Anthropic Examples/Readme Guide Created and Proper Reference Noted. #466
Conversation
Contains 3 journals showcasing sync/async using anthropic and a tool example!
Made better example
This example uses async in a little more meaningful way while also taking care of a few fixes and errors!
Finalized page and made clearer! Also fixed a number of issues within the async function example.
Changed Anthropic example link to point to the guide readme
|
Let's take a look @the-praxs |
|
Thanks @TKTSWalker! The sync notebook looks great but it looks like you're using Anthropic Computer Use in the others, and those don't work. Is that what you're after? Or did you just coincidentally name the roles "computer" |
|
Good afternoon! I named the role computer by coincidence, the name should be able to be replaced with a term like "Machine" if need be! i'm just used to naming the role computer, as with the way the system works, the output shouldn't be affected! If there are any problems with this however, please feel free to let me know and I will act accordingly! |
Hahaha that's what I figured. Can you make the changes and re-run the notebooks end-to-end to make sure they run properly? Will merge ASAP when you push it |
|
Hello, it's no problem! Going to start doing so in a moment, I'll share a
copy with all the output logs once checked!
…On Fri, Nov 8, 2024, 9:05 PM Alex Reibman ***@***.***> wrote:
Good afternoon! I named the role computer by coincidence, the name should
be able to be replaced with a term like "Machine" if need be! i'm just used
to naming the role computer, as with the way the system works, the output
shouldn't be affected!
If there are any problems with this however, please feel free to let me
know and I will act accordingly!
Hahaha that's what I figured. Can you make the changes and re-run the
notebooks end-to-end to make sure they run properly? Will merge ASAP when
you push it
—
Reply to this email directly, view it on GitHub
<#466 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ANRX3WVGRUHZ7ZW5EE5RSN3Z7VUU5AVCNFSM6AAAAABQUHIGCWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRVHE4DSMBXGM>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Files selected (22)
Files ignored (18)
InstructionsEmoji Descriptions:
Interact with the Bot:
Available Commands:
Tips for Using @Entelligence.AI Effectively:
Need More Help?
|
|
Whoops, found my errors! I was thinking about raw text input/output like the Llama.cpp instead of accounting for the fact that anthropic deadlocks assistant and user! Also there are one or two errors but the reason it may not have worked on your end aside from roles is the fact that I was supposed to put client = Anthropic(api_key=ANTHROPIC_API_KEY) but accidently removed everything in the ()! Fixing now and I already got sync working again, it should take a few minutes at least and an hour or two at most (considering I don't get sidetracked)! |
Files selected (1)
Files ignored (0)InstructionsEmoji Descriptions:
Interact with the Bot:
Available Commands:
Tips for Using @Entelligence.AI Effectively:
Need More Help?
|
I made a few fixes which I forgot to account for during development, for example - Accidently using the "computer role" when only "assistant" and "user" role is allowed - Making several major changes for readability - Keeping output strings to show that the scripts work properly and - Properly starting an AgentOps session which properly logs everything
|
Hello @areibman it's all set! I had to handle a few things related to revamping as I realized I made some glaring mistakes that barely kept things working! You can check https://github.com/TKTSWalker/agentops/blob/main/examples/anthropic_examples/anthropic-example-sync.ipynb to see that these were ran and the outputs are still there to view to confirm as working! Besides that, I do believe that the way listed for the tool functions do not work in the default example and want to handle this but it will take a while to properly implement it! For example, I had problems while keeping stream on! Also, the output itself is also not a string but rather a sort of weirdish output (kind of similar to Json)! Even if there is a way to get things from text, the JSON format is extremely valuable as it allows for easily getting the AI's text and related tools! I am completely sure that I can make a system around automatically handling this but it would definitely take a little time and direct focus so i'm hoping to do so in a different push request entirely! |
Files selected (2)
Files ignored (0)InstructionsEmoji Descriptions:
Interact with the Bot:
Available Commands:
Tips for Using @Entelligence.AI Effectively:
Need More Help?
|
![entelligence-ai-pr-reviews[bot]](https://avatars.githubusercontent.com/in/932812?s=60&v=4){kind=small}
Files selected (0)Files ignored (1)
InstructionsEmoji Descriptions:
Interact with the Bot:
Available Commands:
Tips for Using @Entelligence.AI Effectively:
Need More Help?
|
|
Merged-- thanks @TKTSWalker !! |
|
Hey, no problem! Moving onto the next! |
📥 Pull Request
📘 Description
_Changed link on Anthropic example to an anthropic readme I created; aside from the pre existing barebones example, I also included three examples showing how the async, sync and tool function could be used.
The sync example shows a case where a sentence generated by simple mixing and connecting is given as context to create a story. The async example has the LLM generate a script line based off the game "TitanFall", depending on the state of the Titan and the personality. At the same time, 4 UUIDs are generated on async as a simple example. Finally, the tool function gives a piece of "Cyberware" based off the user's requested company. Then with the job role the user wants, the LLM says if it is a good or bad fit.
I purposefully made the examples connect to different media as when starting out, lesser experienced developers often have issues following along without a greater level of explanation and focus is an issue as well. By connecting it to better known medias, it remains more engaging while allowing the reader to possibly create parallels between the instructions being shown and the base concepts they're already aware of.
I am also requesting that I am allowed to use AgentOps to show how to do a few things such as setting up a history to keeping it small but with context through summarization when needed. Then, a few use case examples could be shown to help newer developers both understand how to develop with LLMs and how to do so with AgentOps!_
🧪 Testing
Ran journals on Kaggle piece by piece (Which is why the async example is changed so much aside from having a little more meaningful example) before sending it over to a friend to test as well.