D&D Session Notes

Backend

dnd_rag_ingest

A containerized lambda function that does the following:

• downloads the chromadb data from S3 in the chromadb.zip file
• downloads the list of markdown files from S3 in the session-notes/ S3 prefix
• updates the chromadb data with the markdown files that were updated, deleted, or renamed
• saves the chromadb data back to S3 in the chromadb.zip file

dnd_rag_completion

A containerized lambda function that does the following:

• downloads the chromadb data from S3 in the chromadb.zip file
• queries the chromadb data for the 5 most similar data points closest to the supplied query
• passes the query and the similar data points to the OpenAI API to answer the question

dnd_rag_api

A python runtime lambda function that allows for login, file manageent, and access to the RAG answering utility.

Frontend

The frontend is a CloudFront distribution pointing to an S3 bucket, which talks to the backend.

Scripts

I have some miscellaneous scripts in the session-notes/ directory, used for actually creating the inputs for the other apps above

generate.sh

This file exclusively combines LLM instructions, campaign state, and big-picture notes in one doc for use with the summary-generate.py method

Setup

• create a mainifest.txt which lists all of the files and folders to be combined, example below.

instructions.ignore.md
characters-header.ignore.md
characters/
after-characters-spacing.ignore.md
campaign-setting-map.md
important-places.md
important-items.md
important-groups.md

• .ignore.md is an extension that does not get uploaded when running sync-notes.py, so you can create secret DM-only notes that don't get used in the RAG completion but do get used for summarizing things.
• you can set a directory in the manifest.txt and the generate.sh script will walk th edirectory and find all the *.md files

Running

cd session-notes/
sh generate.sh

This will generate a file in sessions/instructions-and-state.txt

summary-generate.py

This script requires a OPENAI_API_KEY environment variable.

Setup

• set your OPENAI_API_KEY environment variable
• create three files
  • sessions/YYYY-MM-DD-chat-log.md
  • sessions/YYYY-MM-DD-notes.md
  • sessions/YYYY-MM-DD-summary.md
• run the generate.sh script to create sessions/instructions-and-state.txt

Running

Once you run the script with whatever flavor of python (uv, pipenv, whatever idc), it will generate a sessions/YYYY-MM-DD-summary.md

sync-notes.py

This script syncs between your local session-notes folder and a S3 bucket defined on BUCKET and a prefix defined on PREFIX

Once you run the script with whatever flavor of python (uv, pipenv, whatever idc), it will interactively ask you how to resolve conflicts between files.

Hear command

To make the transcripts, i use the hear command, here's an example

./hear -p -d -i ../../dnd-session-notes/session-notes/2025-12-09.mp3 > ../../dnd-session-notes/session-notes/2025-12-09.txt