Engineered Content Generation via Structured Style Deconstruction

Schematic is a content engineering tool designed to replicate specific professional writing styles. Instead of relying on generic AI prompts, it deconstructs successful LinkedIn posts into structured JSON "blueprints." By utilizing Few-Shot Learning, the system guides Large Language Models (LLMs) to mirror the exact hooks, structure, and tone of a target creator.

• Style Analysis: Extracts metadata such as Topic, Language, and Length from existing content.
• Few-Shot Learning: Injects contextually relevant past posts into the LLM prompt to ensure high-fidelity style replication.
• Customizable Generation: High-level control over post parameters including topic selection and language.
• Streamlit Interface: A clean, interactive dashboard for seamless content creation.

The system processes raw LinkedIn content to extract structural fingerprints (Topic, Language, Length). These are stored and managed to serve as a reference library.

When generating a new post, the engine retrieves specific past examples that match the user's selected criteria. These examples are used for few-shot learning, providing the LLM with a "schematic" of the desired output style.

1. Get your API Key: Obtain a GROQ_API_KEY from the Groq Console.

2. Configure Environment: Create a .env file in the root directory and add your key:

   GROQ_API_KEY=your_api_key_here
   

3. Install Dependencies: pip install -r requirements.txt

4. Run the Application: streamlit run main.py

Developed as a portfolio project focusing on LLM Orchestration and Prompt Engineering.

• Terminology: Using words like "Inference Engine," "Structural Fingerprints," and "Metadata" shows you understand the tech stack.
• Focus on Logic: It moves away from the "Mohan" example and explains the logic of the code (the deconstruction and the few-shot learning).
• Clean Formatting: The use of emojis and clear headers makes it scannable for recruiters who only spend 30 seconds on a repo.

Once you paste this in, would you like me to help you prepare for a technical interview by explaining how your few_shot.py actually selects those posts?