python:3.11-slim-bookworm

python:3.11-slim-bookworm, env file use, kubernetes, misc

.dockerignore

@@ -0,0 +1,13 @@
+.env
+.env-example
+.git/
+.github
+.gitignore
+.idea
+.mypy_cache
+.pre-commit-config.yaml
+.ruff_cache
+Dockerfile
+kubernetes
+docker-compose.yml
+junk/

.gitignore

@@ -21,7 +21,6 @@
 .coverage
 .coverage.*
 .dmypy.json
-.dockerignore
 .eggs/
 .env
 .hypothesis/

.pre-commit-config.yaml

@@ -32,7 +32,7 @@ repos:
     -   id: check-docstring-first
     -   id: check-executables-have-shebangs
     -   id: check-json
-    -   id: check-yaml
+#    -   id: check-yaml
     -   id: debug-statements
     -   id: fix-byte-order-marker
     -   id: detect-private-key

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.11-slim-buster
+FROM python:3.11-slim-bookworm
 
 RUN adduser --uid 1000 --disabled-password --gecos '' appuser
 USER 1000

README.md

@@ -42,40 +42,57 @@ This `README` was written by [Claude 2](https://www.anthropic.com/index/claude-2
     - `meta-llama/Llama-2-13b-chat-hf`
     - `meta-llama/Llama-2-70b-chat-hf`
 - Streaming output of assistant responses
-- Leverages LangChain for dialogue management
+- Leverages LangChain for dialogue and memory management
 - Integrates with [LangSmith](https://smith.langchain.com) for tracing conversations
 - Allows giving feedback on assistant's responses
+- Tries reading API keys and default values from environment variables
+- Parameters in sidebar can be customized
+
+# Code Overview
+- `langchain-streamlit-demo/app.py` - Main Streamlit app definition
+- `langchain-streamlit-demo/llm_stuff.py` - LangChain helper functions
+- `Dockerfile`, `docker-compose.yml`: Docker deployment
+- `kubernetes/`: Kubernetes deployment files
+- `.github/workflows/`: CI/CD workflows
+
+# Deployment
+`langchain-streamlit-demo` is deployed as a [Docker image](https://hub.docker.com/r/joshuasundance/langchain-streamlit-demo) based on the [`python:3.11-slim-bookworm`](https://github.com/docker-library/python/blob/81b6e5f0643965618d633cd6b811bf0879dee360/3.11/slim-bookworm/Dockerfile) image.
+CI/CD workflows in `.github/workflows` handle building and publishing the image as well as pushing it to Hugging Face.
 
-# Usage
 ## Run on HuggingFace Spaces
 [![Open HuggingFace Space](https://huggingface.co/datasets/huggingface/badges/raw/main/open-in-hf-spaces-sm.svg)](https://huggingface.co/spaces/joshuasundance/langchain-streamlit-demo)
 
 ## With Docker (pull from Docker Hub)
-1. Run in terminal: `docker run -p 7860:7860 joshuasundance/langchain-streamlit-demo:latest`
-2. Open http://localhost:7860 in your browser.
 
-## Docker Compose
-1. Clone the repo. Navigate to cloned repo directory.
-2. Run in terminal: `docker compose up`
-3. Then open http://localhost:7860 in your browser.
+1. _Optional_: Create a `.env` file based on `.env-example`
+2. Run in terminal:
 
-# Configuration
-- Select a model from the dropdown
-- Enter an API key for the relevant provider
-- Optionally enter a LangSmith API key to enable conversation tracing
-- Customize the assistant prompt and temperature
+`docker run -p 7860:7860 joshuasundance/langchain-streamlit-demo:latest`
 
-# Code Overview
-- `app.py` - Main Streamlit app definition
-- `llm_stuff.py` - LangChain helper functions
+or
 
-# Deployment
-The app is packaged as a Docker image for easy deployment. It is published to Docker Hub and Hugging Face Spaces:
+`docker run -p 7860:7860  --env-file .env joshuasundance/langchain-streamlit-demo:latest`
 
-- [DockerHub](https://hub.docker.com/r/joshuasundance/langchain-streamlit-demo)
-- [HuggingFace Spaces](https://huggingface.co/spaces/joshuasundance/langchain-streamlit-demo)
+3. Open http://localhost:7860 in your browser
 
