Full stack developer | Building scalable applications.

youtube

github

Shubhojeet Bera • Peerlist

<h3>Fieldcraft Overview</h3><p>Fieldcraft is a local-first Chrome extension designed to streamline the job application process. It intelligently reads job pages, evaluates how well a role fits your saved profile, researches the company, and drafts truthful answers for your review before you submit them.</p><ul><li><p><strong>Local-First Approach:</strong> All your profile data and API keys remain securely within your browser, with no server-side storage.</p></li><li><p><strong>Intelligent Evaluation:</strong> The extension scores the fit of a job role against your profile, identifies gaps, and provides a concise company brief.</p></li><li><p><strong>Drafting Assistance:</strong> It drafts answers for form fields based on your profile and research, which you can review and edit before insertion.</p></li><li><p><strong>User Control:</strong> Explicit user approval is required before any drafted answer is inserted, ensuring you remain in control of submissions.</p></li><li><p><strong>Seamless Integration:</strong> Operates as a Chrome side-panel extension directly within the job page, eliminating the need for a separate dashboard.</p></li></ul><h3>Technical Implementation</h3><p>The product is built as a Chrome MV3 side-panel extension with a companion Next.js marketing site. The extension utilizes a Vite + CRXJS build, a React side panel, and background scripts for content injection. The marketing site features GSAP animations, semantic HTML, and optimized metadata for search engines.</p><h3>Impact</h3><ul><li><p><strong>Time Savings:</strong> Cuts per-application time by approximately 55–65% through automation of reading, parsing, scoring, and drafting.</p></li><li><p><strong>Reduced Context Switching:</strong> Decreases research and tab switching by about 70–80% by providing key information within the side panel.</p></li></ul>

Fieldcraft

<h3>Inkdown is a markdown editor and sharing platform designed for efficient document management and collaboration.</h3><ul><li><p><strong>Live Preview:</strong> See your markdown changes in real-time as you type, with scroll position syncing for seamless navigation in long documents.</p></li><li><p><strong>Folder Organization:</strong> Organize your documents with a nested folder structure, making it easy to manage large knowledge bases.</p></li><li><p><strong>GitHub Flavored Markdown (GFM) Support:</strong> Enjoy full support for GFM, including tables, task lists, code blocks, and strikethrough.</p></li><li><p><strong>Private-by-Default Documents:</strong> All documents are private by default, ensuring your content remains secure until you choose to share it.</p></li><li><p><strong>One-Click Sharing:</strong> Easily share your documents with unique, private links with just one click.</p></li><li><p><strong>Custom Markdown Parser:</strong> Built with a custom parser for enhanced GFM compatibility and performance.</p></li><li><p><strong>Real-time Sync:</strong> The editor and preview stay in sync, mapping editor scroll position to preview position for an intuitive editing experience.</p></li><li><p><strong>Public Sharing with Access Tracking:</strong> Generate shareable links with built-in view count and last viewed tracking.</p></li><li><p><strong>Migrated Stack:</strong> Leverages Neon (serverless Postgres), Better Auth, Cloudflare R2 for storage, and Resend for transactional emails.</p></li></ul><p>Inkdown enhances productivity by replacing flat document lists with an intuitive folder hierarchy and keeping long documents readable through synchronized editing and preview views. Its private-by-default approach combined with one-click publishing offers flexibility for both personal notes and public documentation.</p>

