Prompts and generation parameters are assembled in a Python service layer and passed to Hugging Face text-generation pipelines.
- Implemented as a service class around a text-generation pipeline inside a Django app.
Prompt engineering
Designing and combining prompts and generation parameters so that small language models produce useful text.
Project requirement
Generate project documents with small, free models on minimal hardware.
Technical question
Which prompting techniques and parameters give usable output across several small models?
Validation
Techniques were tested alone and in combination on five models; output was judged by how well it met user expectations.
Decision
One shared configuration that performed adequately on all five models was kept.
The findings are task- and model-specific: a setting that helps one model can degrade another.