This MR adjusts the prompt for generating Analysis Support text to include specific instructions for the LLM (adapted from a prompt provided by @gridinoc) as well as examples of desired output. It feeds the LLM the entire proposal again, instead of just the judge comments (a partial revert of !181 (merged)), and it removes introductory lines like "Here is a summary of the proposal:".
pipenv install torqueclient jinja2
sudo apt-get install pandoc
cd scripts/llm
cp config.py.tmpl config.py
$EDITOR config.py
pipenv run ./draft-analysis.py --dry-runCloses ots/llm/meta#88 (closed)