All case studies
AI / Content Automation · Open Source

NewsAgent Pro: autonomous AI newsroom in 60 seconds

Built a fully agentic LangGraph workflow that researches trending stories via NewsAPI + Tavily, drafts articles with Llama 3.3, designs visuals with Flux.1, and enforces quality through a self-correcting Critic Agent loop — all end-to-end in under 60 seconds.

End-to-end latency<60s
Agent nodes5-node pipeline
GitHub stars6

The brief

A working newsroom, not a demo. The pipeline had to go from a trending signal to a publishable draft without a human in the loop, and fail loudly when a source did not hold up.

What we built

A five-node LangGraph circuit: scout (NewsAPI + Tavily), angle, draft (Llama 3.3), art (Flux.1), and a critic that scores the draft against the brief and returns it to the writer when it falls short. Every node writes its own trace, so a weak story is traceable to a specific decision rather than to a vague prompt.

What it proved

End-to-end latency under 60 seconds on a cold run, with the critic loop absorbing most retries. The interesting part is not the speed — it is that the retry budget is bounded, so the system degrades to refusal instead of publishing something it cannot defend.

  • LangGraph
  • Llama 3.3
  • Tavily
  • Flux.1
  • Automation