A simple, text-based system for building an easily-searchable knowlegebase on OS X. CLI, LaunchBar and Alfred included.
The story behind the script is: I forget everything. I forget what I was working on last night. I forget where I left important things. I forget how I solved a major problem and have to work it out again next time it comes up. I forget where I saved the perfect settings I found for h.264 compression. You get the picture. So I've started building an archive of plain text files with questions and answers, and a system which makes it easy to add new knowledge at any time. The question format makes it easy to query, even when I don't remember the answer at all.
The system is very simple. A bunch of plain text files, each titled with one question and containing one answer. Searching and sorting is handled my the
mdfind Spotlight interface with scripts and extensions that focus the search to a narrow scope with predefined options. The knowledgebase that's built is bulletproof and portable, scriptable and easily searchable. Plus, it works the same way as Notational Velocity, so they make a great pair.
I use it with nvALT and sync to Simplenote and Dropbox, which means I can enter new questions and their answers from any mobile device or remote computer, in addition to being able to use the command line tool, Quicksilver actions, LaunchBar actions and Alfred extensions. It keeps me sane.
: Put the
chmod a+x /path/to/qq. Edit the script to set the location of your notes folder and the extension you use. You may want to set a different preferred "question" prefix if you already have one (or don't want filenames that need constant escaping).
: Move the two .scpt files into
~/Library/Application Support/Quicksilver/Actions/. These scripts require that the above command line script is in place and need to be edited with the proper full path to the script.
: Move the two .scpt files into
~/Library/Application Support/LaunchBar/Actions/. These scripts require that the above command line script is in place and need to be edited with the proper full path to the script.
: The Alfred extensions are standalone and do note require the
qa commands will be available. Once imported, there are several configuration variables that need to be set in the top of the script.
: By nature, you don't need anything special to use this system with NV. You just need to save your notes to plain text files if you want them to be accessible to
mdfind and other system tools.
: To get simple command line access to the question in your NV folder, set up the configuration in the
: Add questions and answers using
qq -a "Question" "Answer", e.g.
qq "What is the answer to life, the universe and everything?" "42". If
-a is specified without any additional arguments,
qq fragmented question, e.g.
qq meaning universe. See the Querying section for more information on composing fragmented queries.
Quicksilver : Add a question and answer by launching Quicksilver and entering a string in the format "question: answer", e.g. "what is the meaning of life, the universe and everything: 42" (activate text entry with the "'" key). Press tab, select the "Quick Answer" action and hit Enter. : Ask a question by typing launching Quicksilver and entering a fragmented query (activate text entry with the "'" key). Press tab, select the "Quick Question" action and hit Enter.
LaunchBar : Add a question and answer by launching LaunchBar and typing "qa". Select the "Quick Answer" action and hit Space. Then type a string in the format "question: answer", e.g. "what is the meaning of life, the universe and everything: 42". : Ask a question by typing "qq" and selecting the "Quick Question" action. Hit space and type a fragmented query.
Alfred : To add a question and answer to your archive, open Alfred and type "qa". Enter the question and answer in the format "question: answer", e.g. "qa is there anybody out there?: nope". : To query your archive, type "qq" and enter a fragmented query.
Notational Velocity/nvALT : Notes will be prefixed with the string specified in the configuration (default "??"). You can start a search with the prefix and follow it with a fragmented query to find the question (and answer) you're looking for. Add new questions and answers by typing the prefix and the question in the search field, press enter to create a new note, and type the answer and any additional meta.
Ask yourself what question the search for this answer began with. Will that be the question you ask again when you've forgotten the answer a year (or month) later? Think about how you, personally, would phrase the question and shape it around keywords that you're pretty sure you would repeat.
Keep your questions in a natural language format, but avoid contractions and use "who, when, where, what, how and why" as much as possible. These words make it easy to sort out different types of questions about the same subject.
When querying, only use operative terms to get the best results. "Where did I leave my glasses" will return poor results if the question you labeled it with was "Where did I put my glasses". Instead, query "where glasses" and you'll find the note instantly.
grep to parse out pieces of questions and/or answers that you couldn't find otherwise.
You can add additional meta within the contents of the file to help in locating the answer later by using the format
@(meta information). Anything within the parenthesis will be included in search, but the whole tag will be excluded from results in the
An example of this would be a question like "What brand of cleaner did I use during the move?" The answer you gave was a cleaning solution you were really impressed with. A year later the only thing you may remember to ask is "cleaner," but you might not even remember that was the term you used. So you add "@(solution cleaning favorite)" to the answer. Now you can search for "favorite cleaning" or "cleaning solution" and still get the question you were looking for. It's similar to using tags, but you can be more verbose without cluttering up your pool of existing tags. In the example case, you can also use just the root or part of a word, e.g.
qq fav clean to expand matching possibilities.
If your answer includes a command, url or other piece that you probably want to grab, you can surround it with @copy(part to copy) and the contents of the tag will be copied to the clipboard when the question is answered. The '@copy(' and ')' parts will be stripped from the answer given, but its contents will remain inline.
If you add one or more @Open() tags, the contents will be passed to the
open command with the
-g flag (open in background). You can use this to add related urls to the query. It's not a perfect system because if you ask a broad question and get 20 answers, each with one or more urls, they'll all open. But it's handy once in a while.
If you're using Quicksilver, LaunchBar or Alfred it's assumed that you can locate the question file and open it in an editor of choice easily enough. It's even simpler if your questions are stored in Notational Velocity/nvALT as you can edit them at the same time you search for them.
From the command line you can use the
-e argument to open the first result for the following query in the editor you specified in the configuration (
mate by default). It expects a command-line tool, not an OS X application name. Most editors have these available.