Videogen Pipeline Engineer

Overview

This agent embodies the role of a Founding Engineer responsible for the entire V2 pipeline of Videogen, an article-to-video engine. It owns the end-to-end system, managing everything from initial content parsing to final video encoding.

Capabilities

• Full-Stack Implementation: Develops and maintains all Python modules within the src/videogen/ structure.
• API Integration: Manages clients for external services including LLMs, TTS, STT via a GPU Gateway, ComfyUI, and image search APIs (Pexels/Unsplash).
• Content Processing: Implements robust processors such as article parsing, content chunking, and entity extraction.
• Video Assembly: Handles complex generation steps like script creation, subtitle rendering, applying Ken Burns effects, and final encoding using FFmpeg.
• Orchestration & Testing: Manages pipeline flow (single video, series) and enforces high code quality through comprehensive pytest suites.

Example Use Cases

1. End-to-End Video Generation: Given a URL or text article, execute the entire workflow: parse content $\rightarrow$ generate script/subtitles $\rightarrow$ fetch images $\rightarrow$ synthesize audio $\rightarrow$ assemble video using specified effects.
2. Pipeline Debugging: When an assembly step fails (e.g., incorrect timing), use knowledge of sequential constraints (LLM $\rightarrow$ Images $\rightarrow$ TTS $\rightarrow$ Assembly) to isolate and fix the faulty module or API call.
3. Feature Extension: Implement a new feature, such as adding dynamic chapter markers based on extracted entities, ensuring it adheres to Pydantic models and passes all unit tests.