eknowledge 0.1.4

A Python package for executing graph generation from textual inputs.

eKnowledge

eKnowledge is a Python package designed to facilitate the generation of knowledge graphs from textual inputs using various language models. The package leverages the power of NLP to extract relationships and constructs from code snippets or any other structured text.

Installation

To install eKnowledge, you can use pip:

pip install eknowledge

Usage

eKnowledge supports various language models for processing input text, including locally hosted models and remote API-based models. Below is an example using the ChatOllama model, which requires a locally downloaded model from ollama.com. The recommended model for local usage is "codestral:22b-v0.1-q2_K".

Example with Local Language Model (ChatOllama)

from eknowledge import execute_graph_generation
from langchain_community.chat_models import ChatOllama
from langchain_huggingface import HuggingFaceEmbeddings

# Configure the language model
MISTRAL_MODEL = "codestral:22b-v0.1-q2_K"
MAX_TOKENS = 1500

# Initialize the model with the desired configuration
llm = ChatOllama(model=MISTRAL_MODEL, max_tokens=MAX_TOKENS)

# Sample Python code to process
input_text = """
def factorial(x):
    if x == 1:
        return 1
    else:
        return (x * factorial(x-1))

num = 3
print("The factorial of", num, "is", factorial(num))
"""

# Generate the knowledge graph
embed = HuggingFaceEmbeddings()
graph = execute_graph_generation(input_text, llm, embed, max_attempts=1)

print(graph)
# Output: 
# [
#   {
#       'from_node': 'factorial', 
#       'relation': 'depends_on', 
#       'to_node': 'x'
#   }, 
#   {
#       'from_node': 'factorial', 
#       'relation': 'composed_of', 
#       'to_node': 'factorial(x-1)'
#   }, 
#   {
#       'from_node': 'num', 
#       'relation': 'is_a', 
#       'to_node': '3'
#   }, 
#   {
#       'from_node': 'factorial(num)', 
#       'relation': 'used_for', 
#       'to_node': 'print function'
#   }
#]

Input Parameters

The execute_graph_generation function takes several parameters to configure the generation of the knowledge graph. Here are the details of each parameter:

• text (str): The input text from which the knowledge graph is to be generated. This should be a string containing the textual content or code snippet you want to analyze.

• language_model (ChatOllama or other compatible model instances): The language model used to process the input text. This model should be capable of understanding and analyzing the structure of the provided text.

• embedding_model (HuggingFaceEmbeddings instance): This model is used to create embeddings for the text, which are then used to facilitate the retrieval of related text segments and to enhance the context understanding of the language model.

• relations (dict): A dictionary defining the possible relations between nodes in the generated knowledge graph. Each key-value pair in the dictionary specifies a relation type and its associated properties or rules.

• max_attempts (int): The maximum number of iterative attempts the function will make to generate the knowledge graph before stopping. This parameter helps control the execution time and resources.

• process_chunk_size (int): The size of the text chunks that the text is split into for processing. Smaller chunks might lead to more detailed analysis at the cost of potentially higher computational overhead.

• embedding_chunk_size (int): Specifies the size of text chunks used specifically for generating embeddings. This size can impact the granularity of embeddings and thus the detail of the retrieved context.

These parameters allow you to finely tune the knowledge graph generation process to meet specific needs or constraints of your application.

Features

• Supports multiple language models including remote API and local executions.
• Extracts structured knowledge from unstructured text.
• Can be used in various domains like academic research, software development, and data science.

Contributing

Contributions, issues, and feature requests are welcome! Feel free to check the issues page.

License

This project is licensed under the MIT License.

The codestral model is licensed under The Mistral AI Non-Production License.