Skip to main content

​
Overview

Tool Search allows you to discover API actions using natural language queries. This makes it easy to find the right API endpoints without needing to read through documentation or understand complex API specifications.

​
Action Discovery

The first step is to use natural language to discover available actions. You can do this by making a request to the Wild Card API: 
import requests

query = "I need to create a product called Premium Access for $100"
result = requests.get(
    "https://api.wild-card.ai/search",
    params={
        "query": query,
        "collection_id": "<YOUR-COLLECTION-ID>"
    },
    headers={"X-API-Key": "<YOUR-WILDCARD-API-KEY>"}
)
The API will return relevant actions based on your natural language query. These actions are formatted in a way that can be directly used with language models.

​
Agent Integration

Once you have discovered the relevant actions, you can use them with your agent: 
from openai import OpenAI
from agentsjson.core.models.auth import BearerAuthConfig, AuthType
from agentsjson.core.models.schema import AgentsJson
from agentsjson import ToolFormat, execute, get_tools

# Convert the API response into tool definitions
agents_json = AgentsJson(**result.json())
tools = get_tools(agents_json, ToolFormat.OPENAI)

# Set up your OpenAI client
client = OpenAI(api_key="<YOUR-OPENAI-API-KEY>")
completion = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {
            "role": "system",
            "content": "You are a helpful assistant that completes tasks."
        },
        {
            "role": "user",
            "content": query
        }
    ],
    tools=tools
)

# Execute the chosen action with authentication
auth = BearerAuthConfig(type=AuthType.BEARER, token="<YOUR-STRIPE-API-KEY>")
execute(agents_json, completion, ToolFormat.OPENAI, auth)

​
How It Works

  1. Natural Language Query: You start by describing what you want to do in plain English.
  2. Action Discovery: The Wild Card API searches through its collection of API specifications to find relevant actions.
  3. Tool Definition: The discovered actions are converted into a format that language models can understand and use.
  4. Execution: The language model chooses the appropriate action and executes it with the provided authentication.

​
Benefits

  • Simplified API Integration: No need to read through API documentation or understand complex specifications
  • Natural Language Interface: Describe what you want to do in plain English
  • Automatic Tool Discovery: Find the right API endpoints without manual searching
  • Language Model Ready: Tools are automatically formatted for use with OpenAI and other language models

​
Authentication

The tool search and execution process requires two types of authentication:
  1. Wild Card API Key: Used for searching and discovering actions
  2. Service API Key: Used for executing the discovered actions (e.g., Stripe API key)
Make sure to keep these API keys secure and never expose them in your code.
⌘I