feat(finetune): adapt code for gemma-4 LoRA

feat(finetune): adapt code for gemma-4 full fine-tuning

feat: add deployment files regarding LiteRT-LM

fix: improve the chat template to handle tool responses

fix: normalize tool call output to list and allow single-object JSON

refactor: make chat template function format in a more common manner

doc: add contributing file

Merge pull request #2 from autoharness/feature/fine-tuning

feat: add fine-tuning code and document

Merge pull request #1 from autoharness/initial-commit

Initial commit for generation pipeline

Initial commit