This was a big week for the analyze tool. I think the feature set is really coming together and you can actually feel why pulse is better than other solutions. We recently launched the Create Update - this gives the analyze tool the ability to pipe your conversation history to an artifact generator that makes pdfs, presentations and graphics with human-in-the-loop feedback:

I find the PDFs feature the most helpful/useful so that's what I included a photo of - the feature is available on all nodes now if you want to try it out.

Another big Analyze tool upgrade that was released this week was an updated context management experience. Rather than an endless chat scroll that doesn't show you when you're hitting limits/when your history is getting too long for most models to stay on task I'm testing having a context block limit of 50. Token limits are still set by-model, this is less about the amount of text you're sending and more about forcing the AI you're working with to stay on track by keeping the amount of active context under control/giving it some kind of structure.

The first things you'll probably notice are the new group/group all button - this will group all non-flagged, non-grouped context entries (messages, digests, search results, etc) and add them to the bottom of the conversation as well as the context meter (currently /50) at the top and green/yellow/red context meter indicator around the chat modal:

Context meter at the top center, updated buttons, you can also see the colored border used for visual feedback on how close you are to the limit

This is what a context group looks like once created:

You can also select individual context items to create smaller/more targeted groups:

Selecting a group-able context block highlights it - use the purple "group" icon to create a group out of selected entries

Here's the context meter going to yellow and red as the amount of context entries increases:

Yellow at over 30 messages
Red at 50 messages + Warning and lockout

When you reach this point you can use the group button at the top without any context items selected and it will group everything that isn't flagged - speaking of which - for important responses or messages we have a "flag" option that prevents them from being grouped. This is a way to "anchor" the key context points in your workflow - code examples that really worked, style guides, etc. We also added the much-requested ability to copy and export individual context blocks rather than the whole conversation.

flag/copy/txt options

Boring Stuff - Non-technical audiences may want to avoid

There was an absolutely hellacious bug fixed this week in the code tool. In our local dev environments we were able to upload zip files to the code tool without issue and run them through "pulsification" (our backend api that generates auditable pol files from an existing codebase) but due to strange limitations with our backend framework, or perhaps with cloudflare (we're not 100% sure) we weren't able to receive the entire zip file in a way that wouldn't mangle it and post it to the backend in a usable state.

To get around this I simply finished the git-based pull/pulsify functionality that was promised when the tool first launched and made it the default - let us know on discord/x/private messages if you find any providers that don't work properly. It should support any https-based git request (ssh support is unlikely - sorry!)

Go Forth and Analyze

This is the most real-life friendly version of the analyze tool we've put out so far so I hope everyone gets some use out of it.