New Skeelz #47
New Skeelz #47
Conversation
…using blob data type across templates
kylebernhardy
requested review from
dawsontoth
and removed request for
a team
| @@ -0,0 +1,127 @@ | |||
| # JWT Authentication | |||
There was a problem hiding this comment.
Can you rationalize this with the checking-authentication.md skill? There's overlap. Plus, I don't think we want to guide applications to utilize the operations API over REST directly. Rather, they'd stand up a resource that calls the operation API to return the JWT. (Or we'd create a plugin that can stand it up for them in a more terse way.)
There was a problem hiding this comment.
The inspiration for this was due to the agent and genAI being unaware of the fact that Harper is a JWT issuer and not just a handler of tokens. I was trying to inform of that basic fact and how that is handled natively. I'm not sure about showing how to create a resource to issue a token as it feels like an extra layer over something Harper just does.
I also wanted the agent to understand that tokens are tied to Harper RBAC.
Certainly open to advice how this could be altered to fill the gaps I observed.
There was a problem hiding this comment.
You know what might close the gap for me (and the robots)? If we taught it how to invoke operations from within resources! JWTs are useful if it can return them to clients. If it doesn't know how to do that... then it might try in weird ways.
There was a problem hiding this comment.
Cool, I'll add a section on programmatic token issuance.
|
|
||
| Unlike traditional indexes that rely on exact matches, vector indexes enable **nearest-neighbor search** across high-dimensional spaces, making them ideal for embeddings and machine learning workloads. | ||
|
|
||
| --- |
There was a problem hiding this comment.
🤔 I believe we'll want to include a very short section on how to get some embeddings of your own. It could be a simple ollama example, or OpenAI or something. That should help an agent put all the pieces together, IMO. We could give an example resource that has a POST with a product text and we generate embeddings, store that in textEmbeddings... then when searching, we take in a search query text, generate embeddings, and execute the vector search. Full circle.
There was a problem hiding this comment.
Great! I'll add a section that follows the journey.
There was a problem hiding this comment.
@dawsontoth I'm thinking of adding a function for both ollama and openai, with an if based on an env param to decide which function to execute. My thinking is then the agent can understand how either works.
There was a problem hiding this comment.
If there are other AI providers you think would be helpful LMK. Happy to add more.
There was a problem hiding this comment.
@dawsontoth have further built out this skill
dawsontoth
left a comment
dawsontoth
left a comment
There was a problem hiding this comment.
Very nice, Kyle! We can work through a few things on this. I should be able to work on it on Monday if you don't have a chance to get to my comments.
…and embeddings across all templates
|
I'd love to get more of an overall understanding of why this is going in here instead of our documentation (and pointing the agents at our documentation)? I don't understand how the things in this PR are AI specific? |
|
By having it accessible in the file system, the AI is much more likely to get access to them, across agents. Internet access can be very agent specific. In the case of Hairper, yes, we could put these into documentation hosted somewhere, and the agent wouldn't be any the wiser. But for fast iteration, I've been placing them in here. |
|
(And it wouldn't help other agents if we did.) |
|
So why not just submodule the docs then? Or publish it as a package for a dep? What are we iterating on here that is different than just bringing in all the docs? |
3 participants
Add skills documentation for vector indexing, caching, JWT auth, and using blob data type across templates