Simple RAG

simple RAG using VARAG¶

Requirement to RUN this notebook - CPU or T4(if using OCR and need fast OCR)

In [ ]:

!git clone https://github.com/adithya-s-k/VARAG
%cd VARAG
%pwd

!git clone https://github.com/adithya-s-k/VARAG %cd VARAG %pwd

In [ ]:

!apt-get update && apt-get install -y && apt-get install -y poppler-utils

!apt-get update && apt-get install -y && apt-get install -y poppler-utils

In [ ]:

%pip install -e .

## We will be using Docling for OCR
%pip install docling

%pip install -e . ## We will be using Docling for OCR %pip install docling

In [1]:

from sentence_transformers import SentenceTransformer
from varag.rag import SimpleRAG
from varag.llms import OpenAI
from varag.llms import LiteLLM
from varag.chunking import FixedTokenChunker
import lancedb
import os
from dotenv import load_dotenv

# os.environ["OPENAI_API_KEY"] = "api-key"

load_dotenv()

from sentence_transformers import SentenceTransformer from varag.rag import SimpleRAG from varag.llms import OpenAI from varag.llms import LiteLLM from varag.chunking import FixedTokenChunker import lancedb import os from dotenv import load_dotenv # os.environ["OPENAI_API_KEY"] = "api-key" load_dotenv()

/home/adithya/miniconda3/lib/python3.11/site-packages/sentence_transformers/cross_encoder/CrossEncoder.py:13: TqdmExperimentalWarning: Using `tqdm.autonotebook.tqdm` in notebook mode. Use `tqdm.tqdm` instead to force console mode (e.g. in jupyter console)
  from tqdm.autonotebook import tqdm, trange

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[1], line 2
      1 from sentence_transformers import SentenceTransformer
----> 2 from varag.rag import SimpleRAG
      3 from varag.llms import OpenAI
      4 from varag.llms import LiteLLM

File ~/workspace/VARAG/varag/rag/__init__.py:1
----> 1 from ._simpleRAG import SimpleRAG
      2 from ._colpaliRAG import ColpaliRAG
      3 from ._hybridColpaliRAG import HybridColpaliRAG

File ~/workspace/VARAG/varag/rag/_simpleRAG.py:16
     14 from openai import OpenAI
     15 from dotenv import load_dotenv
---> 16 from varag.chunking import BaseChunker, FixedTokenChunker
     17 from sklearn.metrics import precision_score, recall_score, f1_score
     18 import pandas as pd

ModuleNotFoundError: No module named 'varag.chunking'

In [ ]:

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

In [ ]:

# Initialize OpenAI LLM
# llm = LiteLLM(model="gpt-4o-mini" , is_vision_required=True, api_key=OPENAI_API_KEY) 
llm = LiteLLM(model="gpt-3.5-turbo" , is_vision_required=True, api_key=OPENAI_API_KEY)

# Initialize OpenAI LLM # llm = LiteLLM(model="gpt-4o-mini" , is_vision_required=True, api_key=OPENAI_API_KEY) llm = LiteLLM(model="gpt-3.5-turbo" , is_vision_required=True, api_key=OPENAI_API_KEY)

In [ ]:

response = llm.query(query="What is you name?" , )
print(response)

response = llm.query(query="What is you name?" , ) print(response)

In [ ]:

embedding_model = SentenceTransformer("all-MiniLM-L6-v2", trust_remote_code=True)
# embedding_model = SentenceTransformer("BAAI/bge-base-en", trust_remote_code=True)
# embedding_model = SentenceTransformer("BAAI/bge-large-en-v1.5", trust_remote_code=True)
# embedding_model = SentenceTransformer("BAAI/bge-small-en-v1.5", trust_remote_code=True)

# Initialize shared database
shared_db = lancedb.connect("~/shared_rag_db")

# Initialize TextRAG with shared database
text_rag = SimpleRAG(
    text_embedding_model=embedding_model,
    db=shared_db,
    table_name="textDemo",
)

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
# Initialize OpenAI LLM
llm = LiteLLM(model="gpt-4o-mini" , is_vision_required=True, api_key=OPENAI_API_KEY)

embedding_model = SentenceTransformer("all-MiniLM-L6-v2", trust_remote_code=True) # embedding_model = SentenceTransformer("BAAI/bge-base-en", trust_remote_code=True) # embedding_model = SentenceTransformer("BAAI/bge-large-en-v1.5", trust_remote_code=True) # embedding_model = SentenceTransformer("BAAI/bge-small-en-v1.5", trust_remote_code=True) # Initialize shared database shared_db = lancedb.connect("~/shared_rag_db") # Initialize TextRAG with shared database text_rag = SimpleRAG( text_embedding_model=embedding_model, db=shared_db, table_name="textDemo", ) OPENAI_API_KEY = os.getenv("OPENAI_API_KEY") # Initialize OpenAI LLM llm = LiteLLM(model="gpt-4o-mini" , is_vision_required=True, api_key=OPENAI_API_KEY)

In [ ]:

text_rag.index(
        "./examples/data",
        recursive=False,
        chunking_strategy=FixedTokenChunker(chunk_size=1000),
        metadata={"source": "gradio_upload"},
        overwrite=True,
        verbose=True,
        ocr=True,
    )

text_rag.index( "./examples/data", recursive=False, chunking_strategy=FixedTokenChunker(chunk_size=1000), metadata={"source": "gradio_upload"}, overwrite=True, verbose=True, ocr=True, )

In [ ]:

query = "what is colpali ?"
num_results = 5

search_results = text_rag.search(query, k=num_results)

print("This was the retrieved Context")
for i, r in enumerate(search_results):
    print(f"{'==='*50}")
    print(f"\n\nChunk {i+1}:")
    print(f"Text: {r['text']}")
    print(f"Chunk Index: {r['chunk_index']}")
    print(f"Document Name: {r['document_name']}")
    print(f"\n\n{'==='*50}")

query = "what is colpali ?" num_results = 5 search_results = text_rag.search(query, k=num_results) print("This was the retrieved Context") for i, r in enumerate(search_results): print(f"{'==='*50}") print(f"\n\nChunk {i+1}:") print(f"Text: {r['text']}") print(f"Chunk Index: {r['chunk_index']}") print(f"Document Name: {r['document_name']}") print(f"\n\n{'==='*50}")

In [ ]:

from IPython.display import display, Markdown, Latex

context = "\n".join([r["text"] for r in search_results])
response = llm.query(
    context=context,
    system_prompt="Given the below information answer the questions",
    query=query,
)


display(Markdown(response))

from IPython.display import display, Markdown, Latex context = "\n".join([r["text"] for r in search_results]) response = llm.query( context=context, system_prompt="Given the below information answer the questions", query=query, ) display(Markdown(response))

Run Gradio Demo¶

In [ ]:

%cd examples
!python textDemo.py --share

%cd examples !python textDemo.py --share