Takeout X-Ray reads your 500 MB chat archives without a server
Takeout X-Ray is a free web app that analyzes huge Telegram, ChatGPT, Spotify and Chrome exports inside your browser tab. Nothing is uploaded anywhere.

A developer wanted a "Spotify Wrapped" for his old Telegram chats, ChatGPT conversations and browser history. He ended up with Takeout X-Ray, a web app that chews through gigabyte-sized exports without sending a single byte to a server.
The version 2.0 tool handles 12 services and 13 export formats: Telegram, WhatsApp, Discord, X (Twitter), Spotify, YouTube, TikTok, Chrome, ChatGPT, Claude.ai, plus local logs from Claude Code and Antigravity. It runs on DuckDB-Wasm, a database compiled to WebAssembly, inside a dedicated Web Worker. A live demo with synthetic data is online, and the code is on GitHub.
How does Takeout X-Ray work without a backend?
The author set two hard rules. No backend, because uploading personal messages and prompts for a few charts looks dubious even with a "we store nothing" banner. And no await file.text() followed by JSON.parse(), since a full Telegram export can weigh 300 to 700 MB and crash the tab with an out-of-memory error.
For .zip files, the app reads only the central directory, the index at the end of the archive, so gigabytes of videos and images are never touched. If a Google Takeout archive contains several services, the interface asks which one to analyze. The chosen file is streamed to the parser in 64 KB chunks.
Format detection looks at the first bytes of a file, using a window that grows from 64 KB to 4 MB. Real exports forced that: a Telegram export can carry more than 80 KB of contacts before the chat list, and a 16 KB window wrongly rejected such files.
A streaming parser then turns each message into a flat record and drops the original JSON node. Records go into DuckDB in batches of 5,000 rows, because single inserts are too slow and one giant array eats memory.
Odd corners of real exports
ChatGPT stores a conversation as a directed graph, not a list. Regenerated answers and edited prompts stay behind as side branches. The parser walks back from the final node to rebuild the active thread, and the leftover nodes tell you how often you regenerated a reply.
X's archive files are not JSON at all. They are JavaScript files starting with window.YTD.tweets.part0 = [...], which have to be stripped on the fly.
Apps worth a look
Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.
Talk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.
Open EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.
Pick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.
More from this day
See all →
2.8 million basketwork eels packed onto a Tasmanian seamount

Rotation Station lets you scratch Spotify live on a DIY deck

Voyager 1 is now one light-day from Earth, and still inside

Wylsacom owns 30% of Immortal, the startup behind KloKlo AI bot
More in: Software
See all →
Windows UserAssist tracks every app you open: how to turn it off

Instagram Content Protection tool flags stolen videos for creators

Mozilla Firefox hotline: call 1-855-25-FIREFOX to complain about Edge
