LangChain Unleashed: A Guide To Using Open Source LLM Models

Explore how prompts guide language models using prompt templates for dynamic customization with variables and placeholders. Learn how to construct a chain to apply these prompts in practice.

Learn to add memory to a chatbot by using LangChain's conversation buffer memory with a chat history key, load memory variables via runnable lambda, and save context after each exchange.

Understand retrieval augmented generation, an framework grounding large language models with external data. Follow the workflow: load, split, embed, store in a LangChain vector store, and retrieve with similarity search.

Use mini lm6 v2 open source embeddings from Hugging Face to map text into a 384-dimensional vector, then store and retrieve them with Chroma DB via LangChain embeddings.

Initialize db Init.py to load sample.txt with Lang change document loaders, split into 400-chunk pieces with 100 overlap, create embeddings with all mintum l6 v2 and persist data with chroma.

Build a retrieval augmented generation pipeline in code using lang chain components like llama cpp, hugging face embeddings, and chroma vector store to answer from context.

Explore how agents use a language model as a reasoning engine to decide actions and their order, customize with tools and toolkits, and communicate decisions via xml based chains.

Create a Kaggle notebook, install transformers and related libraries, load the Mistral 7b Instruct model with auto tokenizer and auto model, and stream output via a Hugging Face pipeline.

Explore building an XML agent with LangChain by wiring a word count tool, pulling prompts from LangChain hub, and executing the agent with a configured executor.

Explore how the ReAct agent uses react prompting to couple reasoning and action in a long chain framework, with the language model guiding steps and order.

Build a mini chatbot using a react frontend template with buttons and text areas, sending post requests to chat endpoint; use FastAPI and LangChain to generate and stream output back.

Set up GitHub Codespaces with react template on virtual machine, verify 16gb ram, clone repo into chat, rename folders frontend and backend, install dependencies, and run app on port 5173.

Open a terminal, navigate to chat and back end, install LangChain, llama cpp, Python, and faster pi, create a models directory, and download gemma seven with wget.

Configure a back-end script to generate text with gemma 7b using llama cpp via langchain community lms. Set the model path and max tokens to 50, then invoke and print.

Create a fast api app with cors middleware for localhost 5173 and credentials, install an auto formatter, and build a chat prompt template with a language model, stopping at Q.

Create a faster API chat endpoint with app.post, converting requests to JSON, and streaming LLM output via a generator and server-sent events using FastAPI streaming response.

Configure allow origins to asterisk to make the url accessible, run uvicorn with reload, publish port 8000, and update the frontend to use the copied url with /chat.