Implementing an agent from scratch using an IBM Granite foundation model, LangChain, and watsonx.ai

This repository implements a weather query example application of an agent, including function calling based on the IBM Developer Blog Post Create a LangChain AI Agent in Python using watsonx. Therefore, the endpoint https://wttr.in/ is used as the resource for the weather information.

Related blog post: Implementing LangChain AI Agent with WatsonxLLM for a Weather Queries application

0. Clone the repository

git clone https://github.com/thomassuedbroecker/temp-agent-tools-langchain-watsonx.git
cd agent-tools-langchain-watsonx

1. Environment Setup

Step 1: Generate a virtual Python environment

cd code
python3 -m venv --upgrade-deps venv
source venv/bin/activate

Step 2: Install the needed libraries

python3 -m pip install -qU langchain-ibm
python3 -m pip install -qU langchain
python3 -m pip install langchain_core
# show the agent chain as graph
python3 -m pip install grandalf

Step 3: Generate a .env file for the needed environment variables

cat env_example_template > .env

Insert the values for the two environment variables:

• WATSONX_PROJECT_ID=YOUR_WATSONX_PROJECT_ID
• IBMCLOUD_APIKEY=YOUR_KEY

Content of the environment file.

export IBMCLOUD_APIKEY=YOUR_KEY
export IBMCLOUD_URL="https://iam.cloud.ibm.com/identity/token"

# Watsonx
export WATSONX_URL="https://eu-de.ml.cloud.ibm.com"
export WATSONX_VERSION=2023-05-29
export WATSONX_PROJECT_ID=YOUR_PROJECT_ID

export WATSONX_MIN_NEW_TOKENS=1
export WATSONX_MAX_NEW_TOKENS=300
export WATSONX_LLM_NAME=ibm/granite-13b-chat-v2
export WATSONX_INSTANCE_ID=YOUR_WATSONX_INSTANCE_ID
export WATSONX_USERNAME="exampe_user@exampl.net"

2. Execution

Step 1: Run the example without graph

cd code
bash example_agent_invocation.sh

• Output

> Entering new AgentExecutor chain...


Action:
...json
{
    "action": "weather_service",
    "action_input": {
        "cities": ["LA", "NY"]
    }
}
...
Weather service response:
...json
{
    "result": "LA",
    "cities": ["LA", "NY"]
}
...

[{'city': 'NY', 'temperature': '14'}, {'city': 'LA', 'temperature': '12'}]
 The hottest city is NY with a temperature or 14 degrees Celsius. In the city LA the temperature is 12 degrees Celsius.
Action:
...json
{
    "action": "Final Answer",
    "action_input": "The hottest city is NY with a temperature of 14 degrees Celsius."
}
...




> Finished chain.
16.a Response of the agent executor
{'input': 'Which city is hotter today: LA or NY?', 'history': '', 'output': 'The hottest city is NY with a temperature of 14 degrees Celsius.'}



> Entering new AgentExecutor chain...

{
    "action": "Final Answer",
    "action_input": "The temperature in Berlin is 10 degrees Celsius."
}



> Finished chain.
16.b Response of the agent executor
{'input': 'How hot is it today in Berlin?', 'history': 'Human: Which city is hotter today: LA or NY?\nAI: The hottest city is NY with a temperature of 14 degrees Celsius.', 'output': 'The temperature in Berlin is 10 degrees Celsius.'}



> Entering new AgentExecutor chain...

{
    "action": "Final Answer",
    "action_input": "Weather is a phenomenon that occurs in the atmosphere, primarily near the Earth's surface, and is characterized by a combination of temperature, humidity, precipitation, and wind."
}



> Finished chain.
16.c Response of the agent executor
{'input': 'What is the offical definition of the term weather?', 'history': 'Human: Which city is hotter today: LA or NY?\nAI: The hottest city is NY with a temperature of 14 degrees Celsius.\nHuman: How hot is it today in Berlin?\nAI: The temperature in Berlin is 10 degrees Celsius.', 'output': "Weather is a phenomenon that occurs in the atmosphere, primarily near the Earth's surface, and is characterized by a combination of temperature, humidity, precipitation, and wind."}



> Entering new AgentExecutor chain...

{
    "action": "Final Answer",
    "action_input": "The only way to win a soccer game is to score more goals than the opposing team."
}



> Finished chain.
16.d Response of the agent executor
{'input': 'How to win a soccer game?', 'history': "Human: Which city is hotter today: LA or NY?\nAI: The hottest city is NY with a temperature of 14 degrees Celsius.\nHuman: How hot is it today in Berlin?\nAI: The temperature in Berlin is 10 degrees Celsius.\nHuman: What is the offical definition of the term weather?\nAI: Weather is a phenomenon that occurs in the atmosphere, primarily near the Earth's surface, and is characterized by a combination of temperature, humidity, precipitation, and wind.", 'output': 'The only way to win a soccer game is to score more goals than the opposing team.'}

Step 2: Run the example and show the dependencies graph of the runnables

cd code
bash example_agent_graph_invocation.sh

• Output

...
6. Show the 'simple chain' dependencies graph

  +-------------+    
  | PromptInput |    
  +-------------+    
          *          
          *          
          *          
 +----------------+  
 | PromptTemplate |  
 +----------------+  
          *          
          *          
          *          
   +------------+    
   | WatsonxLLM |    
   +------------+    
          *          
          *          
          *          
+------------------+ 
| WatsonxLLMOutput | 
+------------------+ 
7. Invoke the simple chain by asking a question.
....

16. Show the 'agent chain' dependencies graph 

  +----------------------------------------------+       
  | Parallel<agent_scratchpad,chat_history>Input |       
  +----------------------------------------------+       
               ***        *         ***                  
           ****          *             ****              
         **              *                 **            
+--------+          +--------+          +-------------+  
| Lambda |*         | Lambda |          | Passthrough |  
+--------+ ****     +--------+         *+-------------+  
               ***       *          ***                  
                  ****    *     ****                     
                      **  *   **                         
  +-----------------------------------------------+      
  | Parallel<agent_scratchpad,chat_history>Output |      
  +-----------------------------------------------+      
                          *                              
                          *                              
                          *                              
               +--------------------+                    
               | ChatPromptTemplate |                    
               +--------------------+                    
                          *                              
                          *                              
                          *                              
                   +------------+                        
                   | WatsonxLLM |                        
                   +------------+                        
                          *                              
                          *                              
                          *                              
              +-----------------------+                  
              | JSONAgentOutputParser |                  
              +-----------------------+                  
                          *                              
                          *                              
                          *                              
           +-----------------------------+               
           | JSONAgentOutputParserOutput |               
           +-----------------------------+               
17. Create an agent executor
...