Cybernagle's sharing: [ProConfig generator, state machine for any bot, code widget integration]

[vestal bronze]

for newbie

if you a newbie for proconfig , you can try use pcc cli to help you generate your proconfig.
simply use python tool pip: pip install proconfig
then , use pcc init help you create a fully structured proconfig project.
afterwards, run pcc encode main.yaml , you can get a output.json , and directly put this json on myshell, it will works. 🙂
TLDR:

pip install proconfig 
pip init
pip encode main.yaml
cat output.json 

for senior users

debugging bot sometimes can be really hard, if you have too many states.
now using pcc , you can just run pcc draw your_proconfg.json , and then access your http://localhost:8050
you can see you proconfig state machine, you can try to move each node, zoom in,and zoom out. I've try to analysed April with Her , see picture attached.(I mean, this is the most complicated bot I've ever seen. can't believe it's written in bare json)

for who using code widget

some times , we need integrate several code together, and some times, need encode json into a single string.
with pcc, I've implement !include & !oneline macro.
!include is to include state from /state path
oneline is simple encoder to encode any thing into your json, which means you can use this !oneline to import json, or javascript or other structured file into you proconfig. and then debug your javascirpt or json separately.

or , you can just directly use the code https://github.com/cybernagle/proconfig-generator directly ,feel free to fork if you want impelment it your self. 🙂
with this repo, try to use pcc encode main.yaml, you can directly get the master promting bot's proConfig
https://app.myshell.ai/bot/nMNJNr/1713262101

upon state machine is draw by running pcc draw VN_project.json
which is the state machine of April with Her

[vestal bronze]

using LLM help me analyze big project with complex logic code (ideas)

recently I'm trying to use gpt4o to leverage my daily code reading tasks.
even gpt4o is not works very well when trying to analyze big project. especailly when we are facing a lot of folders, data structures.
thus I'm thinking about , maybe using LLM directly analyze logicially content is not a very good option.
as far as I knwo, LLM is really good at summarize and extract information.
thus, I'm thinking about maybe we can take this advantage to improve my dailly tasks.

precedure

step1

first , when we are facing complex project include a lot logic. we can using this kind of content , and use LLM extract information from such content and reply with a strctured data.
with normal process , we always using regex expression, which it's buggy. but with LLM , it can do much better.

step2

once we build this kind of data structure. next step is to using persistent analyze logic into code(maybe your personal experience about how to read code ?)
and processing this kind of data structure . and transform it into a more LLM facing content.

step 3

give the content code generated , and take this into LLM, let llm summarize it for us .