GPT_RAG.py

# -*- coding: utf-8 -*-
"""nomic_embedding_rag.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1vAQoZx_07yU0nVCkFxJQkcVeymgNpzFF
"""

!pip install nomic
!pip install --upgrade langchain

! nomic login

! nomic login nk-bqukmTuFJHW8tgXzXXBw1qDL062-pth-ACecKP7CkXs

! pip install -U langchain-nomic langchain_community tiktoken langchain-openai chromadb langchain

# Optional: LangSmith API keys
import os

os.environ["LANGCHAIN_TRACING_V2"] = "true"
os.environ["LANGCHAIN_ENDPOINT"] = "https://api.smith.langchain.com"
os.environ["LANGCHAIN_API_KEY"] = "api_key"

"""## Document Loading

Let's test 3 interesting blog posts.
"""

import json
from langchain_community.document_loaders import JSONLoader
from langchain.docstore.document import Document

# Define el JSONLoader para cargar y procesar cada mensaje del JSON
class JSONLoader:
    def __init__(self, message):
        self.message = message

    def load(self):
        # Crear una instancia de Document con el contenido y metadata adecuada
        return Document(
            page_content=self.message['content'],
            metadata={
                'role': self.message['role'],
                'conversation_id': self.message['conversation_id'],
                'message_id': self.message['message_id']
            }
        )

# Cargar el archivo JSON
file_path = 'RAG_Datos.json'  # Asegúrate de que esta ruta sea correcta

with open(file_path, 'r') as file:
    data = json.load(file)

# Procesar los mensajes y crear los documentos
docs_list = []
for conversation in data:
    for message in conversation['messages']:
        docs_list.append(JSONLoader(message).load())

# Verificar el contenido (opcional)
for doc in docs_list:
    print(doc.page_content, doc.metadata)

"""from langchain_community.document_loaders import WebBaseLoader

urls = [
    "https://lilianweng.github.io/posts/2023-06-23-agent/",
    "https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/",
    "https://lilianweng.github.io/posts/2023-10-25-adv-attack-llm/",
]"""

"""docs = [WebBaseLoader(url).load() for url in urls]""

"""docs_list = [item for sublist in docs for item in sublist]

## Splitting

Long context retrieval,
Chunck_size -> tamaño de cada texto
"""

# Ahora puedes usar docs_list con text_splitter
from langchain.text_splitter import CharacterTextSplitter

text_splitter = CharacterTextSplitter(
    chunk_size=7500, chunk_overlap=100
)
doc_splits = text_splitter.split_documents(docs_list)

# Verificar el contenido de los splits (opcional)
for split in doc_splits:
    print(split.page_content, split.metadata)

import tiktoken

encoding = tiktoken.get_encoding("cl100k_base")
encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")
for d in doc_splits:
    print("The document is %s tokens" % len(encoding.encode(d.page_content)))

"""## Index

Nomic embeddings [here](https://docs.nomic.ai/reference/endpoints/nomic-embed-text).
"""

import os

from langchain_community.vectorstores import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableLambda, RunnablePassthrough
from langchain_nomic import NomicEmbeddings
from langchain_nomic.embeddings import NomicEmbeddings

# Add to vectorDB
vectorstore = Chroma.from_documents(
    documents=doc_splits,
    collection_name="rag-chroma",
    embedding=NomicEmbeddings(model="nomic-embed-text-v1"),
)
retriever = vectorstore.as_retriever()

"""## RAG Chain

We can use the
"""

import os
from sklearn.metrics import precision_score, recall_score, f1_score
from nltk.translate.bleu_score import corpus_bleu
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langchain.chains import LLMChain

# Configurar la clave de API como variable de entorno
os.environ['OPENAI_API_KEY'] = 'sk-proj-OaIQbNSKP2uATHxcaUxhT3BlbkFJi2HSSi4zSHSOw9UjtUWn'

# Prompt
template = """Answer the question based only on the following context:
{context}

Question: {question}
"""
prompt = ChatPromptTemplate.from_template(template)

# LLM API
model = ChatOpenAI(temperature=0, model="gpt-4-1106-preview")

# Placeholder para `retriever`
class DummyRetriever:
    def __call__(self, *args, **kwargs):
        return {"context": "This is a test context"}

retriever = DummyRetriever()

