Software Engineer, JavaScript, Linux, React JS, React Native, Node JS, Deno and open source software enthusiast.

github

projects

hashnode

devto

medium

Keshav Ashiya  • Peerlist

Local-first CLI code intelligence tool with LangChain-powered RAG offers a robust solution for developers seeking to enhance their codebase management and search capabilities. This tool indexes your codebase and enables semantic code search using advanced AI technologies. With everything running locally on your machine via Ollama, it ensures privacy and speed without the need for cloud services.<h3>Key Features</h3><ul><li>Semantic Code Search - Utilize natural language queries to find code efficiently.</li><li>AI-Powered Suggestions - Receive intelligent code recommendations directly from your codebase.</li><li>100% Local - Operates entirely on your machine, ensuring data privacy and security.</li><li>Fast Indexing - Supports incremental updates, re-indexing only changed files for efficiency.</li><li>Python Support - Comprehensive Python AST parsing for detailed code analysis.</li></ul>CodeSage is designed for developers who prioritize local-first solutions, offering a seamless integration with Python environments and optional support for OpenAI and Anthropic models. The tool is open source, licensed under the MIT license, and can be easily installed via PyPI or directly from the source.

codesage

Your personal research assistant that remembers everything you've ever read. Docify is an open-source, local-first AI application designed to enhance your research capabilities. This innovative tool allows you to upload various resources such as PDFs, URLs, documents, images, and code, and ask questions about them to receive cited, grounded answers. All processing is done locally, ensuring complete privacy of your data.<h3>Key Features</h3><ul><li>Privacy-First: All processing, including embeddings and storage, happens locally.</li><li>Smart Deduplication: Prevents duplicate processing through content-based fingerprinting.</li><li>Multi-Format Support: Supports PDF, URL, Word, Excel, Markdown, images (OCR), and code.</li><li>Cited Answers: Every response includes citations to source documents.</li><li>Hybrid Search: Combines semantic (vector) and keyword (BM25) search.</li><li>Local LLM: Runs Mistral 7B via Ollama with optional cloud LLM support.</li><li>Workspace Model: Offers personal, team, or hybrid collaboration.</li><li>One-Command Setup: Simplified setup with Docker Compose orchestration.</li></ul><h3>Architecture Overview</h3>Docify's RAG pipeline integrates 11 core services, including resource ingestion, chunking, embeddings, query expansion, hybrid search, re-ranking, context assembly, prompt engineering, LLM service, citation verification, and message generation. For complete technical details, refer to the <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="/keshavashiya/docify/blob/main/ARCHITECTURE.md">ARCHITECTURE.md</a>.

Docify

<h3>Introduction</h3>In the age of information overload, we’re all voracious readers, collectors of bookmarks, and lifelong learners. But if you’re anything like me, you’ve probably faced this: you save a fantastic article on <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="http://dev.to">dev.to</a>, a must-read on <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="http://daily.dev">daily.dev</a>, and a handful of gems elsewhere—only to lose track of them when you need them most. The result? A scattered digital trail and a sense of missed opportunity.That’s where the idea for my open source Reading List project was born: a single, reliable place to automate, curate, and share everything I’m reading—across platforms, in real time, and with the world.<h3>The Spark: Solving a Real Problem</h3>The inspiration was simple but powerful: I wanted a way to find my bookmarks from different platforms, all in one place, whenever I needed them. But I also wanted more. What if this reading list could be public—a living portfolio of my learning journey, a way to show the world what I'm reading, and maybe even inspire others?This project isn't just about personal productivity. It's about storytelling through reading making your learning journey visible, discoverable, and shareable.<h3>The "Now Page" Philosophy</h3>This project draws inspiration from Derek Sivers' concept of the "now page"—a simple, public declaration of what you're currently focused on. Instead of a static "about me," a "now" page answers: What are you working on right now?My reading list applies this to learning: What are you reading right now? It’s a living, breathing snapshot of your current intellectual journey—not what you read last year, but what you’re actively engaging with today. The beauty of this approach is its authenticity. It’s not curated for perfection; it’s real, current, and honest about where your attention is actually going.<h3>From Idea to Architecture</h3>The core challenge was integrating multiple data sources—<a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="http://dev.to">dev.to</a>, <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="http://daily.dev">daily.dev</a>, and potentially more—each with their own APIs and formats. I wanted a solution that would fetch and update my reading list automatically, without manual intervention, and make it public for anyone to see.The answer was to use GitHub Actions as the orchestrator. On a schedule, it fetches data from all sources, normalizes it, and prepares it for publishing. The data is stored as simple JSON, which is then bundled with the site and deployed to GitHub Pages. This means my reading list is always up to date, always available, and always at the same universal path: `{username}.github.io/readinglist`.By keeping everything automated and using GitHub’s infrastructure, the project is both reliable and easy for anyone to fork and adapt. No server maintenance, no manual updates—just a living record of what I’m reading, always fresh.<h3>Roadmap</h3>Here’s what’s next for the project:1. More Data Sources - Integrate with additional platforms like Pocket, Medium, and Twitter bookmarks.2. Access to Browser Bookmarks - Allow users to import or sync bookmarks directly from their browser, making the reading list even more comprehensive.3. Enhanced Filtering - Add advanced filtering and search, so users can quickly find articles by topic, source, or reading time.<h3>Conclusion: A Universal Reading Path</h3>What started as a solution to my own bookmark chaos has become something bigger—a platform for making learning journeys visible and shareable. The beauty of this project is its universal accessibility. Just like the "now page" philosophy, your reading list will be available at a universal path: `{username}.github.io/readinglist`. This consistent URL structure makes it easy for others to discover and follow your learning journey, creating a network of shared knowledge and inspiration.Check out the live demo: <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="https://keshavashiya.github.io/readinglist">https://keshavashiya.github.io/readinglist</a>Explore the source code: <a target="_blank" rel="noopener noreferrer" class="break-all link-break text-gray-500 hover:underline" href="https://github.com/keshavashiya/readinglist">https://github.com/keshavashiya/readinglist</a>If this resonates with you, fork it, customize it, and share your own reading journey. Let’s make learning—and sharing—more visible, more connected, and more meaningful.

Open Source Reading List

Fetch Data from Twitter using Twitter API through Hadoop tool Flume and store it into Mysql. Then manipulate data into Human readable format and print it from different tables. 

Twitter Data Analysis using Hadoop 

Web app 

E-commerce Web app coded in ruby on rails. 

BUILDmyPC

Multi-level campus ambassador program -  Seed Microsoft's new AI chatbot -  Ruuh

Fetch Data from Twitter using Twitter API through Hadoop tool Flume and store it into Mysql. Then manipulate data into Human readable format and print it from different tables.