Anyone using DEVONagent Pro in 2024 on a regular basis? If so, what are you using it for? I’ve had this app for years, and it’s never been able to be what I wanted it to be. I always hoped it would eventually support Confluence wikis, or that I’d be able to develop a plugin to search the wikis, but so far no luck.
On top of that, the results I get back from quick searches of the web seem to always be of poor quality. For example, here is a screenshot of a search with DA for “devonagent pro plugin development”:
Just did, I didn’t see anything in there that’s news to me. Repeated claims that DA gives better results, removes duplicates, and reduces tedious tasks. None of which I’ve found to be true. Which leads me back to my original question, “what are people using this tool for in 2024”?
A lot of my work is technical research. I need to spend a lot of time reading AWS documentation, programming language documentation, Linux documentation, etc… As well as my own archive I’ve built up over 20 years or so. In general, I’ve found search engines to be getting worse over time as spam, seo junk, and their own corporate interests prioritize ads over relevant results, and I think DEVONagent might be a great replacement for what I need to do.
But actually using it is tedious, or at least clunky. Or… as I suspect and which led to this thread, maybe I’m not thinking of it correctly. I know the documentation says that it’s “far from” a replacement to a search engine, but then it immediately states that DEVONagnet does a few things that are exactly what a search engine does.
So, what do you use it for? Where do you draw the distinction between what you’ll use a search engine for and what you’ll use DA for?
Many people use DEVONagent in a more passive way, creating a search set and verifying they like the kinds of results it returns, then scheduling it to run on its own. The results are then sent to them or put into DEVONagent’s Archive for later inspection. In that way, it does act as an “assistant” in their process.
Due to the nature of my job, I don’t think my experience is as useful. I have to do things in a very ad-hoc manner, turning things on and off repeatedly, testing, etc. So my results over time are atypical. And the two I have on schedule are just crawling a few RSS feeds.
I use it when all other browsers fail to return anything useful and I know bloody well there is more out there. It almost always finds something useful that other search tools rank many pages down, if they bother to consider it applicable at all.
In my own experience and use cases, the key is building the right search set. I don’t myself use it as a front end to general search engines, powerful though its multi-engine search is; I use it almost entirely for academic bibliography trawls, at which it’s stunningly good – partly because this is an area well served by online databases and repositories that can be gathered into a custom search set. It’s particularly brilliant for getting an initial state of the conversation on a subject you know nothing about, but also great for finding out-of-the-way items that slip through other bibliographical nets, and the ability to save a digest of all the results is a godsend. But it’s also invaluable for quite mundane things like trawling Wikipedia for all the articles that deal with a particular topic, or (if you have DT Server) generating digests from your own databases.
DevonAgent Pro is a premium product that claims to be the “smartest search assistant for the Mac.” It’s not unreasonable to ask it to justify that description.
(Full disclosure: I also rarely use it and have found search set design to be fairly opaque.)
Search set design, as opposed to plug-in design? The latter I can absolutely see, but search sets are just a matter of checking boxes in the list of plug-ins and adding whatever additional sites you want to include. I suppose you do have to choose between Search and Crawl for the latter, but the answer is nearly always Search.
I have not been successful in consistently getting results from DA that are more useful than what I can get via standard search engines. Part of the opacity is that I’m not sure whether that’s due to poor query/search set design or poor plug-in design. Since my goal is to find information, not to become a DA expert, I have to admit I haven’t put a great deal of effort into understanding the problem.
As @NickLowe implied - DA does not work particularly well for ad hoc searches. Traditional search engines - possibly now assisted by AI such as Perplexity - excel there.
DA excels when there is a particular type of search that you routinely perform- such as academic citations in a particular database or current news stories from particular sources. In that situattion it is worth the effort to learn how to make DA do what you wish.
For a long time, the academic database I use most often was inaccessible to DA. That has been fixed, but the database’s own search tools still give more consistent results.
I am similarly frustrated by DevonAgent Pro, thinking all the time that I must be missing something in how to use it. I find it clunky and laborious. HOWEVER…I recently needed to understand the side effects of some drugs I’ve been prescribed, and the interaction of them. In this case, being able to explore the network diagram of search terms and results enabled me to find results I wouldn’t have found in straight searches with Google. So I resolved to learn how to use DevonAgent Pro better, and work out what I am missing in how to get the best from it.
I’d argue that your first bullet point: