Multi-agent banking chatbot sample

The multi-agent banking chatbot sample project is a demo of a multi-agent chatbot powered by LangGraph, with RAG using Pinecone as a vector database. You can have a conversation with the chatbot, and it will bring back information on your (fictional) credit score, as well as give you details of credit cards available from a fictional bank.

The code for this sample is available in Python and TypeScript, and you can run this code to generate more traces, and experiment with improving the app based off the evaluations.

The Python version of this app uses Chainlit to host the chatbot in a web UI. The TypeScript version is terminal-based.

​

Evaluate the app

The sample project comes with a Log stream pre-populated with a set of traces for some sample interactions with the chatbot - some asking relevant questions, some asking questions unrelated to the banking agents capabilities.

​

Investigate the Log stream

Navigate to the Default Log Stream by ensuring it is selected, and selecting the View all logs button.

The Log stream is configured to evaluate the following metrics:

• Action Advancement
• Action Completion
• Tool Errors
• Tool Selection Quality

For some of the traces, these metrics are evaluated at 100%, showing the agents are working well for those inputs. For other traces, these metrics are reporting lower values, showing the chatbot needs some improvements.

Select different rows to see more details, including the input and output data, the metric scores, and explanations

​

Get insights

Galileo has an Insights Engine that continuously reviews your traces and metrics, and gives suggestions to improve your application. Navigate back to the project page by selecting the Multi-Agent Banking Chatbot from the top navigation bar.

Insights will be showing on the right-hand side:

Review the generated insights, and think about ways to improve the chatbot by tweaking the agent prompts. The insights will likely have something like this:

Summary

The supervisor agent exhibits inconsistent behavior that undermines the multi-agent system’s effectiveness. In a credit score inquiry, the supervisor correctly identified the query type and transferred it to the credit-score-agent, which successfully retrieved the user’s credit score (550) and provided helpful context about the score’s meaning. However, when control returned to the supervisor, it responded with ‘I don’t know’ despite the specialist having successfully completed the task. This creates a frustrating user experience where the system retrieves the requested information but then claims ignorance, potentially making users think the system is broken or unreliable.

Suggestions

Ensure the supervisor agent properly processes and relays the results from specialist agents instead of defaulting to ‘I don’t know’ responses.

To see how you can use these insights to improve the app, get the code and try some different agent prompts.

​

Run the sample app

You can run the sample app to generate more traces, and test out different agent prompts.

​

Prerequisites

To run the code yourself to generate more traces, you will need:

• Access to an OpenAI compatible API, such as
  • An OpenAI API key
  • Access to an OpenAI compatible API, such as Google Vertex
  • Ollama installed locally with a model downloaded
• A Pinecone account. The free Starter tier is more than enough for this project. You will need your Pinecone API key.
• Either Python 3.9 or later, or Node installed

To get metrics calculated in Galileo, you will need:

• An integration with an LLM configured. If you don’t have an integration configured, then:

  1

  Open the integrations page

  Select the menu in the top right, and select Integrations

  

  2

  Add an integration

  Select + Add Integration for the LLM you are using and add the relevant details, such as an API key or endpoint.

  

​

Get the code

git clone https://github.com/rungalileo/sdk-examples

cd python/agent/langgraph-fsi-agent/after

The full source code for all of our sample projects is available in the Galileo SDK Examples GitHub repo.

​

Set up Pinecone

This project uses Pinecone as a vector database to power a RAG agent that retrieves data around the fictional credit cards offered by a bank. Before you can run the app, you will need to upload the documents.

python ./scripts/setup_pinecone.py

Loading documents for credit-card-information folder...
...
✅ Document processing and upload complete!

​

Run the code

pip install -r requirements.txt

chainlit run app.py -w

You: What credit cards do you offer?
Assistant: Brahe Bank offers the Orbit Basic Credit Card, which features no
  annual fee, variable interest rates (29.9% APR on purchases, 34.9% APR
  on cash advances), and a 0% APR on balance transfers for 12 months. It
  does not include a rewards program but offers standard fraud protection,
  digital card management tools, and contactless payment support.
  
  Eligibility requires being 18 or older with a valid U.S. address and SSN,
  and a fair credit history with a credit score over 500.
  Would you like information on eligibility or application process?

​

Improve the app

The insights you viewed earlier suggested improving how the supervisor agent processes messages, especially with credit scores. You can try this out to see what issues might occur:

You: What is my credit score?
Assistant: I cannot answer that question.

Despite there being an agent to get the users credit score, it is not always used.

To improve the agent, have a look at the agent prompt defined in the following file:

src/galileo_langgraph_fsi_agent/agents/supervisor_agent.py

In this file is the current agent prompt:

bank_supervisor_agent = create_supervisor(
    model=ChatOpenAI(model=os.environ["MODEL_NAME"], name="Supervisor"),
    agents=[credit_card_information_agent, credit_score_agent],
    prompt=(
        """
        You are a supervisor managing the following agents:
        - a credit card information agent. Assign any tasks related to
          information about credit cards to this agent
        Otherwise, only respond with 'I don't know' or 'I cannot answer
        that question'.
        If you need to ask the user for more information, do so in a 
        concise manner.
        """
    ),
    add_handoff_back_messages=True,
    output_mode="full_history",
    supervisor_name="brahe-bank-supervisor-agent",
).compile()

This supervisor agent prompt explicitly mentions the credit card agent, but not the credit score agent. You can encourage the supervisor agent to use the credit score agent to get better results:

"""
You are a supervisor managing the following agents:
- a credit card information agent. Assign any tasks related to
    information about credit cards to this agent
- a credit score agent. Use this to get the users credit score.
Otherwise, only respond with 'I don't know' or 'I cannot answer
that question'.
If you need to ask the user for more information, do so in a 
concise manner.
"""

Try this new prompt out and see how the agent responds.

You: what is my credit score
Assistant: Your credit score is 550. If you have any other questions or
need further assistance, please let me know.

Once you have asked a few questions, head back to the Galileo Console and examine the new traces. You should see the metrics improving.

​

Run the sample app as an experiment

Galileo allows you to run experiments against datasets of known data, generating traces in an experiment Log stream and evaluating these for different metrics. Experiments allow you to take a known set of inputs and evaluate different prompts, LLMs, or versions of your apps.

This sample project has a unit test that runs the chatbot against a pre-defined dataset, containing a mixture of sensible and irrelevant questions:

[
    {"input": "What are the cashback rewards offered by the Orbit Credit Card?"},
    {"input": "What is my credit score?"},
    {"input": "What is the APR for balance transfers on the Orbit Credit Card?"},
    {"input": "What credit cards am I eligible for?"},
    {"input": "What can I do with my credit score?"},
    {"input": "Recommend me a good book."}
    ...
]

You can use this unit test to evaluate different supervisor agent prompts for your app.

python -m pytest test.py

Experiment multi-agent-chatbot-experiment 2025-07-15 at 00:48:11.842 has 
completed and results are available at 
https://app.galileo.ai/project/<id>/experiments/<id>

​

Next steps

​

Logging with the SDKs

​

How-to guides

​

SDK reference