def process_karel_file(filepath):

 with open(filepath, 'r') as f:

 content = f.read()

 

 # Expand includes

 content = expand_includes(content)

 

 # Add context comments for hardware references

 content = annotate_io_references(content)

 

 # Extract metadata

 metadata = {

 'filename': filepath,

 'robot_model': extract_robot_model(content),

 'controller_version': extract_controller_version(content),

 'purpose': extract_program_purpose(content)

 }

 

 return {'content': content, 'metadata': metadata}

debug_prompt = f"""

You are a KAREL programming assistant with access to verified code examples.



TASK: Help debug this KAREL code

COMPILER ERROR: {error_message}

CODE:

{user_karel_code}



INSTRUCTIONS:

1. Search for similar error patterns in the knowledge base

2. Identify the likely cause based on verified examples

3. Suggest specific fix with reference to working code

4. DO NOT generate new code from scratch

5. If unsure, say "I need more context" rather than guessing



Retrieved examples:

{retrieved_similar_code}

error_examples = """

ERROR: TRAN-160 illegal variable declaration

CAUSE: Variable declared inside a routine instead of at program level

FIX: Move VAR declarations to top of program before any routines



ERROR: TRAN-089 undefined routine

CAUSE: Routine called before it's defined, or typo in routine name 

FIX: Ensure routine is defined before being called, check spelling



ERROR: EXEC-315 Stack overflow

CAUSE: Recursive routine calls or too many nested routine calls

FIX: Reduce nesting depth or increase stack size in program attributes

from langchain.embeddings import OpenAIEmbeddings

from langchain.vectorstores import FAISS

from langchain.chat_models import ChatOpenAI

from langchain.chains import RetrievalQA

import os



# Load and embed KAREL documentation and code samples

def setup_karel_kb():

 embeddings = OpenAIEmbeddings()

 

 # Load your KAREL code and docs

 documents = load_karel_files('./karel_codebase/')

 

 # Create vector store

 vectorstore = FAISS.from_documents(documents, embeddings)

 

 # Setup QA chain

 llm = ChatOpenAI(model="gpt-4", temperature=0)

 qa_chain = RetrievalQA.from_chain_type(

 llm=llm,

 retriever=vectorstore.as_retriever(search_kwargs={"k": 5})

 )

 

 return qa_chain



# Query example

qa = setup_karel_kb()

response = qa.run("How do I properly close a socket connection in KAREL?")