Write-ups of how my projects work underneath: the architecture, the rewrites, and the decisions that aren't obvious from the code.
Most READMEs tell you how to run a thing. This is the other half — why it's built the way it is, and what earlier versions tried. It's also the source material for posts on my site.
Author: Yashasvi Allen Kujur · yashasviallen.is-a.dev · GitHub @JavaProgswing
| Folder | Project | What's inside |
|---|---|---|
| valorantnarrator/ | Valorant Narrator | Three years and ~40 releases of a desktop app that reads Valorant text chat and speaks it on team voice. Five generations of chat capture, the audio-routing problem, four voice-synthesis backends, and the backend/quotas/payments story. |
| discord-bots/ | VoithosHelper → Vithron → Aestron | One Discord bot across three rewrites and five years: a 5k-line Replit file, a 17k-line / 1,868-commit Heroku monolith with a Postgres schema and dashboard, and a 39-module discord.py 2.7 package with a FastAPI site, Riot RSO linking and tests. |
Drafts for new site posts live in blog/.
More teardowns will land here as I write them.
Everything I've built that's worth pointing at, with a one-line purpose and a rough timeframe. Dates are repo creation → last activity (or local first → last edit for things never pushed). Grouped by theme, newest-ish first within each group.
| Project | When | Purpose |
|---|---|---|
| ValorantNarrator (OPS mirror) | 2023-08 → 2026-07 | The desktop app: Valorant text chat → real-time agent-voice comms. Full teardown in valorantnarrator/. |
| VoiceCloner | 2025-12 → 2026-07 | Generic local XTTS-v2 voice-clone server; shipped as the agent-voice engine. |
| AgentVoiceServer (private) | 2026-07 | Current local NeuTTS Air voice server with a backend authorization gate. |
| api-valnarrator-vercel (private) | 2024-10 → 2026-07 | Current backend: FastAPI on Vercel + Supabase, quotas, temp AWS creds, agent-voice licensing. |
| site-valnarrator-vercel | 2024-10 → 2025-12 | valnarrator.vercel.app — landing, download, ToS/privacy, referrals. |
| payment-valnarrator-vercel (private) | 2024-10 → 2025-11 | Checkout + webhooks: PayPal (international) and Razorpay (INR). |
| ValorantMatchStats | 2025-01 | Match-stats viewer, built on the same Riot local-API knowledge. |
| whispercpp | 2024-08 | whisper.cpp snapshot used for the (short-lived) in-app dictation experiment. |
| Project | When | Purpose |
|---|---|---|
| jailbreaker-llm | 2026-09 | LLM-vs-LLM automated red-teaming research platform (attacker fine-tuned against a target). |
| cineguide-rag | 2026-08 | Evaluated movie-discovery RAG: hybrid retrieval, grounded answers, monitoring. |
| rag-n8n-spring-platform | 2026-09 | Production-style RAG platform: Spring Boot, n8n, pgvector, Mongo, Redis, Ollama, CI. |
| syllabusRet | 2026-09 | Syllabus RAG assistant with cited, evidence-bound answers. |
| Urban-Heat-Mitigation | 2026-06 → 07 | Physics-informed geospatial ML for urban-heat hotspots and mitigation planning. |
| climate-platform | 2026-03 | Flood prediction + urban-heat + energy forecasting, end-to-end. |
| garbage-sorter / waste-classifier | 2026-03 → 07 | Image classifier for waste sorting (HF Space). |
| Claimify | 2026-08 | Evidence-based claim verification: Firecrawl + local NLI, web app + browser extension. |
| StockMarketPredictor | 2025-04 | Stock outlook from latest news + market data via an LLM. |
| ml-* course projects | 2026-07 | Nine from-scratch ML projects (titanic, fraud, wine, digits, gaze/gesture, …). |
| Project | When | Purpose |
|---|---|---|
| aula-f75 | 2026-07 | Reverse-engineered USB-HID protocol for the Aula F75 keyboard — full keymap read/write, no official SDK. |
| snake-io-bot | 2026-07 | Snake.io perception + autopilot over ADB: explicit temporal vision layer, certified per-class precision. |
| ChessAutomation | 2025-07 → 2026-02 | Desktop chess assistant: GUI, analysis, move automation. |
| SRM-WIFI-Login | 2024-12 → 2026-02 | Windows tray app that auto-logs-in to the campus captive portal. |
| SRM-Academia-Timetable | 2025-08 → 2026-01 | Visualizes the SRM academic timetable. |
| gridee-cli | 2026-07 → 08 | ADB/UIAutomator CLI that books badminton slots through the app's own UI. |
| ChatGPTSeleniumAutomation | 2024-08 → 2025-04 | Drive the ChatGPT web UI programmatically. |
| wifi-walk-survey | 2026-03 → 07 | DIY Wi-Fi site-survey tool: walk a building, IDW-interpolated signal heatmap (NetSpot-lite). |
| LenovoLegionToolkit (fork) | 2026-05 | Extended Legion automation (Bluetooth, DND, fan, night-light…). |
| Project | When | Purpose |
|---|---|---|
| Personal-Apex | 2026-06 → 09 | "Apex" — a personal productivity OS for college life (timetable, tasks, focus). |
| RepoSignal | 2026-10 | Turns a repo's issues/PRs into an evidence-backed maintainer attention queue. |
| GotYourBack | 2026-09 | SRM-only marketplace for borrowing/lending/selling within a trusted student community. |
| AssistanceSupport | 2026-02 | AI support portal: Gemini + image analysis for refund claims + policy enforcement. |
| Roadmap-Builder | 2025-08 → 2026-02 | Skill/academics roadmap generator. |
| GDGoC | 2025-04 → 07 | Speaker-session booking platform (OTP auth, calendar, email). |
| CompleteBlogApp | 2025-10 | Full-stack blogging platform (CRUD, auth). |
| ChatAnalysis | 2026-05 | Chat-export analysis dashboard (WhatsApp/Discord/ChatGPT/Claude) + optional local-LLM profiling. |
| Aestron | 2023-01 → 2026-07 | Full-feature Discord bot (moderation, music, giveaways, tickets, Valorant UI) — 3rd generation; teardown in discord-bots/. |
| vithron_webdashboard / VoithosHelper1 | 2021 → 2022 | Earlier generations of the same bot: Quart + Discord OAuth2 dashboard, and the original Replit bot. |
| Project | When | Purpose |
|---|---|---|
| chess-server + Checkora | 2025-12 → 2026-01 | Chess backend (Spring Boot) and a minimax + C++ engine chess platform. |
| EffaceDistractions | 2025-12 | JavaFX app that blocks distracting applications. |
| flight-server + FlightReservationSystem | 2025-12 | Flight reservation: Spring Boot backend + HTML/JS frontend. |
| ChatApp | 2022-10 → 2025-01 | Swing chat application (+ Quarkus API rework). |
| Vosk speaker-recognition / WhisperCPP (local) | 2025 | Speech-to-text and speaker-embedding experiments that fed the dictation work. |
| AdventOfCode2025 | 2025-12 | AoC solutions. |
Forks I've contributed to or studied (classpro, Wavelink, tesseract.js, goscraper, auto-resume, procurement-ai-assistant, and the GSoC-org set) are intentionally left off this index — it's for things I built or led.
- Each teardown is its own folder with a
README.mdentry point and numbered chapters. - Diagrams are Mermaid so they render on GitHub and stay diffable.
- Secrets, keys, account identifiers and user data are never reproduced here. Third-party internals are described at the level needed to understand my design decisions, not to reproduce anyone else's service.