subh05sus pushed a major update on main, adding a full-featured browser extension for tracking browsing activity (including tabs, searches, media, bookmarks and reading depth), introducing an AI-powered session processing API, and delivering a new chat dashboard.

subh05sus pushed a major update on main that implements a full RAG-powered chat system with new dedicated API routes, database-backed chat sessions, and enriched embeddings. The change also overhauls the browser extension to support multi-tab tracking, AI summarization of page content and interactions, and ships a revamped React UI with sidebar session history and management.

subh05sus rolled back the main branch to the previous extension state, undoing the recent v2 “autopilot” overhaul in background.js, content.js, popup UI, manifest.json, and styles.css. this revert restores the original configuration, script logic, and UI elements as they were before the update.

subh05sus pushed a new global CSS setup to main, integrating Tailwind with custom theme variables, utilities and animations. Their update tweaks primary and foreground color values in app/globals.css for a refined look and feel.

subh05sus pushed a commit to main that adds a new globals.css for base styles, theme variables, custom utilities, and enhanced form controls. They also adjusted the primary color from bright blue to black (with white text) for a more streamlined theme.

subh05sus pushed a foundational “feat” commit on main that scaffolds the entire app—adding dashboard, chat and settings pages, core API routes (chat, stats, suggestions), a browser extension, global UI styles, and updated package dependencies. Notable changes include extending shown chat sources from 5 to 8, wiring up real-time memory stats and AI suggestion endpoints, and enriching the UI with new icons, custom components, and theming.

subh05sus pushed a feature to implement RAG context retrieval with natural-language date parsing (via chrono-node) so you can filter embeddings by time ranges. They also introduced a revamped browser extension with authentication and status panels, updating the background and content scripts to handle batching, tab tracking, idle detection, and config synchronization.

subh05sus pushed a major update to main that adds an Autopilot mode via a browser extension—automatically capturing your browsing sessions and wiring up dashboard settings for configuration. They also cleaned house by removing a bunch of outdated or unnecessary documentation from the repo.

subh05sus pushed a new “autopilot” feature on the main branch, adding content scripts for interaction tracking, a background service worker to queue, filter and flush data, plus an options page for user controls. They also enhanced the backend API routes (via Prisma) to support both session- and bearer-token authentication, auto-create defaults in the database, and sync settings seamlessly between the extension and server.

subh05sus introduced a new Retrieval Augmented Generation (RAG) system, adding data ingestion endpoints, context retrieval in the chat API, and RAG-based chat integration. They also beefed up logging and telemetry across the server routes and browser extension to track requests, batching, and performance.

On main, subh05sus introduced a new autopilot tracking API that logs telemetry records and kicks off inline embedding processing via waitUntil (using processAutopilotDataEmbedding instead of the old enqueue queue). They also cleaned up by removing the standalone embeddings processor route and the old Vercel cron configuration.

subh05sus pushed a feature to main that implements a new page ingestion POST endpoint (app/api/pages/ingest/route.ts), enqueuing page embeddings for background cron processing. They also added the @vercel/functions dependency to package.json and updated the lockfile accordingly.

subh05sus pushed a feature update on main that adds new API endpoints for vector search, chat, embedding processing, and page ingestion. The commit also enhances logging throughout the tracking, chat, search, and embeddings pipelines (including Vercel cron triggers and fallback behaviors) and tightens up queueing and RAG context handling.

subh05sus pushed a new feature on main that implements a full Gemini embedding generation and processing pipeline—including a Redis-backed queue for saved pages and autopilot telemetry. They bumped the chat API model from gemini-1.5-flash to gemini-2.5-flash, standardized the embedding model name, added a check for the GOOGLE_API_KEY, and restructured the embedContent payload for better compatibility.

subh05sus pushed a change to main that replaces the old GOOGLE_GEMINI_API_KEY env var with GOOGLE_API_KEY across all documentation (README, SETUP, testing guides) and in the search API code for embeddings, ensuring naming consistency with the Google Generative AI integration.

subh05sus added a browser extension for page ingestion and overhauled the backend embedding services for both saved pages and autopilot telemetry. The update now passes page content and DOM snapshots directly into the embedding queues (avoiding refetching), enriches telemetry metadata, and even switches the extension’s default API URL to production with a version bump to 1.0.2.

subh05sus pushed a new feature to main implementing AI chat with RAG, embedding processing for pages and autopilot telemetry (with improved error logging, API-key checks, and a switch to the stable gemini-1.5-flash model). they also bumped the Cosmyc browser extension to v1.0.1, added Upstash queue validation, and updated the Prisma schema to link embeddings to telemetry records.

subh05sus pushed two commits to main, first adding a full autopilot telemetry tracking feature plus a dedicated embedding generation and processing system for pages and telemetry data (with updated Prisma queries and refined API handlers), then cleaned up build and lint errors by removing unused imports, standardizing function signatures, and tweaking ESLint rules.

subh05sus pushed two refactoring commits to main, tidying up code formatting (adding semicolons, consistent imports and spacing) in route.ts and replacing the old verifyApiToken flow with NextAuth’s getServerSession for user authentication. These updates streamline the chat search API and improve overall code consistency.

On the main branch, subh05sus refactored the autopilot API routes to use NextAuth’s getServerSession with authOptions instead of the custom verifyAuth utility, updating imports and calls in both the config and pause endpoints. The old verifyAuth function was removed entirely, and database lookups now use session.user.id. This streamlines authentication handling across those routes.