This week brings slicker data tables, expanded AI course content, handy WhatsApp utilities, and richer tools setup. It’s a roundup of upgrades that make data science, coding, and message scrapes more insightful and usable.
LLM multiplication results got a style and info refresh for clearer insights and easier reading.
- Rich interactive tables: The main page now shows model scores with right-aligned text and color-coded backgrounds from red to green for easy accuracy spotting, thanks to a concise rewrite split into a modular
script.js(09d3ac6). - Popover details: Hover cells reveal numbered lists of all model responses or errors for each multiplication, offering more transparency on how LLMs answered.
- Accurate stats counting: Multiplications now run five times each, improving stats fidelity and overall accuracy calculations (f9cc4c5).
- Clear column renaming: The first column is now “Model” instead of a vague “m”, and the “%Win” cell disables wrapping for neater tables.
- Minimal and modern code: Uses ESM imports from D3, and bootstrap popovers for UX; the overhaul makes the script concise, readable, and visually crisp.
Yes, you really needed another level of polish on LLM math tables!
The official IITM Tools in Data Science course content receives timely updates and important clarifications.
- Expanded teaching assistants roster: Added two new TAs to the README, helping you get more hands-on support during the course (066078b).
- LLM Agents content introduced: Added comprehensive lecture notes and code for building autonomous agents that intelligently reason, act, and learn—demonstrated by a command-line Python agent executing code on your behalf (8e93540).
- Project 1 points fixed: Content links now point to the correct Jan 2025 term materials, reducing confusion (890689c).
- Improved learning workflow: Rewrote example commands to use correct LLM CLI flags, promoting smoother hands-on experience (720e4bd).
- Prettier reformat: Project 1 documentation reformatted for better readability (27562d8).
Turns out some AI agent tutoring is exactly what you needed for your data science hustle.
WhatsApp scraping and viewing gains new tools for deeper chat analysis.
- WhatsApp scraper bookmarklet added: Open any chat and capture the entire message log as JSON, including quoted messages for context. Copy it straight to clipboard for archiving or analysis (9af824b).
- WhatsApp thread viewer: Paste scraped JSON into this new tool to see messages in neat threaded conversations, revealing reply hierarchies and quoted replies (perfect for untangling long chats) (4535667).
- Refined scraper: The scraper nicely extracts quoted author names, message IDs, and handles system messages cleanly, improving accuracy for the viewer (5c05948 + bf34a72).
- Cyborg Scraping archive: The old scrapers repo now points users to the tools repo with better bookmarklets, paving the way forward (ce157d9).
- UI finesse: Fixed bookmarklet title spacing for cleaner appearance (cdeaf62).
You never thought WhatsApp threads could look this organized, did you?
Shell setup and productivity tweaks smooth daily coding life.
- Conda prompt removed: Simplified prompt by ditching unused Conda environment indicator (897a96c).
- Fish shell setup cleaned: Migrated node version manager and lmstudio to main Fish setup. Also added Google Cloud SDK to path for easy CLI usage (897a96c).
- Fixed variable names in
charsscript: Avoid confusion by clearly using distinct names for line content and line number (98e52b8). - Markdown converter
md2htmlupgraded: Uses GitHub Flavored Markdown with extensions like bracketed spans, fenced divs, subscripts, and superscripts for richer HTML output (4a8c6c0). - Linux setup updated: Added new apps like Opera, ffmpeg, w3m, Google Cloud SDK, PostgreSQL, Supabase, VLC, and Ollama models in setup doc (897a96c).
Less clutter, more power in your terminal. Ah, the sweet smell of effective scripts.
Entity extraction API polished with better docs and smoother deployment.
- Secure token handling: Moved LLMFOUNDRY API token from request header to Cloudflare environment variable, simplifying authorization (5e431e5).
- Expanded README: Added detailed deployment instructions, live curl examples with expected JSON outputs, and Swagger UI docs introduction for easier developer onboarding (97ae7be, e10d936).
- Interactive Swagger UI: Hosted Swagger UI directly within the worker, letting you explore API docs visually at
/docs(97ae7be8). - Updated worker code: Improved code structure for handling CORS, extracting JSON, and serving docs, with neat error handling and consistent responses (e10d936).
JWT tokens are so last year – environment variables FTW!
A personal repo portfolio enhanced with better repo link handling and topic filtering.
- Fixed repo footer links: Cards now link correctly to repo main pages, while clicking card bodies opens homepages if available (8f972d1).
- Show repos without topics: Previously hidden repos with no topics are now displayed under an “(other)” category to avoid confusion (9a42c9c).
A portfolio repo that finally stops playing hide-and-seek.
- Capture and share a screencast demo of the new LLM multiplication table UI to showcase its slick new visuals.
- Develop a tutorial video on building and deploying the LLM command-line agent from the new tools-in-data-science material.
- Write a blog post guiding on how to scrape and visualize WhatsApp threads using the new scraper and thread viewer tools.
- Explore automating your shell environment updates with scripts that integrate the updated Fish and bash setups from
scripts. - Add more interactive examples in the entity extractor Swagger UI to attract developer users and get feedback.