• Introduction
• Prerequisites
  • Install Poetry
• Bootstrap Environment
  • How to use the make file
• Using the Open-Source Models Locally
  • Supported Models
• Supported Response Synthesis strategies
• Example Data
• Build the memory index
• Run the Chatbot
• Run the RAG Chatbot
• How to debug the Streamlit app on Pycharm
• References

This project combines the power of Lama.cpp, LangChain (only used for document chunking and querying the Vector Database, and we plan to eliminate it entirely), Chroma and Streamlit to build:

• a Conversation-aware Chatbot (ChatGPT like experience).
• a RAG (Retrieval-augmented generation) ChatBot.

The RAG Chatbot works by taking a collection of Markdown files as input and, when asked a question, provides the corresponding answer based on the context provided by those files.

The Memory Builder component of the project loads Markdown pages from the docs folder. It then divides these pages into smaller sections, calculates the embeddings (a numerical representation) of these sections with the all-MiniLM-L6-v2 sentence-transformer, and saves them in an embedding database called Chroma for later use.

When a user asks a question, the RAG ChatBot retrieves the most relevant sections from the Embedding database. Since the original question can't be always optimal to retrieve for the LLM, we first prompt an LLM to rewrite the question, then conduct retrieval-augmented reading. The most relevant sections are then used as context to generate the final answer using a local language model (LLM). Additionally, the chatbot is designed to remember previous interactions. It saves the chat history and considers the relevant context from previous conversations to provide more accurate answers.

To deal with context overflows, we implemented three approaches:

• Create And Refine the Context: synthesize a responses sequentially through all retrieved contents.
  • 
• Hierarchical Summarization of Context: generate an answer for each relevant section independently, and then hierarchically combine the answers.
  • 
• Async Hierarchical Summarization of Context: parallelized version of the Hierarchical Summarization of Context which lead to big speedups in response synthesis.

• Python 3.10+
• GPU supporting CUDA 12.1+
• Poetry 1.7.0

Install Poetry with the official installer by following this link.

You must use the current adopted version of Poetry defined here.

If you have poetry already installed and is not the right version, you can downgrade (or upgrade) poetry through:

poetry self update <version>

To easily install the dependencies we created a make file.

• Check: make check
  • Use it to check that which pip3 and which python3 points to the right path.
• Setup:
  • Setup with NVIDIA CUDA acceleration: make setup_cuda
    • Creates an environment and installs all dependencies with NVIDIA CUDA acceleration.
  • Setup with Metal GPU acceleration: make setup_metal
    • Creates an environment and installs all dependencies with Metal GPU acceleration for macOS system only.
• Update: make update
  • Update an environment and installs all updated dependencies.
• Tidy up the code: make tidy
  • Run Ruff check and format.
• Clean: make clean
  • Removes the environment and all cached files.
• Test: make test
  • Runs all tests.
  • Using pytest

We utilize the open-source library llama-cpp-python, a binding for llama-cpp, allowing us to utilize it within a Python environment. llama-cpp serves as a C++ backend designed to work efficiently with transformer-based models. Running the LLMs architecture on a local PC is impossible due to the large (~7 billion) number of parameters. This library enable us to run them either on a CPU or GPU. Additionally, we use the Quantization and 4-bit precision to reduce number of bits required to represent the numbers. The quantized models are stored in GGML/GGUF format.

You could download some Markdown pages from the Blendle Employee Handbook and put them under docs.

Run:

python chatbot/memory_builder.py --chunk-size 1000

To interact with a GUI type:

streamlit run chatbot/chatbot_app.py -- --model openchat-3.6 --max-new-tokens 1024

To interact with a GUI type:

streamlit run chatbot/rag_chatbot_app.py -- --model openchat-3.6 --k 2 --synthesis-strategy async-tree-summarization

• LLMs:
  • Calculating GPU memory for serving LLMs
  • Building Response Synthesis from Scratch
  • GPT in 60 Lines of NumPy
  • Introduction to Weight Quantization
• LLM integration and Modules:
  • LangChain:
    • MarkdownTextSplitter
    • Chroma Integration
    • The Problem With LangChain
• Embeddings:
  • all-MiniLM-L6-v2
    • This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search.
• Vector Databases:
  • Chroma
  • Food Discovery with Qdrant
  • Indexing algorithms:
    • There are many algorithms for building indexes to optimize vector search. Most vector databases implement Hierarchical Navigable Small World (HNSW) and/or Inverted File Index (IVF). Here are some great articles explaining them, and the trade-off between speed, memory and quality:
      • Nearest Neighbor Indexes for Similarity Search
      • Hierarchical Navigable Small World (HNSW)
      • From NVIDIA - Accelerating Vector Search: Using GPU-Powered Indexes with RAPIDS RAFT
      • From NVIDIA - Accelerating Vector Search: Fine-Tuning GPU Index Algorithms

      • PS: Flat indexes (i.e. no optimisation) can be used to maintain 100% recall and precision, at the expense of speed.

• Retrieval Augmented Generation (RAG):
  • Rewrite-Retrieve-Read

    • Because the original query can not be always optimal to retrieve for the LLM, especially in the real world, we first prompt an LLM to rewrite the queries, then conduct retrieval-augmented reading.

  • Rerank
  • Conversational awareness
  • Summarization: Improving RAG quality in LLM apps while minimizing vector storage costs
  • RAG is Dead, Again?
• Chatbot Development:
  • Streamlit:
    • Build a basic LLM chat app
    • Layouts and Containers
    • st.chat_message
    • Add statefulness to apps
      • Why session state is not persisting between refresh?
    • st.cache_resource
    • Handling External Command Line Arguments
  • (Investigate) FastServe - Serve Llama-cpp with FastAPI
  • (Investigate) Ollama
• Text Processing and Cleaning:
  • clean-text
• Open Source Repositories:
  • llama.cpp
  • llama-cpp-python
  • GPT4All
  • chroma
  • Inspirational repos:
    • lit-gpt
    • api-for-open-llm
    • AnythingLLM