Found a model.

.gitignore

@@ -0,0 +1,6 @@
+CLAUDE.md
+notes-for-blog.md
+.venv
+venv
+__pycache__
+.DS_Store

app.py

@@ -0,0 +1,32 @@
+"""
+Goals: Off-Brand, Llama Champion, Field Notes
+
+The plan
+Create an LLM powered transpiler
+Taking input code (Python, JS, etc) and translate it into Mermaid.js syntax
+Gradio then interprets that visually.
+
+The Pipeline:
+1. Input. User pastes code into a gradio.Code() block.
+2. Process. Send that code to Small Model with a specific system prompt
+3. Graph. Capture the resulting mermaid string and visualize it
+
+- include few shot examples in system prompt
+- master prompting for the system prompt and test differnet ones
+
+Future ideas include allowing user to edit the resulting flowchart and thereby edit code
+
+Model:
+https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct-GGUF
+
+To do
+- need to create the basic gradio looks
+- need to use llama.cpp
+- code and pipeline
+
+"""
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
+
+
+

requirements.txt

@@ -0,0 +1,6 @@
+# Pre-built CPU wheels for llama-cpp-python (avoids compiling from source on Spaces)
+--extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
+llama-cpp-python
+
+gradio
+huggingface_hub