Build RAG Chatbot with Llamaindex, Pgvector, DeepSeek V3, and Ollama all-minilm

Introduction to RAG

Retrieval-Augmented Generation (RAG) is a game-changer for GenAI applications, especially in conversational AI. It combines the power of pre-trained large language models (LLMs) like OpenAI’s GPT with external knowledge sources stored in vector databases such as Milvus and Zilliz Cloud, allowing for more accurate, contextually relevant, and up-to-date response generation. A RAG pipeline usually consists of four basic components: a vector database, an embedding model, an LLM, and a framework.

Key Components We'll Use for This RAG Chatbot

This tutorial shows you how to build a simple RAG chatbot in Python using the following components:

Llamaindex: a data framework that connects large language models (LLMs) with various data sources, enabling efficient retrieval-augmented generation (RAG). It helps structure, index, and query private or external data, optimizing LLM applications for search, chatbots, and analytics.
Pgvector: an open-source extension for PostgreSQL that enables efficient storage and querying of high-dimensional vector data, essential for machine learning and AI applications. Designed to handle embeddings, it supports fast approximate nearest neighbor (ANN) searches using algorithms like HNSW and IVFFlat. Since it is just a vector search add-on to traditional search rather than a purpose-built vector database, it lacks scalability and availability and many other advanced features required by enterprise-level applications. Therefore, if you prefer a much more scalable solution or hate to manage your own infrastructure, we recommend using Zilliz Cloud, which is a fully managed vector database service built on the open-source Milvus and offers a free tier supporting up to 1 million vectors.)
DeepSeek-V3: DeepSeek-V3 is a cutting-edge, open-weight large language model (LLM) with 685 billion parameters, excelling in code generation, mathematical reasoning, and long-context understanding (up to 128K tokens). It adopts the MIT license, enabling free modification, distribution, and commercial use.
Ollama all-minilm: A compact language model optimized for efficient NLP tasks, offering fast inference and low resource consumption. Strengths include scalability, versatility in text processing, and seamless integration. Ideal for real-time applications, edge deployments, and scenarios requiring lightweight yet robust language understanding.

By the end of this tutorial, you’ll have a functional chatbot capable of answering questions based on a custom knowledge base.

Note: Since we may use proprietary models in our tutorials, make sure you have the required API key beforehand.

Step 1: Install and Set Up Llamaindex

pip install llama-index

Step 2: Install and Set Up DeepSeek V3

%pip install llama-index-llms-deepseek

from llama_index.llms.deepseek import DeepSeek

# you can also set DEEPSEEK_API_KEY in your environment variables
llm = DeepSeek(model="deepseek-chat", api_key="you_api_key")

# You might also want to set deepseek as your default llm
# from llama_index.core import Settings
# Settings.llm = llm

Step 3: Install and Set Up Ollama all-minilm

%pip install llama-index-embeddings-ollama

from llama_index.embeddings.ollama import OllamaEmbedding

embed_model = OllamaEmbedding(
    model_name="all-minilm",
)

Step 4: Install and Set Up Pgvector

%pip install llama-index-vector-stores-postgres

from llama_index.core import VectorStoreIndex
from llama_index.vector_stores.postgres import PGVectorStore

vector_store = PGVectorStore.from_params(
    database=db_name,
    host=url.host,
    password=url.password,
    port=url.port,
    user=url.username,
    table_name="your_table_name",
    embed_dim=1536,  # openai embedding dimension
    hnsw_kwargs={
        "hnsw_m": 16,
        "hnsw_ef_construction": 64,
        "hnsw_ef_search": 40,
        "hnsw_dist_method": "vector_cosine_ops",
    },
)

Step 5: Build a RAG Chatbot

Now that you’ve set up all components, let’s start to build a simple chatbot. We’ll use the Milvus introduction doc as a private knowledge base. You can replace it with your own dataset to customize your RAG chatbot.

import requests

from llama_index.core import SimpleDirectoryReader

# load documents
url = 'https://raw.githubusercontent.com/milvus-io/milvus-docs/refs/heads/v2.5.x/site/en/about/overview.md'
example_file = 'example_file.md'  # You can replace it with your own file paths.
response = requests.get(url)
with open(example_file, 'wb') as f:
    f.write(response.content)
documents = SimpleDirectoryReader(
    input_files=[example_file]
).load_data()

print("Document ID:", documents[0].doc_id)

storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context, embed_model=embed_model
)

query_engine = index.as_query_engine(llm=llm)
res = query_engine.query("What is Milvus?")  # You can replace it with your own question.
print(res)

Example output

Milvus is a high-performance, highly scalable vector database designed to operate efficiently across various environments, from personal laptops to large-scale distributed systems. It is available as both open-source software and a cloud service. Milvus excels in managing unstructured data by converting it into numerical vectors through embeddings, which facilitates fast and scalable searches and analytics. The database supports a wide range of data types and offers robust data modeling capabilities, allowing users to organize their data effectively. Additionally, Milvus provides multiple deployment options, including a lightweight version for quick prototyping and a distributed version for handling massive data scales.

Optimization Tips

As you build your RAG system, optimization is key to ensuring peak performance and efficiency. While setting up the components is an essential first step, fine-tuning each one will help you create a solution that works even better and scales seamlessly. In this section, we’ll share some practical tips for optimizing all these components, giving you the edge to build smarter, faster, and more responsive RAG applications.

