Making a Information Graph Utilizing an LLM

!pip set up "mirascope[openai]" matplotlib networkx 

import os
from getpass import getpass
os.environ["OPENAI_API_KEY"] = getpass('Enter OpenAI API Key: ')

from pydantic import BaseModel, Subject

class Edge(BaseModel):
    supply: str
    goal: str
    relationship: str

class Node(BaseModel):
    id: str
    sort: str
    properties: dict | None = None

class KnowledgeGraph(BaseModel):
    nodes: listing[Node]
    edges: listing[Edge]

patient_log = """
Mary referred to as for assist at 3:45 AM, reporting that she had fallen whereas going to the toilet. This marks the second fall incident inside per week. She complained of dizziness earlier than the autumn.

Earlier within the day, Mary was noticed wandering the hallway and appeared confused when requested primary questions. She was unable to recall the names of her drugs and requested the identical query a number of occasions.

Mary skipped each lunch and dinner, stating she did not really feel hungry. When the nurse checked her room within the night, Mary was mendacity in mattress with gentle bruising on her left arm and complained of hip ache.

Important indicators taken at 9:00 PM confirmed barely elevated blood stress and a low-grade fever (99.8°F). Nurse additionally famous elevated forgetfulness and attainable indicators of dehydration.

This habits is just like earlier episodes reported final month.
"""

from mirascope.core import openai, prompt_template

@openai.name(mannequin="gpt-4o-mini", response_model=KnowledgeGraph)
@prompt_template(
    """
    SYSTEM:
    Extract a data graph from this affected person log.
    Use Nodes to characterize folks, signs, occasions, and observations.
    Use Edges to characterize relationships like "has symptom", "reported", "famous", and many others.

    The log:
    {log_text}

    Instance:
    Mary stated assist, I've fallen.
    Node(id="Mary", sort="Affected person", properties={{}})
    Node(id="Fall Incident 1", sort="Occasion", properties={{"time": "3:45 AM"}})
    Edge(supply="Mary", goal="Fall Incident 1", relationship="reported")
    """
)
def generate_kg(log_text: str) -> openai.OpenAIDynamicConfig:
    return {"log_text": log_text}
kg = generate_kg(patient_log)
print(kg)

@openai.name(mannequin="gpt-4o-mini")
@prompt_template(
    """
    SYSTEM:
    Use the data graph to reply the person's query.

    Graph:
    {knowledge_graph}

    USER:
    {query}
    """
)
def run(query: str, knowledge_graph: KnowledgeGraph): ...

query = "What well being dangers or issues does Mary exhibit primarily based on her current habits and vitals?"
print(run(query, kg))

import matplotlib.pyplot as plt
import networkx as nx

def render_graph(kg: KnowledgeGraph):
    G = nx.DiGraph()

    for node in kg.nodes:
        G.add_node(node.id, label=node.sort, **(node.properties or {}))

    for edge in kg.edges:
        G.add_edge(edge.supply, edge.goal, label=edge.relationship)

    plt.determine(figsize=(15, 10))
    pos = nx.spring_layout(G)
    nx.draw_networkx_nodes(G, pos, node_size=2000, node_color="lightgreen")
    nx.draw_networkx_edges(G, pos, arrowstyle="->", arrowsize=20)
    nx.draw_networkx_labels(G, pos, font_size=12, font_weight="daring")
    edge_labels = nx.get_edge_attributes(G, "label")
    nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels, font_color="blue")
    plt.title("Healthcare Information Graph", fontsize=15)
    plt.present()

render_graph(kg)