code snippets and highlight breaking changes only."
  • "Extract text from this image, remove noise, and summarize in bullet points."
  • "Compress this 45‑second video for web under 5MB, keep 720p."
  • "Create an event called ‘Dev Friday’, add 3 agenda items and a poll question."

How RAG + agentic tools stay reliable

  • Grounding: The bot retrieves on‑site docs (tool guides, FAQs, specs) before answering.
  • Deterministic tools: Converters, decoders, formatters, and encoders run locally where possible.
  • Structured results: Many tools return files, URLs, or formatted blocks the bot can reuse.
  • Privacy by default: Client‑side processing is favored; server routes only when necessary.

Where to start

Two easy ways:

  1. Use the floating chat button anywhere on the site to open the unified chatbot.
  2. Prefer a UI? Open tools directly from the Tools page and switch back to chat when you want the agent to chain actions for you.

You can also try the contact form automation to see a full end‑to‑end flow: Contact form.

FAQ

Does the chatbot work with files? Yes. Drop images, PDFs, or videos directly into chat. The agent will pick file‑aware tools like OCR, image/video compression, or upload + short link.

Can I combine multiple steps in one prompt? Absolutely. The agent will plan and execute a tool chain to reach your goal. If it needs clarification, it will ask.

Is it safe? Inputs are validated, and most processing happens in your browser. Sensitive operations are routed to secure API endpoints with sane limits.

Tell me what to build next

I’d love your feedback, bug reports, or suggestions. Drop a note here:

👉 Share feedback or issues

If you found this useful, please share it with a friend or teammate. It helps a lot! 🙌

About the Author

Mohammed Razi Kallai is a Senior Software Engineer with 10+ years of experience in full-stack development, cloud architecture, and AI automation.