Operator Log #0004
New Object Types, Periodic Indexing, Community Discord
Hey y'all — hope everyone is having a great Friday! Hard to believe it’s already July 15th, this month is going by super quick. This past week has been super busy on our end, a lot of behind the scenes work in addition to a bunch of user facing stuff. Super excited to share what we’ve been up to this week.
Before we get into the updates, wanted to give a quick shoutout to Samarth Jajoo for shipping the first version of riff.quest! The tool works by indexing Wikipedia articles on the fly (using Operand) and passing this information into large language model(s), similar to the approach that we outlined in one of our recent blog posts. After using the tool a good bit internally, we were blown away by how good it is, and given that this is only the first version, we’re super excited to see what comes next!
Spent a bit of time this week putting together our community Discord server, a place for our users to bounce ideas off one another, give feedback / suggest improvements, and get help with any bugs or issues that arise. Our team is pretty active in the server, so feel free to drop in and say hello!
Here’s the invite link to join: https://discord.gg/WpaFpt5C9M.
Looking forward to seeing all of y’all in there!
New Object Types, Periodic Indexing
We’ve recently added support for periodic indexing, i.e. being able to periodically refetch data from a source and index it incrementally. This opens us up considerably to a number of new object types, specifically:
- RSS Feeds. It’s now possible to copy+paste an RSS feed link directly into Operand, and we’ll take care of the rest. Periodically, we’ll re-fetch the RSS feed and index any new data.
- Email Inboxes. We now have support for *.mbox file types, which is mainly useful for one-time dumps of historical email messages. If you did want to index your entire email inbox with Operand, we encourage a one-time dump of historical messages + a trigger which creates new
- Notion. You can now index your entire Notion workspace directly into Operand, and yet again, we’ll make sure that it’s kept up to date with all your latest changes. This is part of a much larger project which we’ll be talking more about next week. Stay tuned!
One note, we currently check for new updates every two hours. If your use case needs faster updating, send us a message or join our community Discord server, we’ll be happy to chat and possible accommodate.
Bug Fixes and Improvements
- Updated our API docs, fixed a number of typos and added a bunch of missing endpoints and object types. Added links to our two SDKs: Typescript and Go.
- Fixed an issue where API key “last used” date would be reported inaccurately.
- Fixed an issue with our blog where the date of a post would be shown as the day before the post was actually published.
- Added support for a
urlparameter within the
htmlobject type. If the HTML content is unspecified, but we have a URL, we’ll try our best to populate the content by fetching it from the URL.
- Added support for fetching JS-based websites, i.e. websites who rely on client-side rendering to display the contents of a page. If we detect that a website is JS-based, we fire requests via a headless chrome machine.
- Made a bunch of reliability improvements to our serverless infrastructure, still seeing a few issues with the latency, but overall, the system should be much more stable.