This setup is quite similar to the one fabric [1] patterns, the limitation i encountered while testing those with local llms was the prompt efficacy. More speci

This setup is quite similar to the one fabric [1] patterns, the limitation i encountered while testing those with local llms was the prompt efficacy. More specifically the output format of the prompt is rarely respected properly. In addition to this the tone/sentiment of the answers vary a lot between models, as usual. Are there more compliant models wrt respecting prompt instructions? (assuming comparable parameter sizing) [1] https://github.com/danielmiessler/fabric https://github.com/danielmiessler/fabric