To run a high-quality media channel today, you need a Researcher to find facts, a Writer to draft hooks, an Editor to fix mistakes, and a Designer to make thumbnails. Doing this manually takes hours. Most "AI Writers" just hallucinate generic slop.
NewsAgent Pro is not a chatbot. It is a Multi-Agent Swarm that mimics a real newsroom. It reads the internet, plans a strategy, drafts content, critiques its own work, and designs branded visuals—all in 60 seconds.
- 🔄 Self-Correction Loop: The Critic Agent reads the draft and grades it (1-10). If the score is low, it sends it back to the Writer with specific feedback. It iterates until perfection.
- ⚡ Hyper-Fast Inference: Powered by Groq (Llama 3.3) for sub-second logic and planning.
- 🕵️♂️ Real-Time Truth: Uses Tavily API to scrape live news (last 48 hours), preventing hallucinations.
- 🎨 AI Graphic Design: Uses Flux.1-Schnell to generate cinematic backgrounds, then uses Python Pillow to programmatically overlay "Newsflash" headlines.
The system uses a Stateful Graph (LangGraph) with conditional routing.
graph LR
A[User Input] --> B(🧠 Planner)
B -->|Strategy| C(🕵️♂️ Researcher)
C -->|Facts| D(✍️ Writer)
D --> E{⚖️ Critic}
E -->|Score < 8| D
E -->|Score > 8| F(🎨 Designer)
F -->|Visuals| G[Final Output]
style E fill:#ff9999,stroke:#333,stroke-width:2px,color:black
style F fill:#99ff99,stroke:#333,stroke-width:2px,color:black
| Role | Model / Tool | Function |
|---|---|---|
| Planner | Llama 3.3 (Groq) | Analyzes the topic and determines the "Viral Angle." |
| Researcher | Tavily API | Scrapes the web for facts/quotes from the last 48 hours. |
| Writer | Gemini 2.5 / Groq | Drafts platform-specific content (Threads vs Posts). |
| Critic | Llama 3.3 (Groq) | The Gatekeeper. Rejects low-quality drafts and forces rewrites. |
| Designer | Flux.1-Schnell | Generates 16:9 cinematic cover art in <4 steps. |
| Twitter Thread (Viral Style) | LinkedIn Post (Professional) |
|---|---|
Topic: "AI Agents 2026"
|
Topic: "Venezuela Oil Crisis"
|
| Short, punchy, thread-formatted. | Deep dive, strategic analysis. |
AI agents aren't hype anymore—they're quietly taking over dev workflows in 2026.
(thread 🧵)
|||Groq just dropped free-tier Llama 3.1 70B inference that's faster than most paid APIs.
Sub-100ms latency. No card needed.
This changes everything.
|||LangGraph + critique loops = agents that self-improve until viral-ready.
No more generic slop.
|||Real wins: Teams reporting 40–60% faster prototyping.
The gap between indie hackers and big tech is closing FAST.
|||But risks remain: Hallucinations without strong guardrails.
The best setups route Groq ↔ Gemini for speed + context.
|||2026 prediction: Every company ships internal agent tools.
The ones who master hybrid routing win.
|||I'm building with this exact stack daily.
What's your biggest agent win so far? Reply below 👇
#AI #Agents #Productivity
Try the Production Build on Hugging Face Spaces:
Try searching: "DeepSeek vs OpenAI" or "SpaceX Starship Launch"
git clone https://github.com/Eatosin/NewsAgent-Pro.git
cd NewsAgent-Pro
pip install -r requirements.txtCreate a .env file with your keys (Get Groq for free speed!):
GROQ_API_KEY=gsk_...
GEMINI_API_KEY=AIza...
TAVILY_API_KEY=tvly-...
HF_TOKEN=hf_...streamlit run src/app.pyWe use a custom multi-stage build to handle system dependencies (Fonts, Pillow):
docker build -t newsagent .
docker run -p 7860:7860 --env-file .env newsagentOwadokun Tosin Tobi Senior AI Engineer & Product Builder
Built with the Lexpertz R&D Stack.