Model Switching & First Real Use Cases in Newelle (with Ollama)

Model Switching & First Real Use Cases in Newelle (with Ollama)

Date: September 6, 2025
Category: Uncategorized

Body

Following the creation of the practical model menu, the next step was to validate model switching in real use cases.

Tests Performed

  1. General Use (qwen2.5:7b-instruct)
    • Tested with balanced reasoning prompts.
    • Returned correct structured answers with cost-efficient local inference.
  2. Coding / Development (codellama:13b-instruct & codellama:7b-instruct)
    • Used for programming-focused tasks.
    • Successfully produced code snippets and explanations tailored for development.
  3. Fast / Lightweight (tinyllama:1.1b)
    • Verified with quick utility prompts (e.g., short text transformations).
    • Delivered rapid responses with minimal overhead.

Result

Switching between models worked seamlessly, with no need to reconfigure endpoints.
Each category could be activated on-demand, proving the menu setup is not only practical but also production-ready for everyday tasks.

This milestone confirms:

  • Newelle + Ollama integration now supports real-world use.
  • Switching between categories allows the system to optimize speed, accuracy, or programming context depending on the task.

📌 This achievement brings the workflow closer to a fully dynamic JARVIS AI Assistant, where the right model is automatically chosen based on intent.