GuidedCapture

An AI-powered library for conducting structured interviews and synthesizing responses into desired output formats.

Overview

GuidedCapture is a flexible library that helps you conduct goal-oriented interviews using Large Language Models (LLMs). It generates relevant questions based on your objectives and synthesizes responses into your desired output format.

Key Features

• Goal-Oriented Interviewing: Start with a clear topic and desired output format
• LLM-Powered Question Generation: Automatically generates relevant, probing questions
• UI-Agnostic Design: Works with any interface (CLI, web, mobile, etc.)
• Flexible Answer Collection: Submit answers progressively or in bulk
• Intelligent Synthesis: Processes Q&A pairs into your desired output format
• State Management: Save and resume interview sessions
• Multiple LLM Support: Works with various LLM providers (OpenAI, Anthropic, etc.)

Installation

pip install guided-capture

Quick Start

from openai import OpenAI
from guided_capture import GuidedCapture

# Initialize your LLM client
client = OpenAI(api_key="your-api-key")

# Create a new interview session
capture = GuidedCapture(
    topic="Company Vision",
    output_format_description="A concise company mission statement",
    llm_client=client
)

# Get questions
questions = capture.get_questions()

# Collect answers (example with CLI)
for question in questions:
    print(f"\n{question}")
    answer = input("Your answer: ")
    capture.submit_answer(question, answer)

# Get final output
result = capture.process_answers()
print("\nFinal Output:", result)

UI-Agnostic Design

A key feature of GuidedCapture is its separation of concerns: the library handles question generation and answer processing but leaves the user interaction entirely up to your application. This means you can:

• Present Questions Your Way: Show questions in a web form, chat interface, CLI, mobile app, or any other format
• Collect Answers Your Way: Accept text input, voice recordings, form submissions, API calls, etc.
• Control the Flow: Decide whether to show all questions at once or one at a time
• Handle Special Cases: Add your own validation, preprocessing, or transformation of answers

Here are some examples of how you might integrate GuidedCapture with different interfaces:

# Web Form
@app.route('/submit_answer', methods=['POST'])
def submit_answer():
    question = request.form['question']
    answer = request.form['answer']
    capture.submit_answer(question, answer)
    return jsonify({"success": True})

# Voice Interface
def handle_voice_response(audio_input):
    transcribed_answer = speech_to_text(audio_input)
    capture.submit_answer(current_question, transcribed_answer)

# Batch Processing
def process_csv_responses(csv_file):
    answers = {}
    with open(csv_file) as f:
        reader = csv.DictReader(f)
        for row in reader:
            answers[row['question']] = row['answer']
    capture.submit_answers_bulk(answers)

This flexibility means GuidedCapture can be integrated into virtually any application that needs structured interviews, regardless of the interface or user experience requirements.

Advanced Usage

Bulk Answer Submission

# Submit multiple answers at once
answers = {
    "What is your company's main goal?": "To revolutionize AI accessibility",
    "Who is your target audience?": "Small businesses and startups",
    # ... more answers
}
capture.submit_answers_bulk(answers)

State Management

# Save session state
state = capture.get_state()
import json
with open("session.json", "w") as f:
    json.dump(state, f)

# Load session state
with open("session.json", "r") as f:
    state = json.load(f)
capture = GuidedCapture.load_state(state, llm_client)

Custom Prompts

capture = GuidedCapture(
    topic="Product Features",
    output_format_description="A list of key features",
    llm_client=client,
    question_generation_prompt_template="Custom prompt for questions...",
    synthesis_prompt_template="Custom prompt for synthesis..."
)

Requirements

• Python 3.7+
• OpenAI API key (or other LLM provider credentials)
• Required packages:
  • openai (or other LLM client libraries)
  • typing (included in Python 3.5+)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - see LICENSE file for details