Name: Griptape
Uploaded: 2024-12-20T05:59:00.812Z
Duration: 2 min 33 s
Description: A modular open-source Python framework for building and deploying LLM-based agents, pipelines, and workflows.

griptape

A modular open-source Python framework for building and deploying LLM-based agents, pipelines, and workflows.

Griptape

Hey my name is Zach and I'm a software engineer here at GripTape. Today I'm going to show you an overview of GripTape Cloud highlighting some of the features that GripTape Cloud provides and the essential building blocks that you can utilize for quickly spinning up AI driven applications. 

Here on the left we're looking at the landing page you'll see when you first log into GripTape Cloud. This page provides an overview of various workflows you can accomplish on Cloud including ETL, Rag, and running custom LLM powered agents. 

We'll start with data sources. This component enables you to

Bring your own data into Cloud to ingest and transform. You can select from various supported data sources or author your own code to bring in data with structures which we'll cover soon enough. I've already got some data sources in my account. The succeeded status indicates that the most recent data job for that source completed as expected. 

Ingesting and storing the data vectorized in Cloud with data sources prepared, we can then configure a knowledge-based resource to organize multiple data sources and make them accessible for querying by AI applications. This particular knowledge base.

Was configured with a data source containing information from a PDF about retrieval augmented generation. And we can test this knowledge base and data source right here in the UI by asking something relevant to the contents of that PDF in which it'll generate a response relevant to retrieval augmented generation. Things get more interesting as we move o over to explore structures. 

Now structures in the cloud come from the concept of the same name in The Python based GripTape framework. You can read more about structures in the grip tape documentation, but they're effectively custom.

Python applications powered by llms grip tape Cloud enables you to take any grip tape structure that you authored in a GitHub repo and deployed as a hosted endpoint. We've got some samples that you can get started with and I've already got a sample deployed in my account called grip tape chat memory agent. 

So this structure is configured only with my grip tape Cloud API key as well as my open AI API key which are both stored securely in the cloud as secrets. The deployment history as well as the Run history of the structure available right in the UI and this structure is configured.

Out of the box to make use of threads rule sets and rules. And I'll explain what those resources are momentarily. But first I'll show them in action with the help of the grip tape chat application that I have pulled up on the right hand side here. 

So this is a gradio application configured to invoke a hosted grip tape Cloud structure and it's available for use as an open-source repo. I already have it configured to target my chat memory agent structure. So I'll just go ahead and send it a message here. Can you explain retrieval augmented generation to me? And there's a couple things.

Going on here with this application. Firstly, the gradio app is configured to invoke the structure using a thread resource. The thread is used as a container to hold messages. So this enables the structure to make use of conversation memory, which is another topic you can read about in the documentation. But it allows structures to remember conversation across runs. 

So if I then follow up with what is my name as a follow up to the first message that I asked the structure, it should be able to retrieve that conversation memory and tell me that my name is Zach. Um additionally this.

Structure is configured to make use of rule sets. Another framework concept that you can read about. Rule sets and rules they are vital for providing guard rails for the LMS to achieve more consistent behavior. Having these as Cloud resources enables you to iterate on and reuse rules and rule sets among multiple structures or applications. 

I'm using this rule set which directs the LLM to make use of knowledge bases when applicable to site knowledge bases when referencing information from them and to always mention how much they love uh grip tape cloud in the response as if that wasn't.

Already part of the lm's initial response. So with these Cloud resources you can spin up your own custom AI application in a matter of minutes. If you have any questions feel free to reach out to us. Uh we're available and active in the grip tape Discord server. So come and say hi. Thank you for watching.

Note, that there may be slight differences in text and video. Watch video for the original content.

Griptape Video