# Crear una cadena LLM
llm_chain = LLMChain(
    prompt=prompt,
    llm=model,
)

# Datos de prueba
test_data = [
    {"context": "Write a Python function to sum all prime numbers up to 1000.", "question": "How to write a function to sum all prime numbers up to 1000?", "expected_answer": "def sum_primes(limit):\n    def is_prime(n):\n        if n <= 1:\n            return False\n        for i in range(2, int(n**0.5) + 1):\n            if n % i == 0:\n                return False\n        return True\n    return sum(x for x in range(limit) if is_prime(x))\n\nprint(sum_primes(1000))"},
    {"context": "Write a Python function to calculate the factorial of a number.", "question": "How to write a function to calculate the factorial of a number?", "expected_answer": "def factorial(n):\n    if n == 0:\n        return 1\n    else:\n        return n * factorial(n-1)\n\nprint(factorial(5))"},
    {"context": "Write a Python function to check if a number is palindrome.", "question": "How to write a function to check if a number is palindrome?", "expected_answer": "def is_palindrome(n):\n    return str(n) == str(n)[::-1]\n\nprint(is_palindrome(121))"},
    {"context": "Write a Python function to generate Fibonacci sequence up to n.", "question": "How to write a function to generate Fibonacci sequence up to n?", "expected_answer": "def fibonacci(n):\n    fib_sequence = [0, 1]\n    while len(fib_sequence) < n:\n        fib_sequence.append(fib_sequence[-1] + fib_sequence[-2])\n    return fib_sequence\n\nprint(fibonacci(10))"},
    {"context": "Write a Python function to find the greatest common divisor (GCD) of two numbers.", "question": "How to write a function to find the greatest common divisor (GCD) of two numbers?", "expected_answer": "def gcd(a, b):\n    while b:\n        a, b = b, a % b\n    return a\n\nprint(gcd(48, 18))"},
    {"context": "Write a Python function to check if a string is an anagram of another string.", "question": "How to write a function to check if a string is an anagram of another string?", "expected_answer": "def is_anagram(str1, str2):\n    return sorted(str1) == sorted(str2)\n\nprint(is_anagram('listen', 'silent'))"},
    {"context": "Write a Python function to find the maximum element in a list.", "question": "How to write a function to find the maximum element in a list?", "expected_answer": "def find_max(lst):\n    return max(lst)\n\nprint(find_max([3, 5, 7, 2, 8]))"},
    {"context": "Write a Python function to reverse a string.", "question": "How to write a function to reverse a string?", "expected_answer": "def reverse_string(s):\n    return s[::-1]\n\nprint(reverse_string('hello'))"},
    {"context": "Write a Python function to merge two sorted lists.", "question": "How to write a function to merge two sorted lists?", "expected_answer": "def merge_sorted_lists(lst1, lst2):\n    return sorted(lst1 + lst2)\n\nprint(merge_sorted_lists([1, 3, 5], [2, 4, 6]))"},
    {"context": "Write a Python function to remove duplicates from a list.", "question": "How to write a function to remove duplicates from a list?", "expected_answer": "def remove_duplicates(lst):\n    return list(set(lst))\n\nprint(remove_duplicates([1, 2, 2, 3, 4, 4, 5]))"},
]

# Evaluar la precisión, recall y F1-score de la recuperación
retrieved_contexts = [retriever()["context"] for _ in test_data]
expected_contexts = [item["context"] for item in test_data]
precision = precision_score(expected_contexts, retrieved_contexts, average='macro', zero_division=1)
recall = recall_score(expected_contexts, retrieved_contexts, average='macro', zero_division=1)
f1 = f1_score(expected_contexts, retrieved_contexts, average='macro')

print(f"Retrieval Precision: {precision}")
print(f"Retrieval Recall: {recall}")
print(f"Retrieval F1 Score: {f1}")

# Evaluar la generación de respuestas
generated_answers = []
for item in test_data:
    output = llm_chain.run({"context": item["context"], "question": item["question"]})
    generated_answers.append(output)

# BLEU Score
reference_answers = [[item["expected_answer"].split()] for item in test_data]
generated_answers_tokens = [answer.split() for answer in generated_answers]
bleu_score = corpus_bleu(reference_answers, generated_answers_tokens)

print(f"BLEU Score: {bleu_score}")