Inkdown

<h3>Overview</h3><p>Agentic Chat is an AI workspace that combines document search, long-term memory, web research, and OAuth-connected apps so it can answer with better context and take real action on your behalf.</p><p>The system intelligently decides what context to pull in before the model sees a message, including past conversation memory, relevant document chunks, live web searches, or URL content, and assembles it into the prompt automatically.</p><p>A LangGraph orchestrator manages the process, running a planner, an agent, and a tool execution loop. The agent can connect to various applications like Gmail, Google Calendar, Drive, Docs, Sheets, Slack, Notion, GitHub, and Linear through Composio. Crucially, any action that writes, sends, or deletes data requires explicit user approval before execution.</p><p>The technology stack includes Next.js 16, React 19, PostgreSQL with pgvector for vector search, Prisma, Better Auth for Google sign-in, and UploadThing for file storage. Users have the option to bring their own OpenAI key, which is encrypted at rest and never sent to the client. LangSmith is used to trace every step of the graph.</p><h3>How It Was Built</h3><ul><li>Developed the LangGraph orchestrator from scratch, featuring a planner node for request breakdown, a supervisor agent for decision-making, and a tool node for execution, all within a loop with PostgreSQL checkpointing for conversation resumption after approvals.</li><li>Built an end-to-end document pipeline: file upload via UploadThing, parsing for various file types (PDF, DOCX, XLS, CSV), 800-token semantic chunking, OpenAI embeddings stored in pgvector, hybrid retrieval, and Cohere reranking for relevant chunk surfacing.</li><li>Integrated Composio for dynamic loading of user-connected tools based on OAuth accounts, including routes for connection, callbacks, and status, along with a dangerous-action allowlist and verb-based detection for write operations requiring approval.</li><li>Implemented the BYOK (Bring Your Own Key) flow with AES-256-GCM encryption for user API keys, and a server-side proxy for secure decryption and forwarding to OpenAI.</li><li>Integrated SSE streaming from LangGraph's event stream to the client for real-time updates on token chunks, tool progress, approval requests, and planning steps, with full abort support.</li></ul><h3>Impact</h3><ul><li>A semantic cache with pgvector similarity search efficiently reuses prior answers for repeated or similar queries, reducing redundant embedding and generation calls.</li><li>Composio streamlined OAuth integration, replacing per-service plumbing with a single integration covering 9 toolkits and 94 actions.</li><li>Approval cards provide a critical safety layer, gating all write/send/delete actions proposed by the agent for explicit user confirmation.</li><li>BYOK enables teams to leverage their existing OpenAI contracts without sharing plaintext keys.</li></ul><h3>Highlights</h3><ul><li>The LangGraph loop with PostgreSQL checkpointing ensures conversation state is maintained across approval interrupts, allowing the agent to resume seamlessly.</li><li>A three-tier web scraping mechanism (Readability → Firecrawl → Jina Reader) enables the agent to access content from most web pages, including those rendered with JavaScript.</li><li>A two-layer safety system for Composio actions, including an explicit allowlist and a verb detector, ensures secure operation.</li></ul><h3>Tech Stack</h3><ul><li>Next.js 16</li><li>React 19</li><li>TypeScript 6</li><li>LangGraph 1.3.2</li><li>LangChain</li><li>LangSmith</li><li>OpenAI SDK v6</li><li>Composio</li><li>Exa Search</li><li>Firecrawl</li><li>Cohere Reranking</li><li>Mem0</li><li>PostgreSQL + pgvector</li><li>Prisma v6</li><li>Better Auth</li><li>RAG</li><li>BYOK (AES-256-GCM)</li><li>Tailwind CSS 4</li></ul>

Agentic Chat

<h3>Overview</h3><ul><li><p>Edward was built as a place to actually make software, not just chat about it. A user can describe an app in plain language, generate code, inspect and edit files, run it in an isolated environment, publish a live preview, and sync the result back to GitHub without leaving the product.</p></li><li><p>The system is made up of a Next.js web app for chat and UI, an Express API for managing runs and previews, shared packages for auth and data access, and Docker-based environments for running generated apps.</p></li><li><p>Behind the scenes, it uses Postgres for product data, Redis queues for long-running work, Docker for isolated runs, and AWS plus Cloudflare for shareable previews. That makes the output something you can actually use and build on, not just model text in a chat window.</p></li></ul><h3>How It Was Built</h3><p>The main technical choices behind the product, from system design to the parts that make it work day to day.</p><ul><li><p>Split the product into a web app, an API, background workers, and Docker-based run environments so each part has a clear job.</p></li><li><p>Built the full flow from prompt to files to running app. Users can inspect code, edit it, see logs, preview the result, and sync it to GitHub.</p></li><li><p>Used Redis queues for slower work like builds, sandbox startup, and preview setup so the main product stays responsive.</p></li><li><p>Set up preview publishing through AWS and Cloudflare so users get stable preview links instead of a temporary local run.</p></li></ul><p>Can read more <a target="_blank" rel="noopener noreferrer nofollow ugc" class="break-all link-break text-gray-500 hover:underline" href="https://blog.shubhojeet.com/edward">here</a></p><h3>Impact</h3><ul><li><p>Turned prompt-to-application into a single workflow: generate, edit, run, preview, and sync back without switching tools.</p></li><li><p>Removed blocking build and preview steps from the chat loop by handling them asynchronously, so the conversation stays responsive.</p></li><li><p>Kept the output as a real repo with shareable preview links, so generated work can be reviewed and continued in normal engineering workflows.</p></li></ul><h3>Highlights</h3><ul><li><p>Built around execution, not just generation.</p></li><li><p>Combines product design with the backend systems needed to actually run apps.</p></li><li><p>Hands off cleanly into normal code review and engineering workflows.</p></li></ul><h3>Tech Stack</h3><ul><li><p>Next.js 16</p></li><li><p>Express.js</p></li><li><p>BullMQ</p></li><li><p>Docker Sandboxes</p></li><li><p>OpenAI &amp; Gemini</p></li><li><p>PostgreSQL</p></li><li><p>Redis</p></li><li><p>GitHub Sync</p></li><li><p>AWS S3 + CloudFront</p></li><li><p>Cloudflare Preview Routing</p></li></ul>