Making use of Tools
a guide on using the integrated tools for media creation, enhanced information or to perform actions.
Making Use of Tools
How Tools Work in SShift GPT
SShift GPT is designed to intelligently detect when a tool is needed to enhance your experience. Whether it's fetching live data, generating images, or running calculations, the system will automatically determine when to use the appropriate tool.
Specifying Tools Manually
While the app auto-detects the need for tools, you can also explicitly request a tool (or multiple tools) in your prompt to ensure it is used in the response.
Examples:
Automatic Tool Use:
User: "What's the current price of Aptos?"
SShift GPT: (Automatically fetches live price data for Aptos)
Manually Specifying a Tool:
User: "Use the image generator to create an illustration of a futuristic city."
SShift GPT: (Generates an image as requested)
Using Multiple Tools:
User: "Fetch me the price of BTC and tell me what NFTs are trending."
SShift GPT: (Fetches live BTC price and retrieves a list of trending NFTs)
Best Practices for Using Tools
If you want a tool to be used, mention it explicitly in your prompt.
You can combine multiple tools in a single request for richer responses.
If a tool isn’t triggered automatically but you think it should be, try rephrasing your request to be more direct.
By understanding how tools function within SShift GPT, you can optimize your interactions and get the most out of the platform.
Tool Use Examples
Example 1: Creations
Why Didn’t This Generate an Image?
Example:
User: "Draw apple."
SShift GPT: "To create the image you're looking for, could you provide more details about the style or specific characteristics of the apple you want drawn?"
Explanation:
In this case, the request was too vague for the image generation tool to trigger. The model requested more clarification about the style and context of the apple instead of immediately generating an image. If you want to ensure an image is created, specify the tool and provide more descriptive details, such as:
"Use the image generator to draw a realistic red apple on a wooden table."
"Generate a cartoon-style apple with a smiling face and sunglasses."
By refining your prompts and specifying tool usage more explicitly, you can ensure your requests lead to the desired results.
Example 2: Full Information vs. Specified Data Points
Full Information Request:
User: "Look up Microsoft stock."
SShift GPT: (Fetches a comprehensive stock overview, including price, market cap, P/E ratio, and historical trends.)
Specified Data Points Request:
User: "Look up the price and splits of Microsoft stock."
SShift GPT: (Fetches only the current stock price and historical split data.)
Explanation:
When making a request, specifying the exact data points you need can streamline the response and avoid unnecessary information. If you need full stock details, a broad query like “Look up Microsoft stock” works well. However, if you’re looking for specific aspects, such as price and splits, refining the request ensures a more targeted and efficient result.
Example 3: onchain actions
Example:
Request onchain actions
User: "swap 0.1 APT to USDT on Panora"
SShift GPT: Retrieves the specific transaction details and the required parameters, initiates a call to the browser wallet, prompting it to sign and execute the transaction and provides a detailed explanation of the entire process to the user.
Ask data from onchain
User: "I want to know my balance of USDC"
SShift GPT: Uses the connected wallet address to access the Aptos blockchain after fetches and returns the latest USDC balance information.
Explanation:
Users can issue various onchain transaction commands to SShift GPT, such as functions for swapping, transferring, lending, borrowing, and querying balances. The chat assistant interacts with the blockchain to retrieve necessary data, and when a transaction needs to be executed, it securely passes the required parameters to the user's wallet for signing and execution.
Last updated