In this notebook we'll create an example of an agent that uses Python to solve a problem that an LLM can't solve on its own: counting the number of 'r's in the word "strawberry."
Make sure you have the necessary dependencies installed.
Set up your API keys as an environment variable.
Initialize the ExecPython tool.
Initialize an agent using Anthropic's Claude Haiku model.
llm = ChatAnthropic(model="claude-3-haiku-20240307", temperature=0)
prompt_template = ChatPromptTemplate.from_messages(
[
(
"system",
"You are a helpful assistant. Make sure to use a tool if you need to solve a problem.",
),
("human", "{input}"),
("placeholder", "{agent_scratchpad}"),
]
)
agent = create_tool_calling_agent(llm, tools, prompt_template)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
[1m> Entering new AgentExecutor chain...[0m
[32;1m[1;3m
Invoking: `riza_exec_python` with `{'code': 'word = "strawberry"\nprint(word.count("r"))'}`
responded: [{'id': 'toolu_01JwPLAAqqCNCjVuEnK8Fgut', 'input': {}, 'name': 'riza_exec_python', 'type': 'tool_use', 'index': 0, 'partial_json': '{"code": "word = \\"strawberry\\"\\nprint(word.count(\\"r\\"))"}'}]
[0m[36;1m[1;3m3
[0m[32;1m[1;3m[{'text': '\n\nThe word "strawberry" contains 3 "r" characters.', 'type': 'text', 'index': 0}][0m
[1m> Finished chain.[0m
The word "strawberry" contains 3 "r" characters.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4