LlamaIndex optimization tips

To optimize LlamaIndex for a Retrieval-Augmented Generation (RAG) setup, structure your data efficiently using hierarchical indices like tree-based or keyword-table indices for faster retrieval. Use embeddings that align with your use case to improve search relevance. Fine-tune chunk sizes to balance context length and retrieval precision. Enable caching for frequently accessed queries to enhance performance. Optimize metadata filtering to reduce unnecessary search space and improve speed. If using vector databases, ensure indexing strategies align with your query patterns. Implement async processing to handle large-scale document ingestion efficiently. Regularly monitor query performance and adjust indexing parameters as needed for optimal results.

pgvector optimization tips

To optimize pgvector in a Retrieval-Augmented Generation (RAG) setup, consider indexing your vectors using GiST or IVFFlat to significantly speed up search queries and improve retrieval performance. Make sure to leverage parallelization for query execution, allowing multiple queries to be processed simultaneously, especially for large datasets. Optimize memory usage by tuning the vector storage size and using compressed embeddings where possible. To further enhance query speed, implement pre-filtering techniques to narrow down search space before querying. Regularly rebuild indexes to ensure they are up to date with any new data. Fine-tune vectorization models to reduce dimensionality without sacrificing accuracy, thus improving both storage efficiency and retrieval times. Finally, manage resource allocation carefully, utilizing horizontal scaling for larger datasets and offloading intensive operations to dedicated processing units to maintain responsiveness during high-traffic periods.

DeepSeek V3 optimization tips

DeepSeek V3 benefits from optimized retrieval and structured prompting to generate high-quality responses in RAG workflows. Improve retrieval accuracy by using domain-specific embeddings and reranking retrieved documents for relevance. Implement hierarchical chunking to structure long-form documents while keeping inputs within token limits. Use prompt tuning to guide the model’s reasoning and reduce hallucinations. Enable caching for frequently accessed knowledge to minimize latency and API costs. Experiment with retrieval augmentation techniques, such as query expansion, to improve recall. Monitor system performance and refine retrieval parameters continuously to maintain an efficient and cost-effective setup.

Ollama all-minilm optimization tips

Optimize Ollama all-minilm in RAG by preprocessing input text into concise, semantically meaningful chunks (256-512 tokens) to align with its context window. Use metadata filtering during retrieval to prioritize relevant documents and reduce noise. Fine-tune the model on domain-specific data to enhance answer relevance. Adjust temperature (0.2-0.5) and top-p (0.85-0.95) for balanced creativity and accuracy. Cache frequent queries to reduce latency, and employ batch processing for parallel inference. Quantize the model to 8-bit for faster inference with minimal accuracy loss. Regularly evaluate retrieval hit rates and answer quality to iteratively refine the pipeline.

By implementing these tips across your components, you'll be able to enhance the performance and functionality of your RAG system, ensuring it’s optimized for both speed and accuracy. Keep testing, iterating, and refining your setup to stay ahead in the ever-evolving world of AI development.

RAG Cost Calculator: A Free Tool to Calculate Your Cost in Seconds

Estimating the cost of a Retrieval-Augmented Generation (RAG) pipeline involves analyzing expenses across vector storage, compute resources, and API usage. Key cost drivers include vector database queries, embedding generation, and LLM inference.

RAG Cost Calculator is a free tool that quickly estimates the cost of building a RAG pipeline, including chunking, embedding, vector storage/search, and LLM generation. It also helps you identify cost-saving opportunities and achieve up to 10x cost reduction on vector databases with the serverless option.

Calculate your RAG cost now.

Calculate your RAG cost

What Have You Learned?

Wow, what a journey it's been exploring the world of Retrieval-Augmented Generation (RAG) systems! Through this tutorial, you’ve learned how to elegantly weave together a powerful framework with LlamaIndex and a robust vector database like Pgvector, unleashing the true potential of your data. The integration of the DeepSeek V3 LLM and the Ollama all-minilm embedding model shines a light on how these tools can bring your applications to life, enabling you to capture and generate contextually rich responses. Each component plays a unique role: LlamaIndex streamlines your data retrieval, while Pgvector ensures fast and efficient querying. Meanwhile, the synergy between the LLM and embedding model means you’re not just pulling data, but creating meaningful conversations—making your RAG system truly dynamic!

You also picked up some handy optimization tips along the way—these insights can help reduce costs and improve performance. Don’t forget to leverage the free RAG cost calculator we included, a fantastic resource to keep your projects on budget and efficient! Now that you’re equipped with these skills and knowledge, the possibilities are endless! So, get out there and start building, optimizing, and innovating your own RAG applications. The future of intelligent, responsive systems is in your hands, and I can't wait to see what you create! Let’s turn those brilliant ideas into reality!

Further Resources

🌟 In addition to this RAG tutorial, unleash your full potential with these incredible resources to level up your RAG skills.

How to Build a Multimodal RAG | Documentation
How to Enhance the Performance of Your RAG Pipeline
Graph RAG with Milvus | Documentation
How to Evaluate RAG Applications - Zilliz Learn
Generative AI Resource Hub | Zilliz

We'd Love to Hear What You Think!

We’d love to hear your thoughts! 🌟 Leave your questions or comments below or join our vibrant Milvus Discord community to share your experiences, ask questions, or connect with thousands of AI enthusiasts. Your journey matters to us!

If you like this tutorial, show your support by giving our Milvus GitHub repo a star ⭐—it means the world to us and inspires us to keep creating! 💖