sarav18/git-sre-llama3-lora

Git SRE Autonomous Agent (Fine-tuned Llama-3-8B)

This model is a specialized SRE (Site Reliability Engineer) agent designed to resolve Git repository disasters autonomously.

Core Capabilities:

• Autonomous Reasoning: Uses the ReAct (Reasoning + Acting) framework to analyze terminal errors.
• Safety First: Fine-tuned to prefer non-destructive recovery (e.g., git reflog over git reset --hard).
• Structured Output: Always provides a Thought, Action, and Correction/Safety Tip.

Technical Details:

• Base Model: Llama-3-8B (Unsloth 4-bit)
• Training Method: LoRA fine-tuning on a custom Git-Trace dataset.
• Intended Use: Powering VS Code extensions and autonomous terminal agents.

This llama model was trained 2x faster with Unsloth