Excited to share a recent side project — a mini CodeRabbit clone built using Python, FastAPI, LangChain, and LangGraph. live: https://lnkd.in/gSP_sxec github: https://lnkd.in/gBpRDP63 Key highlights: Clones GitHub PRs Parses code using Tree-sitter to build ASTs and semantic graphs with networkx Enriches context for LLM-driven analysis Uploads processed data to S3 Triggers multi-agent AI code reviews via Redis/RQ The goal was to explore how AI agents can collaborate to review pull requests intelligently by combining static analysis with contextual reasoning. This project reflects my ongoing interest in developer tools, AI automation, and building practical, self-hostable AI review systems. #Python #FastAPI #LangChain #LangGraph #AI #DevTools #CodeReview #LLM #NEXTJS

How Pydantic AI Turned My Chaotic Data Into a Super‑Smart Python Model 🤖 I was juggling a legacy API that returned nested JSON like a tangled ball of yarn—lists inside dicts, optional fields, and a few hidden “type‑mismatch” bugs that broke the whole pipeline. Every time I wrote a new class, I added manual checks, and the codebase grew into a nightmare of try/except blocks. Enter Pydantic AI. I fed it a single example payload, and it instantly generated a hierarchy of BaseModel classes with proper type hints, default values, and validators for the edge cases. The next day, the same API response passed through the model without a single runtime error, and the auto‑generated docs showed exactly what each field meant. Adding a new optional field? Just update the example and let Pydantic AI regenerate—no more hand‑rolled parsing logic. Now my services serialize, deserialize, and validate data in one line, and the code reads like a story instead of a maze. #Python #PydanticAI #DataValidation

Why I Use TypedDict Instead of a Regular Dictionary in LangGraph (and Python) When working with LangGraph, we often pass data between nodes — inputs, outputs, and intermediate states. At first glance, a plain dict seems fine: flexible, simple, and quick to write. But that flexibility comes at a cost — lack of structure. Here’s what happens when you use a normal dict: You might mistype a key name ("usre_id" instead of "user_id") — no warnings. You might forget which fields a node expects — no hints. You only find out about issues at runtime, not during development. That’s where TypedDict shines. 💡 Now your editor (and LangGraph’s type system) can: 1. Warn you about missing or extra fields 2. Autocomplete field names 3. Make refactoring much safer 4. Keep large graphs consistent as they grow In short — TypedDict makes your LangGraph pipelines self-documenting and type-safe, reducing bugs before they even happen. If your LangGraph is getting complex, switching to TypedDicts is one of those small changes that pay off big in the long run. #LangGraph #LangChain #Python #AI #SoftwareEngineering #AIEngineering #GenAI #CleanCode #DeveloperTips #LLM

🎉 New little-scripts drop - DeepSeek-OCR! I’ve just added a standalone DeepSeek-OCR FastAPI service to my little-scripts repo; bundled with a lean Docker Compose setup for super-easy spin-up. 🚀 If you’re tinkering with OCR or need a quick API you can self-host, this might save you some time. 👇 Check it out: https://lnkd.in/eKCPB6DW Found it useful? A ⭐ and/or feedback is always welcome! 🙏 #OpenSource #OCR #DeepSeekOCR #FastAPI #Docker #Python #AI #MLOps

🌟 Day 16 of my AI Fullstack Journey 🌟 Today I explored Constructors, Instance Methods, and Class Methods in Python 🐍 Here’s what I learned 👇 🔹 __init__() is the constructor — it initializes object attributes automatically when an object is created. 🔹 self is used to access instance variables (unique to each object). 🔹 @classmethod with cls helps modify class-level attributes shared by all objects. 🔹 Methods like deposit() and withdraw() define object behavior in a clean, reusable way. 🛠️ Mini Project of the Day: 👉 Bank Account Management System ✔ Created a class BankAccount with features like: Deposit & Withdraw methods Display account info Change bank name using a class method This project helped me understand the power of OOP — how data and behavior are encapsulated within classes. 💡 Next, I’m moving to Inheritance (Day 17) — one of the most exciting OOP concepts! #Python #OOP #AI #FullStack #LearningJourney #engineeringinkannada

🚀 Just launched my first Python package — coding-agent-python! 🐍 An AI-powered coding assistant that can generate, read, write, and even run Python code — all from a simple natural language command. Just type 👇 > coding-agent "make a file for writing the roots of quadratic equation" — powered by Gemini AI 🤖 (Add your GEMINI_API_KEY in the environment) 💡 It can even generate LeetCode solutions by giving inputs and outputs ! This was my first deep dive into Python packaging, learning the role of core files like __init__.py, pyproject.toml, LICENSE, mkdocs.yml, and README.md while exploring: 🧩 uv by Astral — for building & publishing 📘 mkdocs — for simple documentation 📦 PyPI — for sharing with the world 🌍 Open source is the best way to learn — contributions are welcome! 🔗 GitHub: https://lnkd.in/gju9xqXe 🔗 Docs: https://lnkd.in/gejgdf32 🔗 PyPI: https://lnkd.in/gvdkCSFb #Python #OpenSource #AI #Gemini #DevTools #PyPI #Astral #mkdocs #uv

Excited to share my latest project! I built a webpage summarization tool using Python and LLMs that can automatically condense website content into clear, professional summaries. Here’s what it does: Fetches content from any URL using a custom scraper. Sends the content to a local LLM (llama3.2 via Ollama) for summarization. Outputs a clean, readable summary in Markdown for easy reading. The workflow leverages: Python for scripting and automation Ollama and LLaMA 3.2 for local AI-powered summaries This tool is perfect for quickly digesting large amounts of web content, whether for research, competitive analysis, or personal learning. Check out the code here: https://lnkd.in/etKVFH_A #AI #MachineLearning #Python #LLM #WebScraping #ProductivityTools #Ollama #LLaMA3

💡 Today’s Learning: Binary Search in a Rotated Sorted Array As part of strengthening my problem-solving and algorithmic thinking, I explored how to efficiently search for an element in a rotated sorted array — a variation of the classic Binary Search problem. 🧠 Key Takeaways: 🔹 Even though the array is rotated, one half (left or right) always remains sorted. 🔹 By identifying the sorted half in each iteration, we can decide where to continue searching. 🔹 This approach retains the efficiency of Binary Search — achieving O(log n) time complexity and O(1) space complexity. 📘 I’ve implemented the optimized Python solution and added it to my GitHub repository: https://lnkd.in/geZUM25R This kind of practice not only sharpens my problem-solving and analytical thinking, but also helps me write clean, efficient, and scalable code — skills I continue to build as I grow in the data and AI space. #Python #LeetCode #ProblemSolving #BinarySearch #Algorithms #DataStructures #LearningJourney #GitHub #TechSkills #SoftwareDevelopment

⚡ Just published my first Python package https://lnkd.in/dTe9VvAJ TOON (Token-Oriented Object Notation) is a lightweight format that represents JSON data in a more compact, LLM-friendly way reducing tokens and making structured prompts cleaner to read. This library lets you easily convert: from json_toon import json_to_toon, toon_to_json ✅ JSON → TOON ✅ TOON → JSON Why it matters: Large-language-model prompts often waste tokens repeating JSON keys. TOON trims that overhead while keeping data human-readable and reversible. Give it a try, star it if you like the idea, and let me know what you’d build with it 💡 #Python #AI #OpenSource #LLM #DeveloperTools