generating function signatures and descriptions for internal target code functions.
adding this information to the migration prompt based on the internal file dependencies in the source repository
Design decisions
Use an LLM to get function signatures instead of static code analysis as they should work with any language.
Use prompts to guarantee JSON output format instead of OpenAI's function calling feature. This makes it LLM-vendor independent in case other models are supported in the future.
This PR addresses #5 and #26 by
Design decisions