Automation · Artificial Intelligence · Large Language Models · No-Code · Make.com · Ollama · Google Docs · JavaScript
Turn any topic into a complete structured learning guide — automatically — using a locally hosted open-source LLM. Zero paid API costs. Runs entirely on your own machine.
Live demo: pathaks.co.in Built by: Dev Pathak — B.Tech CSBS, Jain University Bangalore
Type any topic. The pipeline runs it through 7 evidence-based learning steps and returns a complete guide in ~20 seconds.
| Step | Output Generated |
|---|---|
| 1 | Beginner-friendly overview |
| 2 | Reliable sources to study |
| 3 | Deep-thinking questions |
| 4 | Practical exercises |
| 5 | Common misconceptions |
| 6 | Spaced-repetition study plan |
| 7 | Self-test quiz |
Result: delivered instantly to browser + saved to Google Docs.
User Input (pathaks.co.in)
|
v
Make.com Webhook
|
v
8 Parallel HTTP Calls
→ Ollama (qwen2.5:0.5b) — running locally
→ Exposed via ngrok tunnel
Steps 1–7 run in parallel + Step 8 compiles output
|
v
[Two outputs]
Webhook Response Google Docs
(instant to frontend) (permanent copy)
Every one of the 8 calls runs against Ollama on your own machine — zero per-token cost, no API key management, no rate limits from a provider.
ngrok exposes localhost:YOUR_MAKE_WEBHOOK_URL to Make.com's cloud, making the local model reachable from any workflow.
| Layer | Tool | Role |
|---|---|---|
| Automation / Orchestration | Make.com | Fans out 8 parallel LLM calls |
| Large Language Model | Ollama (qwen2.5:0.5b) | Local inference — no API cost |
| Tunnel | ngrok | Exposes localhost to Make.com |
| Storage | Google Docs via Make.com | Permanent guide archive |
| Frontend | HTML / CSS / JavaScript | User interface |
| Hosting | Netlify | Static site deployment |
| Domain | pathaks.co.in | Live production URL |
ai-research-pipeline/
├── README.md — Project overview (you are here)
├── index.html — Frontend — live at pathaks.co.in
├── make-blueprint.json — Import directly into Make.com
├── setup.bat — Starts Ollama + ngrok together
└── .env.example — Configuration template
- Ollama installed (ollama.ai)
- ngrok account — free tier works (ngrok.com)
- Make.com account — free tier works (make.com)
- Google account for Docs storage
Step 1 — Pull the model
ollama pull qwen2.5:0.5bStep 2 — Start services Run setup.bat — opens Ollama and ngrok in separate windows
Step 3 — Get ngrok URL Copy the HTTPS forwarding URL from the ngrok window (format: YOUR_MAKE_NGROK_URL)
Step 4 — Import Make.com blueprint New scenario → menu → Import Blueprint → select make-blueprint.json → update all 8 HTTP module URLs to: your-ngrok-url/api/chat
Step 5 — Connect Google Docs Reconnect the Google Docs module to your Google account Set your destination folder ID
Step 6 — Configure frontend Replace YOUR_MAKE_WEBHOOK_URL in index.html with your Make.com scenario webhook URL
Step 7 — Deploy and test Open index.html locally or deploy to Netlify Turn scenario ON → type a topic → submit
- Frontend POSTs topic as JSON to Make.com webhook
- Make.com fans out 7 parallel prompts to local Ollama — each prompt targets one learning step
- 8th call compiles all outputs into one formatted guide
- Guide returned synchronously to frontend via Webhook Respond module
- Simultaneously saved to Google Docs automatically
- Frontend caches last guide in localStorage — guide survives page refresh
- Multi-step prompt engineering for consistent cross-topic output
- Parallel API call orchestration in Make.com
- Local LLM deployment and tunneling via ngrok
- Full-stack JavaScript — custom markdown renderer, localStorage caching, async fetch with error handling
- Production deployment on custom domain via Netlify
- No-code automation architecture design
- qwen2.5:0.5b chosen for speed on modest hardware Swap to qwen2.5:7b or llama3.1:8b for better quality (one-line change per HTTP module)
- ngrok free tier URL changes on every restart Update Make.com HTTP modules each time Or use paid ngrok plan with static domain
- Google Docs auth tied to original account Reconnect after importing blueprint
- AI Email Automation Pipeline Gmail → Gemini AI → Google Sheets → Telegram github.com/Devpathak18/email-ai-automation
Dev Pathak B.Tech Computer Science and Business Systems (CSBS) Jain Deemed-to-be University, Bangalore Program co-designed with Tata Consultancy Services
GitHub: github.com/Devpathak18 LinkedIn: linkedin.com/in/devpathak18 Email: devpathakpersonal@gmail.com Portfolio: pathaks.co.in
Open to remote internships in AI Automation, Data Analytics, and Business Analysis.