A containerized lambda function that does the following:
- downloads the chromadb data from S3 in the
chromadb.zipfile - 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.zipfile
A containerized lambda function that does the following:
- downloads the chromadb data from S3 in the
chromadb.zipfile - queries the chromadb data for the 5 most similar data points closest to the supplied
query - passes the
queryand the similar data points to the OpenAI API to answer the question
A python runtime lambda function that allows for login, file manageent, and access to the RAG answering utility.
The frontend is a CloudFront distribution pointing to an S3 bucket, which talks to the backend.
I have some miscellaneous scripts in the session-notes/ directory, used for actually creating the inputs for the other apps above
This file exclusively combines LLM instructions, campaign state, and big-picture notes in one doc for use with the summary-generate.py method
- create a
mainifest.txtwhich 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.mdis an extension that does not get uploaded when runningsync-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.txtand thegenerate.shscript will walk th edirectory and find all the*.mdfiles
cd session-notes/
sh generate.sh
This will generate a file in sessions/instructions-and-state.txt
This script requires a OPENAI_API_KEY environment variable.
- set your
OPENAI_API_KEYenvironment variable - create three files
sessions/YYYY-MM-DD-chat-log.mdsessions/YYYY-MM-DD-notes.mdsessions/YYYY-MM-DD-summary.md
- run the
generate.shscript to createsessions/instructions-and-state.txt
Once you run the script with whatever flavor of python (uv, pipenv, whatever idc), it will generate a sessions/YYYY-MM-DD-summary.md
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.