Semantic Source Code Summarizer

• 

  Loading the generated yml file into big-agi

• 

  Generate sequence diagram based on description of yml

Inspiration

Needs to navigate mid to large code base. LLM's context window is too small to upload all sources codes in it, and fine tuning is too expensive/slow.

What it does

Generate a token-size sensitive summary of your code base in yml format, which can be used to describe your code base to LLM.

How we built it

python script. https://github.com/fredliubojin/src_sem_sum

Challenges we ran into

Slow GPT4 API.

Accomplishments that we're proud of

Mid size code base transformed into small yml file that fit into context window and still retain semantic and logical information. Can be used to understand code base along with tools such as big-agi

What we learned

The idea looks promisingl

What's next for Semantic Source Code Summarizer

Keep iterating and adding more functions, making it automatically parsing any code base.

Built With

• openai
• python