DEV Community

Cover image for 🌿 Herbal Remedy Advisor – Grandma's Wisdom Meets LLMs
Avradeep Nayak
Avradeep Nayak

Posted on

🌿 Herbal Remedy Advisor – Grandma's Wisdom Meets LLMs

“Because your grandma’s tea deserves LLM-level respect.”

A few weeks ago, while sipping ginger tea during a coding session (thanks, Grandma!), a curious thought struck me:
What if ancient herbal remedies could be queried like ChatGPT?
What if we could combine AI, semantic search, and knowledge graphs to revive traditional wisdom in a modern, developer-friendly way?

That's how Herbal Remedy Advisor was born. 💡

🔮 Meet the App
Herbal Remedy Advisor is an AI-powered herbal medicine search engine. It's like if ChatGPT trained with your grandma and also learned SQL.

It lets you:

🧠 Ask questions like “what helps with a sore throat?” and get meaningful, filtered results.

🌿 Browse a full knowledge base of natural remedies with safety and usage info.

➕ Add your own remedies—because healing wisdom shouldn’t retire.

🤖 Chat with a helpful agent powered by Gemini and Ollama, trained on herbal context.

⚡ Enjoy fast semantic queries with vector-powered SQL magic via MindsDB.

🧠 Under the Hood
I didn’t want to just throw another Flask app into the wild. I wanted this to be smooth, fast, and hackable.

Stack Highlights:
Layer What I Used
LLM Agent gemini-2.0-flash
Embeddings deepseek-r1:1.5b via Ollama
AI Database MindsDB + native Knowledge Base
Backend Flask + Jinja2
UI Bootstrap 5 (quick and clean)
Dev Tooling uv (because pip deserves better)

💻 Dev Magic – Fast Setup
I wanted the setup to be beginner-friendly but still "cool dev-approved".

bash
Copy
Edit

clone the repo

git clone https://github.com/Zedoman/Herbal.git
cd Herbal

install with uv

uv venv
uv pip install . # or compile with pyproject.toml

run MindsDB and Ollama locally

docker run -p 47334:47334 mindsdb/mindsdb
ollama run deepseek-r1:1.5b
Then just run the Flask app and boom — you’re in herbal heaven.

🌱 Features I Loved Building
🔍 Semantic Search via SQL — semantic_search('cold remedy', content) — yep, it's a real thing.

🛡️ Safety filters — because not everything natural is safe for everyone.

🤖 Agent mode — ask about pregnancy-safe remedies, and it checks context from the KB.

📦 Auto init — first run sets up everything: knowledge base, LLM engine, sample data.

📸 A Peek into the UI

Herbal Remedy Screenshot

Responsive cards, clear safety info, and minimal fuss.

🧪 SQL, but Cool
Want to find a remedy that helps with "headache", is marked safe, and feels semantic?

sql
Copy
Edit
SELECT *
FROM herbal_remedy_kb
WHERE semantic_search('headache relief', content)
AND symptom = 'Headache'
AND safety LIKE '%Safe%'
LIMIT 20;
LLM power, SQL-style. 😎

🙏 Shoutouts
MindsDB – ML meets SQL without the drama.

Ollama – Local models that just work.

uv – My new favorite Python package manager.

🚀 What’s Next?
Add user accounts and favorites

More detailed interaction metadata (e.g., drug interactions)

Support for Ayurveda & TCM

Maybe even turn this into a mobile app?

🧝‍♂️ Final Thought
If you're into AI, dev tooling, or you’ve ever been cured by a cup of clove tea—
you’ll enjoy building on this. 🌿

Check it out on GitHub →
🔗 github.com/Zedoman/Herbal

Let me know your thoughts, feature ideas, or which remedy you’d love to see next!

Top comments (0)