campaign-generator 0.1.0

A FastAPI service for campaign generation, competitor research, and audio content.

Campaign Generator

Campaign Generator is a Python FastAPI application for creating content based on competitor research and user-generated audio. It provides a web interface for generating questions, summarizing text, transcribing audio, and fetching the latest news. The app uses local LLMs via Ollama and supports both Mac and Linux environments.

Features

• Audio Transcription: Upload or record audio and transcribe it using Whisper (macOS: whisper-mps, Linux: faster-whisper).
• Text Summarization: Summarize any text using local LLMs (Gemma 3 1B, GPT-OSS 20B) via Ollama.
• Question Generation: Generate questions from text using LLMs.
• Latest News: Fetch recent news on any topic.
• Modern Web UI: Beautiful interface built with Tailwind CSS.
• OS-aware: Automatically selects the best transcription backend and Ollama API endpoint for your platform.

Prerequisites

• Python 3.10+
• Homebrew (for macOS)
• pipx
• ffmpeg (for audio conversion)
• Ollama installed and running with required models (gemma3:1b, gpt-oss:20b)

Install system dependencies (macOS)

brew install pipx
brew install ffmpeg
pipx ensurepath
source ~/.zshrc  # or restart your terminal

Install with pipx

pipx install .

Usage

Start the app with:

campaign-generator

By default, the API will run on http://localhost:8080.

Open your browser and navigate to /frontend to use the web interface:

http://localhost:8080/frontend

How It Works

• Transcription: Uses whisper-mps on macOS and faster-whisper on Linux for audio transcription.
• LLM Integration: Connects to Ollama API for text summarization and question generation. The API endpoint is chosen automatically based on your OS.
• Templates: The web UI is served from templates/home.html.

Development

• All source code is in the root and routers/ directory.
• The entrypoint is main.py, which runs the FastAPI app.
• You can also run with Docker or Docker Compose (see Dockerfile and docker-compose.yml).

Troubleshooting

• If you see errors about missing templates, ensure you installed with pipx after running pipx ensurepath and that your terminal session is up to date.
• Make sure Ollama is running and the required models are pulled.
• If you see errors about ffmpeg, make sure it is installed and available in your PATH.

License

MIT