AI Research Assistant & Podcaster

A FastAPI-powered web application that generates comprehensive research reports and converts them into engaging podcast-style discussions using OpenAI's text-to-speech capabilities. Also allows you to download the report if reading is preferred.

---

Overview

This application transforms research topics into dynamic content through multiple stages:

1. Research Analysis: Generates a team of AI analysts with diverse perspectives/personas
2. Report Generation: Creates a detailed research report with citations
3. Podcast Conversion: Transforms the report into a natural conversation
4. Audio Synthesis: Converts the podcast script into spoken audio using distinct voices for each analyst

Research Graph Visualization

Below is a visualization of the research workflow and how different components interact:

Setup Instructions

Clone and install

git clone https://github.com/andrewlee977/scout.git
cd scout
poetry install

Set up environment variables Create a .env file in the root directory:

OPENAI_API_KEY=your_openai_api_key
NEWS_API_KEY=your_newsapi_key
TAVILY_API_KEY=your_tavily_key

Get your API keys from:

• OpenAI
• NewsAPI
• Tavily

Run the app

poetry run uvicorn app.main:app --reload

Features

Backend (FastAPI)

• Dynamic analyst generation with customizable roles and perspectives
• Research report generation with structured sections (introduction, insights, conclusion)
• Podcast script generation with natural dialogue
• OpenAI TTS integration with voice personality matching
• PDF report generation and download functionality

Frontend

• Clean, responsive HTML interface
• Real-time loading indicators
• Custom audio player with progress tracking
• Interactive feedback system for analyst selection
• Download options for research reports

Audio Processing

• Multi-voice podcast generation using OpenAI's TTS API
• Voice assignment based on analyst gender and role
• Custom voice instructions for personality matching
• Automatic audio segment combination with natural pauses

Research Workflow

• Initial topic submission by the user
• Analyst team generation with feedback options
• Report generation with cited sources
• Podcast script creation
• Multi-voice audio synthesis

Technical Features

• FastAPI for async request handling
• Jinja2 templating for dynamic HTML
• PDF report generation with FPDF
• Audio processing with pydub
• Structured logging system
• State management for research workflow