This is a step-by-step guide to help you create your AI Agents. You'll learn how to configure them, ensuring they are tailored to assist with your specific tasks and meet your unique needs and preferences.
My Agents
How to create my AI Agent
Hover over Agents' section and choose My Agents tab
Click on
+ Create Agentbutton.
A pop-up window will appear, displaying various templates. (Select the one that best suits your needs.)
Important!
Unless it's a New Agent the templates come with a Pre-Made Instructions Prompt. You can always change it!
And that easy you got your new Agent, Now let's customize it!
Agent Personalization
AI Model: Here, you'll find a list of all the available AI Models. Choose the one that suits you the best! You can try to change it depending on the tasks that the agent is carrying out.
Conversation Style: Here, you'll see a bar that you can drag to select how precise or creative you want your agent's answers to be. Feel free to test it and see what option suits you the best! You can always change it!
Instruction Preview: This is essentially your Agent's prompt! Here you'll give them instructions on what's their Mission, how to act, respond, and context!
Example: Here's the pre-made prompt for a Coding-Agent
An AI Agent requires a well-written and specific prompt about their responsibilities in order to fulfill them.
So, we suggest that your agents be assigned 1 or 2 specific tasks. Here is an article that can help you understand more about the structures of prompts: How to make a good prompt
You will also find in this link a trained agent to help you write your prompt: Prompt Agent.
Agent Knowledge
You can add documentation and retrain your chosen model to create domain-specific or knowledge-specific agents. This way, the agent will be able to answer specific questions.
In CodeGPT, you can add 2 types of knowledge:
Code through repositories (Github, Bitbucket, and GitLab)
Documentation from files and URLs
How to add a repository to my new Agent
Click on Agent Knowledge
To upload your repository, you may access the Graphs sub-section.
You can create a new graph or add to an existing one. For more information, see this article.
Select the desired Repository.
Click on Plug Selected Graph.
Now your Agent can access your Github Repository and will be able to provide a tailored response based on your project or use it as a reference.
Important!
Each agent can only support one graph at a time!
How can I add files or docs to my new Agent?
Navigate to the Agent Knowledge section in the chat.
Stay on the Files sub-section.
Select the option to Add New.
You can select an existing file, create one, or add a URL.
Refine Data, ensure that the information you're uploading is correct.
Optional!
If you click on Show advanced settings the following sections will appear next to Content:
Metadata
Metadata
It is a set of data that describes and gives information about other data. ( A summary of your files)
This comes in handy if you want to see the content of your files after you upload them.
If you don't fill in this part, you'll only be able to see the Document ID later, and the content visualization will not appear.
Chunks
Chunks
Chunking involves segmenting large volumes of complex information into smaller, more manageable units.
(Here you'll see how your document is segmented to be more manageable)
You can even customize the Split Settings change the size of the chunks, the overlap and separators.
Confirm the upload by clicking on Start Training.
That's it, you made it!
How do I remove files from my agent?
You can Unplug the file in case you want to use it later or in a different Agent. Or Delete it!
Click on Agent knowledge.
To Delete a file, click on the file and then select the trash-can icon in the upper-right corner.
To unplug a file, click on the file, and the Unplug option will appear. Click on it, and you're done! The file will be saved and available to train another agent.
Conversation Starters
You can craft up to 4 introductory questions for your chatbot. These will appear as interactive buttons, guiding users through their engagement with your chatbot. Here's how:
Click on Conversation Starters.
Fill the required fields and save.
Now the button will appear every time you open your agents chat like this:
Just click on it and your agent will respond
Train your agent with this Question/Answer
If you want to train you agents with Questions-Answers you can do it by clicking on the button Learn of each response
A pop-up will appear where you can edit the agent's response and question and train them to be as custom as you want.
Similarly, when you provide your agents with knowledge, click on Start Training, and your agents will respond the way you taught them.
Chat Skin
Have fun customizing the ✨aesthetics✨ of your agent!
Click on Chat Skin.
Select the color palette you like the most! Change your Agent's Name, Picture and Description!
Chat History
-Here, you'll see a list of all your chat interactions with your agent.
If you select a message from the list, it will open on the right side, and you can see the response, Prompt, Completion, total Tokens, and even Debugging, Embeddings, and Duration.
How did we do? If you want, you can give us your review here: CodeGPT
Join our Discord Community here!: Join the CodeGPT Discord Server!
If you have an issue with this you can document it in: Issues · JudiniLabs/code-gpt-docs
Or feel free to contact us if you need further assistance!