Building My Own RAG With OpenAI, Qdrant and LangChain

import uuid
import os
from openai import OpenAI
from qdrant_client.models import PointStruct, VectorParams, Distance
from qdrant_client import QdrantClient

collection_name = "rag"

texts = [
	"On a Friday evening, Joe Doe, a passionate guitarist, decided to visit a local music shop known for its unique instruments. While browsing, he stumbled upon a vintage guitar that had an intriguing backstory. The shopkeeper shared how the guitar was once owned by a famous musician who had played it during a legendary concert. Inspired by the guitar's history, Joe decided to purchase it, believing it would bring a new depth to his music. That night, he spent hours playing and composing, feeling a deep connection to the instrument and its past.",
	"Rashid Brown, a dedicated volunteer, spent his Friday at a local community center, organizing a food drive for families in need. He rallied a group of friends to help him sort through donations and prepare care packages. As they worked, Rashid shared stories of the families they were helping, emphasizing the importance of community support. By the end of the day, they had assembled over a hundred packages, and Rashid felt a profound sense of fulfillment, knowing they were making a tangible difference in their community.",
	"On a seemingly ordinary Friday, Mark Felix, an office worker, decided to take a different route home from work. As he walked, he discovered a small art gallery hosting an opening night for local artists. Intrigued, he stepped inside and was captivated by the vibrant artwork. He struck up a conversation with one of the artists, who invited him to join a community art class. Mark, who had always enjoyed painting but never pursued it seriously, took the leap and signed up. This spontaneous decision led to a new passion that transformed his weekends into creative explorations."
]

oclient = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

response = oclient.embeddings.create(
    input=texts,
    model="text-embedding-3-small"
)

points = [
    PointStruct(
        id=str(uuid.uuid4()),
        vector=data.embedding,
        payload={"text": text, "doc_type": "stories"},
    )
    for _, (data, text) in enumerate(zip(response.data, texts))
]

qclient = QdrantClient(
	url=os.getenv("QDRANT_URL"),
	api_key=os.getenv("QDRANT_API_KEY")
)

# Create a collection
qclient.create_collection(
    collection_name,
    vectors_config=VectorParams(
        size=1536,  # Size of the embedding vector
        distance=Distance.COSINE  # Distance metric for searching
    )
)

# Insert points into Qdrant
qclient.upsert(collection_name, points)

import os
from openai import OpenAI
from qdrant_client.models import PointStruct
from qdrant_client import QdrantClient, models

collection_name = "rag"

oclient = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

query_vector = oclient.embeddings.create(
    input="Who helped some families on his Friday? Give me his name only..",
    model="text-embedding-3-small"
).data[0].embedding

qclient = QdrantClient(
	url=os.getenv("QDRANT_URL"),
	api_key=os.getenv("QDRANT_API_KEY")
)


# Perform the search
search_results = qclient.search(
    collection_name=collection_name,
    query_vector=query_vector,
    query_filter=models.Filter(
        must=[
            models.FieldCondition(
                key="doc_type",
                match=models.MatchValue(
                    value="stories",
                ),
            )
        ]
    ),
    limit=1
)

# Process search results
for result in search_results:
    print(f"ID: {result.id}, Score: {result.score}, Text: {result.payload['text']}")

import os
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from openai import OpenAI
from qdrant_client.models import PointStruct
from qdrant_client import QdrantClient, models


collection_name = "rag"

oclient = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

query_vector = oclient.embeddings.create(
    input="Who helped some families on his Friday? Give me his name only..",
    model="text-embedding-3-small"
).data[0].embedding

qclient = QdrantClient(
    url=os.getenv("QDRANT_URL"),
    api_key=os.getenv("QDRANT_API_KEY")
)

# Perform the search
search_results = qclient.search(
    collection_name=collection_name,
    query_vector=query_vector,
    query_filter=models.Filter(
        must=[
            models.FieldCondition(
                key="doc_type",
                match=models.MatchValue(
                    value="stories",
                ),
            )
        ]
    ),
    limit=1
)

# Process search results
for result in search_results:
    context = result.payload['text']
    question = "Who helped some families on his Friday? Give me his name only.."

    prompt_template = [
        ("system", "You are a helpful assistant that answers questions based on provided context."),
        ("system", f"context: {context}"),
        ("user", f"answer this question:\n{question}"),
    ]

    prompt = ChatPromptTemplate.from_messages(prompt_template)

    llm = ChatOpenAI(
        openai_api_key=os.getenv("OPENAI_API_KEY"),
        model_name="gpt-4o-mini",
        temperature=0
    )

    chain = prompt | llm | StrOutputParser()

    print(chain.invoke({}))