-CI workflows in `.github/workflows` handle building and publishing the image.
+## Docker Compose (build locally)
+1. Clone the repo. Navigate to cloned repo directory
+2. _Optional_: Create a `.env` file based on `.env-example`
+3. Run in terminal:
+
+`docker compose up`
+
+or
+
+`docker compose up --env-file env`
+
+4. Open http://localhost:7860 in your browser
+
+## Kubernetes
+1. Clone the repo. Navigate to cloned repo directory
+2. Create a `.env` file based on `.env-example`
+3. Run in terminal: `cd kubernetes && kubectl apply -f resources.yaml`
+4. Get the IP address for your new service: `kubectl get service langchain-streamlit-demo`
 
 # Links
 - [Streamlit](https://streamlit.io)
@@ -84,6 +101,3 @@ CI workflows in `.github/workflows` handle building and publishing the image.
 - [OpenAI](https://openai.com/)
 - [Anthropic](https://www.anthropic.com/)
 - [Anyscale Endpoints](https://endpoints.anyscale.com/)
-
-# TODO
-1. More customization / parameterization in sidebar

docker-compose.yml

@@ -4,8 +4,6 @@ services:
   langchain-streamlit-demo:
     image: langchain-streamlit-demo:latest
     build: .
-    env_file:
-      - .env
     ports:
       - "${APP_PORT:-7860}:${APP_PORT:-7860}"
     command: [

kubernetes/deploy.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+set -euo pipefail
+IFS=$'\n\t'
+
+# Create a secret for environment variables
+secretExists=$(kubectl get secret langchain-streamlit-demo-secret --ignore-not-found)
+
+if [ -n "$secretExists" ]; then
+  echo "Secret 'langchain-streamlit-demo-secret' already exists. Deleting and recreating."
+  kubectl delete secret langchain-streamlit-demo-secret
+else
+  echo "Secret 'langchain-streamlit-demo-secret' does not exist. Creating."
+fi
+
+kubectl create secret generic langchain-streamlit-demo-secret --from-env-file=../.env
+
+
+# Deploy to Kubernetes
+kubectl apply -f resources.yaml

kubernetes/resources.yaml

@@ -0,0 +1,84 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: langchain-streamlit-demo-deployment
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: langchain-streamlit-demo
+  template:
+    metadata:
+      labels:
+        app: langchain-streamlit-demo
+    spec:
+      containers:
+      - name: langchain-streamlit-demo
+        image: joshuasundance/langchain-streamlit-demo:latest
+        imagePullPolicy: Always  # get latest on restart
+        resources:
+          requests:
+            cpu: "100m"
+            memory: "200Mi"
+          limits:
+            cpu: "500m"
+            memory: "500Mi"
+        env:
+          - name: OPENAI_API_KEY
+            valueFrom:
+              secretKeyRef:
+                name: langchain-streamlit-demo-secret
+                key: OPENAI_API_KEY
+          - name: ANTHROPIC_API_KEY
+            valueFrom:
+              secretKeyRef:
+                name: langchain-streamlit-demo-secret
+                key: ANTHROPIC_API_KEY
+          - name: ANYSCALE_API_KEY
+            valueFrom:
+              secretKeyRef:
+                name: langchain-streamlit-demo-secret
+                key: ANYSCALE_API_KEY
+          - name: LANGCHAIN_API_KEY
+            valueFrom:
+              secretKeyRef:
+                name: langchain-streamlit-demo-secret
+                key: LANGCHAIN_API_KEY
+          - name: LANGCHAIN_PROJECT
+            value: "langchain-streamlit-demo"
+        securityContext:
+          runAsNonRoot: true
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: langchain-streamlit-demo-service
+  # configure on Azure and uncomment below to use a vnet
+#  annotations:
+#    service.beta.kubernetes.io/azure-load-balancer-internal: "true"
+#    service.beta.kubernetes.io/azure-load-balancer-ipv4: vnet.ip.goes.here
+#    service.beta.kubernetes.io/azure-dns-label-name: "langchain-streamlit-demo"
+spec:
+  selector:
+    app: langchain-streamlit-demo
+  ports:
+    - protocol: TCP
+      port: 80
+      targetPort: 7860
+  type: LoadBalancer
+---
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: langchain-streamlit-demo-network-policy
+spec:
+  podSelector:
+    matchLabels:
+      app: langchain-streamlit-demo
+  policyTypes:
+  - Ingress
+  ingress:
+  - from: []  # An empty array here means it will allow traffic from all sources.
+    ports:
+    - protocol: TCP
+      port: 7860