calculusphile

AI Chat Exporter

A Python CLI tool that converts saved HTML chat logs from ChatGPT, Gemini, Claude, Copilot, and DeepSeek into structured Markdown notes.

AI Chat Exporter

v3.2.0 — Convert AI chat HTML exports into clean, tagged Markdown notes.

A modern Python CLI tool that converts saved HTML chat logs from ChatGPT, Gemini, Claude, Copilot, and DeepSeek into structured Markdown files — ready for Obsidian, Notion, or any knowledge base.

Features

Feature Description
Live Watch Mode Auto-detects new HTML files in your Downloads folder via watchdog
Full-Page Export Convert an entire HTML chat page to Markdown — no search needed
Platform Cleanup Auto-strips sidebars, branding, overlays, and input areas
Content-Safe Cleanup Text-length guard prevents accidental removal of large content containers
Gemini DOM Support Handles Gemini’s custom Angular web components
User-Code Dedup Removes code blocks from user messages so only the AI’s code appears
Batch Processing Process every HTML file in a directory at once
CLI + Interactive Full argparse CLI flags or guided interactive menu
Smart Extraction Finds specific AI responses by search phrase
AI Smart Titles Generates clean headings from verbose questions (AI or heuristic)
Session Merging Append multiple extractions into a single “Master Note”
15+ Language Detection Python, C++, JS, TS, Rust, Go, Java, SQL, Bash, Ruby, C#, Kotlin, Swift, and more
3-Tier Detection HTML class → proximity search → syntax analysis
YAML Frontmatter Auto-generated tags, date, source for Obsidian compatibility
Zero Config Start Works out of the box — config.json is optional

Quick Start

1. Clone & Install

git clone https://github.com/calculusphile/AI-Chat-Exporter.git
cd AI-Chat-Exporter
pip install -r requirements.txt

2. Run (Interactive)

python watcher.py

3. Run (CLI)

# Live watch mode
python watcher.py --watch

# Process a single file
python watcher.py --file "path/to/chat.html"

# Full-page export (entire HTML → Markdown)
python watcher.py --file "chat.html" --full-page

# Batch process a folder
python watcher.py --batch "path/to/folder"

# Merge all extractions into one file
python watcher.py --file "chat.html" --merge "StudyNotes.md"

CLI Reference

Flag Short Description
--version -v Print version and exit
--watch -w Start live-watch mode on Downloads folder
--file PATH -f Process a single HTML file
--batch PATH -b Process all HTML files in a directory
--full-page -p Export entire page instead of searching for phrases
--merge NAME -m Merge all extractions into one .md file
--downloads PATH   Override the watched Downloads directory
--debug   Enable verbose debug logging

How It Works

  1. Input — Save any AI chat page as .html (Ctrl+S in browser)
  2. Detection — The watcher picks it up, or you pass it via --file
  3. Mode — Choose between search-based extraction or full-page export
  4. Smart Titles — Verbose questions are auto-cleaned into concise headings (AI or heuristic)
  5. Language Detection — Code blocks are analyzed with a 3-tier strategy:
    • HTML class attributes (language-python)
    • Proximity search (nearest text label above the block)
    • Syntax pattern matching (regex on code content)
  6. Output — Clean Markdown with YAML frontmatter, auto-tags, and proper code fences

Project Structure

AI_Chat_Exporter/
├── watcher.py            # CLI entry point + file watcher
├── converter.py          # HTML → Markdown conversion engine
├── config_loader.py      # Typed config management
├── title_generator.py    # AI + heuristic smart title generation
├── logger.py             # Centralised logging
├── config.json           # User settings
├── requirements.txt      # Dependencies
├── ARCHITECTURE.md       # Developer guide & change-impact map
├── README.md
├── LICENSE
└── Exported_Notes/       # Output (git-ignored)

Tech Stack

  • Python 3 — Core language
  • BeautifulSoup — HTML parsing & DOM manipulation
  • Watchdog — File system event monitoring
  • OpenAI API (optional) — AI-powered title generation