ECIU_AI_Hackathon

This notebook demonstrates the construction and execution of a state graph for an interview and document retrieval process. The graph is built using the StateGraph class and involves several nodes and edges to manage the flow of the process.

Table of Contents

• Installation
• Usage
• Graph Nodes
• Graph Edges
• Graph Compilation
• Visualization

Installation

To run this notebook, you need to have the following dependencies installed:

• IPython
• langchain
• pydantic
• PIL (optional, for graph visualization)

You can install these dependencies using pip:

pip install ipython langchain pydantic pillow

State Graph for ECIU AI Hackathon

Usage

Build the Graph

The graph is constructed using the StateGraph class. Nodes and edges are added to define the flow of the process.

Define Nodes

Each node represents a specific step in the process:

• greet: Handles the initial greeting of the user.
• judge_node: (Commented out) Would handle the judging logic.
• answer_interview: Handles answering interview questions.
• create_context: Creates the context for document retrieval.
• retrieve_documents: Retrieves relevant documents based on the context.
• rerank_documents: Reranks the retrieved documents.
• llm_answer: Generates answers using a language model.
• interview: Handles the interview process.

Define Edges

Edges define transitions between nodes. Conditional edges determine the next step based on the output of the current node.

• START -> greet: The process starts with the greeting node.
• greet -> judge_node: Conditional edges determine whether to proceed to the interview or document retrieval based on the output of the greeting node.
• interview -> answer_interview: Transition from the interview node to the answer interview node.
• answer_interview -> create_context: Transition from the answer interview node to the create context node.
• create_context -> retrieve_documents: Transition from the create context node to the retrieve documents node.
• retrieve_documents -> rerank_documents: Transition from the retrieve documents node to the rerank documents node.
• rerank_documents -> llm_answer: Transition from the rerank documents node to the LLM answer node.
• llm_answer -> END: The process ends after generating the answer using the LLM.

Graph Compilation

The graph is compiled using the compile method of the StateGraph class.

Visualization

The graph can be visualized using the draw_mermaid_png method from the IPython.display module.

Conclusion

This document demonstrates the construction and execution of a state graph for an interview and document retrieval process. The graph is built using the StateGraph class and involves several nodes and edges to manage the flow of the process. The graph can be visualized using the draw_mermaid_png method from the IPython.display module.