commit 388e0fa8be6645b06f3bae663afdb8955720e3ea
Author: Erik Stambaugh <erik@erikstambaugh.com>
Date:   Mon Feb 19 19:48:10 2024 -0800

    Initial version -- can create a jupyter lab container

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..0251df0
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,2 @@
+notebooks
+.*.sw*
diff --git a/Makefile b/Makefile
new file mode 100644
index 0000000..96b2da6
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,19 @@
+
+
+default: up
+
+up: run
+
+run: build
+	docker compose up
+
+build: docker-compose.yaml jupyterlab.Dockerfile
+	docker compose build
+
+down:
+	docker compose down
+
+
+.PHONY: default up run down build
+
+
diff --git a/ROADMAP.md b/ROADMAP.md
new file mode 100644
index 0000000..a685cd6
--- /dev/null
+++ b/ROADMAP.md
@@ -0,0 +1,89 @@
+
+## the basic path
+
+- [X] make a jupyterlab docker container!
+    - [ ] volume mount for notebooks
+    - [ ] volume for training data?
+
+- [ ] get a model working in jupyter notebook
+- [ ] make a dockerfile to set up the base image
+- [ ] script to train and validate the model, packaging up the trained model alongside a report of its accuracy etc.
+    - [ ] include the package requirements in the artifact
+
+- [ ] taking that trained model and putting it in a flask or fastAPI container
+
+- [ ] how about using my laptop-docker-compose pytest script thing? that's pretty great for what i need
+
+## next steps
+
+- [ ] get the data into redis for training -- or another data store?
+- [ ] how about a clean path from jupyter to docker?
+- [ ] CI/CD: get it working with git and webhooks
+- [ ] git workflow with webhooks
+    - [ ] what is a good project model?  here's a guess:
+        - base dockerfiles for conda and build scripts (?)
+        - model definition
+        - dataset definitions
+        - rest API
+        - test container
+- [ ] whylogs to track model drift?
+
+## questions
+
+
+- [ ] how do we get from jupyter to docker?  what's a clean path?
+- [ ] what's a good practice for making training data available?  Just have a redis container?
+
+- [ ] does it make sense to have each component be a separate CI/CD managed project?
+    - data ingest (for each dataset?)
+    - feature extraction
+    - labeling
+    - jupyter notebook for experimentation
+    - converter for jupyter-> app model  (or do we just rely on people copypasting)
+    - model trainer (+verification?)
+    - [ ] maybe we just define a production application and worry about the rest later?
+        - model + API (or messaging system)
+        - 
+
+- [ ] what's a good way to get project dependencies into a jupyter container?
+
+- [ ] does a setup like above help with parameter optimization?
+    - [ ] should an optimizer be its own containerized service?
+
+- [ ] how can we reuse feature extraction for regular model input?
+    - [ ] should a feature extractor be a service?
+
+
+## What needs doing?
+
+- [ ] step one is writing out all these steps!
+
+- [ ] docker image build pipeline:
+    - [ ] miniconda image
+    - [ ] how do we pull in the requirements for a particular model?
+        - honestly just ship a Dockerfile alongside the model
+            - but it needs to pull from a manifest that an individual developer using something like jupyter would use
+            - [ ] verify this part of the pipeline according to the training materials
+
+- [ ] feature extraction
+
+- [ ] training phase
+    
+
+- [ ] packaging up a pre-trained model artifact
+
+- [ ] CI/CD
+    [ ] webhook plugin for gitea?  https://github.com/adnanh/webhook
+
+
+- [ ] missing from the course????
+    - [ ] training data store!
+    - [ ] labeling phase (is that the right term?) -- unattended?
+    - [ ] feature extraction phase
+
+## refer:
+
+- https://github.com/docker-science/cookiecutter-docker-science?tab=readme-ov-file
+- https://docker-science.github.io/
+
+
diff --git a/docker-compose.yaml b/docker-compose.yaml
new file mode 100644
index 0000000..1a0172b
--- /dev/null
+++ b/docker-compose.yaml
@@ -0,0 +1,15 @@
+---
+version: "3.5"
+services:
+  jupyter:
+    container_name: jupyter
+    build:
+      context: .
+      dockerfile: jupyterlab.Dockerfile
+    restart: unless-stopped
+    ports:
+      - 0.0.0.0:9001:9001
+    volumes:
+      - ./notebooks:/notebooks
+
+
diff --git a/jupyterlab.Dockerfile b/jupyterlab.Dockerfile
new file mode 100644
index 0000000..ad63e9a
--- /dev/null
+++ b/jupyterlab.Dockerfile
@@ -0,0 +1,19 @@
+
+FROM continuumio/miniconda3
+
+# docker run -i -t -p 8888:8888 continuumio/miniconda3 /bin/bash \
+# -c "/opt/conda/bin/conda install jupyter -y --quiet && mkdir \
+# /opt/notebooks && /opt/conda/bin/jupyter notebook \
+# --notebook-dir=/opt/notebooks --ip='*' --port=8888 \
+# --no-browser --allow-root"
+
+RUN    /opt/conda/bin/conda install jupyterlab -y --quiet \
+    && mkdir /opt/notebooks
+
+CMD /opt/conda/bin/jupyter lab \
+        --notebook-dir=/notebooks \
+        --ip='*' \
+        --port=9001 \
+        --no-browser \
+        --allow-root
+