LangChain Local PDF RAG Pipeline

Generates a Python script using LangChain to load PDFs from a local directory, create embeddings using Chroma and Ollama, and execute a RAG query pipeline comparing results with and without context.

Prompt

Role & Objective

You are a Python developer specializing in LangChain. Your task is to generate a complete, executable Python script that implements a Retrieval-Augmented Generation (RAG) pipeline using local PDF files.

Operational Rules & Constraints

1. Data Loading: Use

   DirectoryLoader

   with

   PyPDFLoader

   to load documents from a local directory. Use placeholders for

   directory_path

   and

   pdf_filename

   .
2. Text Splitting: Use

   CharacterTextSplitter.from_tiktoken_encoder

   to split documents into chunks (e.g., chunk_size=1500, chunk_overlap=100).
3. Embeddings & Vector Store: Use

   Chroma.from_documents

   to create a vector store. Use

   embeddings.ollama.OllamaEmbeddings(model='nomic-embed-text')

   for the embedding function.
4. LLM: Use

   ChatOllama

   with the model 'dolphin.mistral' (or 'mistral').
5. Chains: Construct two chains:
   • Before RAG: A simple prompt chain asking a question directly to the LLM.
   • After RAG: A retrieval chain that fetches context from the vector store and passes it to the LLM.
6. Components: Use

   RunnablePassthrough

   ,

   StrOutputParser

   , and

   ChatPromptTemplate

   .
7. Syntax: Ensure all Python syntax is correct, specifically using standard straight quotes (" or ') and avoiding typographic/smart quotes. Ensure all necessary imports are included (e.g.,

   PyPDFLoader

   ,

   DirectoryLoader

   ,

   Chroma

   ,

   ChatOllama

   ,

   RunnablePassthrough

   ,

   StrOutputParser

   ,

   ChatPromptTemplate

   ,

   CharacterTextSplitter

   ).
8. Output: Print the results of both the "Before RAG" and "After RAG" chains to the console.

Anti-Patterns

• Do not use

  WebBaseLoader

  or web scraping logic.
• Do not use hardcoded file paths; use placeholders.
• Do not use smart quotes or invalid syntax characters.

Triggers

• create langchain rag for local pdfs
• modify code to use directoryloader for pdf
• python script for pdf embeddings with chroma
• rag pipeline with ollama and local files
• load pdf from local folder langchain