search

neuralmesh / apimesh

Serves as the template to enable llms in any gihub project
GNU General Public License v3.0
0 stars 2 forks source link

index available tools toolindexer #11

Open m-c-frank opened 9 months ago

m-c-frank commented 9 months ago

i need some form of index that helps llms understand which tools are available to complete a particular task. the repo itself should also help future completely different projects to scope their own available actions to complete any specific task

available actions

the most important functionality here is the retrieval of information. this should be handled for now by getting the raw content of a file in a github repo.

the script should do the following: you should jsut be able to run curl http://textindexlocation/githubowner/reponame/branchname/pathtofile to get the file content. this makes it github agnostic for the future and can be used for any remote git repository no matter if its hosted on github or on a custom site.

for i just want it to actually get the README of any project i suppose! that already helps in identifying which tools could be relevant, dont you think?

i think a more scalable way is to dockerize this into its own microservice? or im not sure what the best way to go ahead is. the ultimate goal is to have everything run offline anyways using some llm and langchain to do the apimesh routing, and have like a bunch of nice tools to kickstart any project. so in essence i now have a way to get a llm to give me good answers in the comments here but i need to be able to make this scalable and give it tools. my instinct tells me that docker or kubernetes will be the platform of the future because you can run the services on just old phones or something and in the future we will have lots of old phones that are actually really capable of compute.

start by just writing a README that details the utility of actionindexer for creating new tools using the new capabilities of LLMs

the clear and concise README.md of the actionindexer is the following:


# actionindexer
this is a microservice that can be called via a simple http post request with
github-actions[bot] commented 9 months ago

.

actionindexer is a free and open source microservice that can be used to quickly and easily index available tools and actions to complete any task. It can be used as an input for Machine Learning models, such as LLMs, to help them understand the actions available for a particular task.

actionindexer is available as a docker container and can be quickly deployed in any environment. To get started, run the following command: 

docker run -it actionindexer:latest

Once deployed, actionindexer will be listening for HTTP POST requests on port 8080. To request the actions available for a particular task, make a POST request to http://localhost:8080/actions with a body containing a JSON object with the task name.

actionindexer will respond with a JSON containing a list of all the available actions related to the task.

actionindexer is an open source project that is actively maintained and regularly updated with new features. You can find the source code and issue tracker on the actionindexer GitHub repository.