Archiving Twitter likes

I use Twitter to keep track of the latest tech information by following a tailored list of hash tags and people. When interesting information is shared, I like to have the possibility to archive the content for later retrieval. Liking a tweet is the closest thing Twitter offers to bookmark tweets. However, after some time you end with up with a long list of likes which quickly outgrows a usable, searchable archive.

Therefor, to have a solution which suits my needs, I created a Wishbone server to collect and archive my Twitter likes into a simple "grep-able" text file.

The plan

The plan is to have a way to monitor my Twitter events, filter out the like events, process and archive them to disk into a simple text format using following conditions:

  • For each URL in the 'liked' tweet a new entry must be made into a text file. Each entry should be suffixed by a comma separated list consisting out of the author, significant words and the hash tags.

    For example:

    - (elight, working, remotely, terrific, document)
  • A 'liked' tweet without URLs must be archived verbatim, prefixed with date and author.

    For example:

    - Sun Mar 11 04:41:24 +0000 2007 netik:  There is no Internet of Things. There are only many unpatched, vulnerable small computers on the Internet.



pip is used to install Wishbone including the necessary modules for this setup: 1

Although not required, I cannot recommend enough to install Wishbone into a virtual env so you do not "pollute" your system Python installation.

$ pip install wishbone-input-twitter wishbone-flow-jq wishbone-function-twitterbookmark wishbone-output-file

Bootstrap file

Starting the server

$ wishbone start --config bootstrap.yaml

The bootstrap file:

Explaining the module instances


The Twitter input module (line 7) needs the necessary authentication information you have to acquire by creating the OAuth access tokens on


The filter_favs module (line 17) is an instance of wishbone.flow.jq and is used to filter out the created favorite events. It uses jq for pattern matching the data structure the Twitter API returns. The defined condition (line 23) filters out the events of type "favorite" which have been created by user "smetj".

Obviously you will have to alter the query to match your Twitter name.


The bookmark instance of wishbone.flow.twitterbookmark is where the information I (personally) find most interesting is extracted from the Twitter event. Arguably this processing is perhaps not what you personally want. Writing your own Wishbone processing module is quite easy if you want to do custom processing.

split_type, construct_bookmark, construct_text

The split_type module instance splits the stream into bookmark and text types and routes the event to the construct_bookmark or construct_text respectively to construct the correct format (line 50 and line 58) and set the filename to write the desired output to (line 51 and line 59).

funnel, output

The events coming out of construct_bookmark and construct_text are then both send to the output module over the funnel module. The output module writes the value of @tmp.result into the filename defined in @tmp.filename by the construct_bookmark and construct_text modules.

Final words

Using this setup we have seen how to bootstrap a server which collects your Twitter likes and stores them into an easily "grep-able" file for later reference.

If required, it should be really easy to modify this setup into something that suits your specific needs.

If this is useful to you or if you have any questions about the setup, don't hesitate to drop me a line or add a comment.


  1. wishbone.flow.jq requires some dependencies