Facebook Message Scraper

A simple python script to download the entire conversation from Facebook, not limited like the one in the data dump provided by Facebook

Outputs the conversation in a JSON format, as well as the JSON for each individual chunk.

Initial Setup

Run for both dumper.py and group_dumper.py

1. In Chrome, open facebook.com/messages and open any conversation with a fair number of messages
2. Open the network tab of the Chrome Developer tools
3. Scroll up in the conversation until the page attempts to load previous messages
4. Look for the POST request to thread_info.php
5. You need to copy certain parameters from this request into the python script to complete the setup:
   1. Set the cookie value to the value you see in Chrome under Request Headers
   2. Set the __user value to the value you see in Chrome under Form Data
   3. Set the __a value to the value you see in Chrome under Form Data
   4. Set the __dyn value to the value you see in Chrome under Form Data
   5. Set the __req value to the value you see in Chrome under Form Data
   6. Set the fb_dtsg value to the value you see in Chrome under Form Data
   7. Set the ttstamp value to the value you see in Chrome under Form Data
   8. Set the __rev value to the value you see in Chrome under Form Data

You're now all set to start downloading messages.

Downloading Messages

1. Get the conversation ID for those messages by opening http://graph.facebook.com/{username-of-chat-partner}
2. Copy the id value from there
3. For group conversations, the ID can be retrieved from the messages tab, as part of the URL. You must use group_dumper.py instead.
4. Run the command python dumper.py {id} 2000, and put the value you retrieved for ID earlier

Messages are saved by default to Messages/{id}/

Known Issues

The script sometimes has trouble with very large conversations (>100k messages). Facebook seems to rate limit this, and returns empty responses. In such cases, the script will retry after 30s until it gets a valid response.

It may take the script several tries to get a valid response. DO NOT PANIC.

Interrupting the execution before completion only leaves the JSON chunks, not the stitched file.