Reference

__main__.pychain.pyfile_io.pyutils.py

main.py

Entry point that passes info read from files to chain, then passes LLM result to be written

`main()`

Create chain, read info from files, append generated questions, then write to new file

Source code in src/chatbot_util/__main__.py

def main():
    """Create chain, read info from files, append generated questions, then write to new file"""
    # Expanded file paths
    filenames = {
        "readfile": os.path.expanduser("~/.chatbot-util/Chatbot FAQ - Enter Here.csv"),
        "readfile2": os.path.expanduser("~/.chatbot-util/Other.txt"),
        "writefile": os.path.expanduser("~/.chatbot-util/Permutated.csv"),
    }

    # Read info, generate questions, then write final output
    print("\nReading topics, questions, employees, and answers...")
    store, employees, phrases, answers, nums = file_io.read(filenames)
    permutated_store = chain.generate(store, phrases)
    print(f'\nWriting to "{filenames["writefile"]}"...')
    file_io.write(filenames, permutated_store, employees, answers, nums)
    print("Done.\n")

chain.py

Setup language model and output parser, then generate and append new questions

`generate(store, phrases)`

Generate and append new questions to store

Source code in src/chatbot_util/chain.py

def generate(store, phrases):
    """Generate and append new questions to store"""

    def progress(index, total):
        """Updates generation progress"""
        sys.stdout.write(f"\rGenerating similar queries for: {index}/{total}...")
        sys.stdout.flush()

    # Calculate total number of questions to generate
    index, total = 1, 0
    for topic in store:
        for question in store[topic]:
            total += 1

    for topic in store:
        new_questions = []
        for question in store[topic]:
            progress(index, total)
            prompt = INSTRUCTION + question
            new_questions.append(invoke(prompt, phrases))
            index += 1
        for new_sub_question in new_questions:
            for new_question in new_sub_question:
                store[topic].append(new_question)
    return store

`invoke(prompt, phrases)`

Define chat model, then create the chain

Source code in src/chatbot_util/chain.py

def invoke(prompt, phrases):
    """Define chat model, then create the chain"""
    options = {"seed": 39}
    response = ollama.generate(
        model="mistral",
        prompt=prompt,
        options=options,
    )

    cleaned_response = parse(response["response"], phrases)
    return cleaned_response

`parse(response, phrases)`

Parse lines from LLM and clean them before returning

Source code in src/chatbot_util/chain.py

def parse(response, phrases):
    """Parse lines from LLM and clean them before returning"""

    def remove_empty(lines):
        """Remove empty lines"""
        for line in lines:
            if line == "":
                lines.remove(line)
        return lines

    def remove_numbers(lines):
        """Remove numbers"""
        for i, line in enumerate(lines):
            if line[1] == ".":
                lines[i] = line[3:]
        return lines

    def remove_phrases(lines, phrases):
        for i, line in enumerate(lines):
            for phrase in phrases:
                line = line.replace(phrase[0], phrase[1])
            lines[i] = line
        return lines

    lines = response.strip().split("\n")
    nonempty_lines = remove_empty(lines)
    no_num_lines = remove_numbers(nonempty_lines)
    cleaned_lines = remove_phrases(no_num_lines, phrases)
    return cleaned_lines