Custom tools

1
import os
2
from superagent.client import Superagent
3
4
client = Superagent(
5
    base_url="https://api.beta.superagent.sh",
6
    token=os.environ["SUPERAGENT_API_KEY"]
7
)
8
9
llm = client.llm.create(request={
10
    "provider": "OPENAI",
11
    "apiKey": os.environ["OPENAI_API_KEY"]
12
})
13
14
agent = client.agent.create(
15
    name="Stock Assistant",
16
    description="An Assistant that can fetch stock prices",
17
    avatar="https://mylogo.com/logo.png",  # Replace with a real image
18
    is_active=True,
19
    llm_model="GPT_3_5_TURBO_16K_0613",
20
    initial_message="Hi there, how can I help you?",
21
    prompt="Use the Stock API to answer the users question."
22
)
23
24
client.agent.add_llm(agent_id=agent.data.id, llm_id=llm.data.id)

1
        tool = client.tool.create(
2
    name="Stock API",
3
    description="Useful for answering questions about a specific stock",
4
    type="FUNCTION",
5
    return_direct=False,
6
    metadata={
7
        "functionName": "get-stock",
8
        "args": {
9
            "ticker": {
10
                "type": "string",
11
                "description": "The stock ticker to search for"
12
            }
13
        }
14
    }
15
)
16
17
client.agent.add_tool(agent_id=agent.data.id, tool_id=tool.data.id)

1
prediction = client.agent.invoke(
2
    agent_id=agent.data.id,
3
    enable_streaming=False,
4
    input="What's the current stock price of Apple?",
5
    session_id="my_session_id"
6
)
7
8
output = prediction.data.get("output")
9
steps = prediction.data.get("intermediate_steps")

1
# Implementation of the get_stock function
2
def get_stock(ticker):
3
    print(f"Getting stock information for {ticker}")
4
5
# Create a dispatch table
6
tool_dispatch = {
7
    "get-stock": get_stock,
8
    # Add more tools here as needed
9
}
10
11
# Check the steps and run the function
12
def handle_tool_actions(steps):
13
    for item, _ in steps:
14
        tool_name = item.get('tool')
15
        tool_function = tool_dispatch.get(tool_name)
16
17
        if tool_function:
18
            tool_input = item.get('tool_input', {})
19
            tool_function(**tool_input)
20
        else:
21
            print(f"No function defined for tool: {tool_name}")
22
23
# Run your custom tool
24
handle_tool_actions(steps)

1
import os
2
from superagent.client import Superagent
3
4
client = Superagent(
5
    base_url="https://api.beta.superagent.sh",
6
    token=os.environ["SUPERAGENT_API_KEY"]
7
)
8
9
llm = client.llm.create(request={
10
    "provider": "OPENAI",
11
    "apiKey": os.environ["OPENAI_API_KEY"]
12
})
13
14
agent = client.agent.create(
15
    name="Stock Assistant",
16
    description="An Assistant that can fetch stock prices",
17
    avatar="https://mylogo.com/logo.png",  # Replace with a real image
18
    is_active=True,
19
    llm_model="GPT_3_5_TURBO_16K_0613",
20
    initial_message="Hi there, how can I help you?",
21
    prompt="Use the Stock API to answer the users question."
22
)
23
24
# Create your custom tool
25
tool = client.tool.create(
26
    name="Stock API",
27
    description="Useful for answering questions about a specific stock",
28
    type="FUNCTION",
29
    return_direct=False,
30
    metadata={
31
        "functionName": "get-stock",
32
        "args": {
33
            "ticker": {
34
                "type": "string",
35
                "description": "The stock ticker to search for"
36
            }
37
        }
38
    }
39
)
40
41
client.agent.add_llm(agent_id=agent.data.id, llm_id=llm.data.id)
42
client.agent.add_tool(agent_id=agent.data.id, tool_id=tool.data.id)
43
44
prediction = client.agent.invoke(
45
    agent_id=agent.data.id,
46
    enable_streaming=False,
47
    input="What's the current stock price of Apple?",
48
    session_id="my_session_id"
49
)
50
51
output = prediction.data.get("output")
52
steps = prediction.data.get("intermediate_steps")
53
54
# Implementation of the get_stock function
55
def get_stock(ticker):
56
    print(f"Getting stock information for {ticker}")
57
58
# Create a dispatch table
59
tool_dispatch = {
60
    "get-stock": get_stock,
61
    # Add more tools here as needed
62
}
63
64
# Check the steps and run the function
65
def handle_tool_actions(steps):
66
    for item, _ in steps:
67
        tool_name = item.get('tool')
68
        tool_function = tool_dispatch.get(tool_name)
69
70
        if tool_function:
71
            tool_input = item.get('tool_input', {})
72
            tool_function(**tool_input)
73
        else:
74
            print(f"No function defined for tool: {tool_name}")
75
76
# Run your custom tool
77
handle_tool_actions(steps)