!pip install google-search-results

from langchain.agents import load_tools from langchain.agents import initialize_agent from langchain.llms import OpenAI

Load the model

llm = OpenAI(temperature=0)

Load in some tools to use

os.environ["SERPAPI_API_KEY"] = "" tools = load_tools(["serpapi", "llm-math"], llm=llm)

Finally, let's initialize an agent with:

1. The tools

2. The language model

3. The type of agent we want to use.

agent = initialize_agent(tools, llm, agent="zero-shot-react-description", verbose=True)

See list of agents types here

Now let's test it out!

agent.run("Who is the current leader of Japan? What is the largest prime number that is smaller than their age?")

here is the return:

Entering new AgentExecutor chain... I need to find out who the leader of Japan is and then calculate the largest prime number that is smaller than their age. Action: Search Action Input: "current leader of Japan" Observation: Fumio Kishida Thought: I need to find out the age of the leader of Japan Action: Search Action Input: "age of Fumio Kishida" Observation: 65 years Thought: I need to calculate the largest prime number that is smaller than 65 Action: Calculator Action Input: 65

ValueError Traceback (most recent call last) Cell In[18], line 2 1 # Now let's test it out! ----> 2 agent.run("Who is the current leader of Japan? What is the largest prime number that is smaller than their age?")

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/chains/base.py:213, in Chain.run(self, *args, **kwargs) 211 if len(args) != 1: 212 raise ValueError("run supports only one positional argument.") --> 213 return self(args[0])[self.output_keys[0]] 215 if kwargs and not args: 216 return self(kwargs)[self.output_keys[0]]

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/chains/base.py:116, in Chain.call(self, inputs, return_only_outputs) 114 except (KeyboardInterrupt, Exception) as e: 115 self.callback_manager.on_chain_error(e, verbose=self.verbose) --> 116 raise e 117 self.callback_manager.on_chain_end(outputs, verbose=self.verbose) 118 return self.prep_outputs(inputs, outputs, return_only_outputs)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/chains/base.py:113, in Chain.call(self, inputs, return_only_outputs) 107 self.callback_manager.on_chain_start( 108 {"name": self.class.name}, 109 inputs, 110 verbose=self.verbose, 111 ) 112 try: --> 113 outputs = self._call(inputs) 114 except (KeyboardInterrupt, Exception) as e: 115 self.callback_manager.on_chain_error(e, verbose=self.verbose)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/agents/agent.py:792, in AgentExecutor._call(self, inputs) 790 # We now enter the agent loop (until it returns something). 791 while self._should_continue(iterations, time_elapsed): --> 792 next_step_output = self._take_next_step( 793 name_to_tool_map, color_mapping, inputs, intermediate_steps 794 ) 795 if isinstance(next_step_output, AgentFinish): 796 return self._return(next_step_output, intermediate_steps)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/agents/agent.py:695, in AgentExecutor._take_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps) 693 tool_run_kwargs["llm_prefix"] = "" 694 # We then call the tool on the tool input to get an observation --> 695 observation = tool.run( 696 agent_action.tool_input, 697 verbose=self.verbose, 698 color=color, 699 **tool_run_kwargs, 700 ) 701 else: 702 tool_run_kwargs = self.agent.tool_run_logging_kwargs()

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/tools/base.py:107, in BaseTool.run(self, tool_input, verbose, start_color, color, kwargs) 105 except (Exception, KeyboardInterrupt) as e: 106 self.callback_manager.on_toolerror(e, verbose=verbose) --> 107 raise e 108 self.callback_manager.on_toolend( 109 observation, verbose=verbose, color=color, name=self.name, kwargs 110 ) 111 return observation

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/tools/base.py:104, in BaseTool.run(self, tool_input, verbose, start_color, color, *kwargs) 102 try: 103 tool_args, tool_kwargs = _to_args_and_kwargs(tool_input) --> 104 observation = self._run(tool_args, **tool_kwargs) 105 except (Exception, KeyboardInterrupt) as e: 106 self.callback_manager.on_toolerror(e, verbose=verbose)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/agents/tools.py:31, in Tool._run(self, *args, kwargs) 29 def _run(self, *args: Any, *kwargs: Any) -> str: 30 """Use the tool.""" ---> 31 return self.func(args, kwargs)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/chains/llm_math/base.py:130, in LLMMathChain._call(self, inputs) 126 self.callback_manager.on_text(inputs[self.input_key], verbose=self.verbose) 127 llm_output = llm_executor.predict( 128 question=inputs[self.input_key], stop=["```output"] 129 ) --> 130 return self._process_llm_result(llm_output)

File /opt/miniconda3/envs/py38_langchain/lib/python3.8/site-packages/langchain/chains/llm_math/base.py:86, in LLMMathChain._process_llm_result(self, llm_output) 84 answer = "Answer: " + llm_output.split("Answer:")[-1] 85 else: ---> 86 raise ValueError(f"unknown format from LLM: {llm_output}") 87 return {self.output_key: answer}

ValueError: unknown format from LLM: This is not a math problem and cannot be translated into an expression that can be executed using Python's numexpr library.`

gkamradt / langchain-tutorials

could not get the answer 61-- the largest prime number that is smaller than their age #22

Chains: Combine LLMs and prompts in multi-step workflows