How ibrahimsobh/llms works
• Notebooks / Demo Modules – At the root level you have individual demo notebooks such as “01_Hello_GPT.ipynb”, “02_Hello_BERT.ipynb”, etc. These represent the individual demonstration modules for specific LLMs.
• Agentic Workflows – This directory contains notebooks that illustrate how agents or chains of actions can be implemented (for example, simple language graph demonstrations, title generation, self-ask multi-agents, etc.).
• MultiAgents – This part of the project shows how multiple LLMs or agents can interact, manage contexts, or manage asynchronous operations.
• LMMs – Although similar in purpose to the other demos, these notebooks focus on “Large Multimodal Models”, which could be extended to cover visual as well as textual data in the future.
Relationships and Interactions
• That the user (or developer) interacts with each notebook module. Each notebook serves as a self-contained demonstration where the LLM is loaded, provided with inputs (prompts) and then returns outputs.
• That underlying libraries (such as Hugging Face Transformers, gpt4all library, and even LangChain-like workflows mentioned in the README) serve as “core services” or infrastructure enabling these demos.
• That Agentic Workflows and MultiAgents are layered on top of the base LLM demos; they not only call individual models but also chain actions, manage contexts, and coordinate between multiple models.
• If desired, add an “External Dependencies” component that shows integration points for tools like Google Palm, Streamlit, etc., all referenced in the README examples.
Architectural Patterns and Design Principles
• That the system is organized in a loosely coupled, plug-and-play style where one can run a selected notebook independently while the overall repository offers a comprehensive learning resource covering tasks such as prompt engineering, retrieval augmented generation, and even prompting chains.
Relevant Technologies and Frameworks
• Clearly label technologies like Hugging Face Transformers, gpt4all, and any mentioned frameworks (such as LangChain) which provide context in the README.