Streamline Dockerfiles into one for the whole project. Modify compose file accordingly

Author: evan-caswell

.env.docker.example

@@ -1,4 +1,7 @@
-DMR_BASE_URL=
 DMR_API_KEY=
+API_BASE_URL=
 MODEL_ID=
-API_BASE_URL=
+
+# Docker Model Runner env variables will be automatically created based on the names
+# given in the compose file
+# DMR_BASE_URL

Dockerfile

@@ -0,0 +1,16 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PYTHONPATH=/app
+
+RUN apt-get update && apt-get install -y build-essential curl && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN pip install -r requirements.txt
+
+COPY backend ./backend
+COPY frontend ./frontend
+

backend/Dockerfile

@@ -3,9 +3,7 @@
 import json
 from typing import Any, AsyncGenerator
 from backend.schemas.chat import LLMRequest
-from backend.settings import get_settings
-
-settings = get_settings()
+from backend.settings import settings
 
 HEADERS = {
     "Content-Type": "application/json",

backend/external/llm_client.py

@@ -1,7 +1,5 @@
 from fastapi.routing import APIRouter
-from backend.settings import get_settings
-
-settings = get_settings()
+from backend.settings import settings
 
 router = APIRouter()
 

backend/routers/health.py

@@ -1,5 +1,4 @@
-from typing import Literal, Any
-from pydantic import BaseModel, Field
+from pydantic import BaseModel
 
 
 class Recipe(BaseModel):

backend/schemas/structured.py

@@ -1,13 +1,12 @@
 from functools import lru_cache
 from pydantic_settings import BaseSettings, SettingsConfigDict
-from pydantic import AnyHttpUrl
 
 
 class Settings(BaseSettings):
-    DMR_BASE_URL: AnyHttpUrl
+    DMR_BASE_URL: str
     DMR_API_KEY: str
     MODEL_ID: str
-    API_BASE_URL: AnyHttpUrl | None = None
+    API_BASE_URL: str | None = None
 
     model_config = SettingsConfigDict(env_file=".env", extra="ignore")
 
@@ -16,3 +15,5 @@ class Settings(BaseSettings):
 def get_settings() -> Settings:
     """Load settings from environment (cached for process lifetime)."""
     return Settings()  # pyright: ignore[reportCallIssue]
+
+settings = get_settings()

backend/settings.py

@@ -1,31 +1,33 @@
 services:
   api:
-    build: ./backend
-    container_name: smolchat-api
+    build: .
     env_file: .env.docker
     environment:
-      PYTHONPATH: /app            # parent of "backend" is on the path
-      PORT: 8000
-    working_dir: /app
+      - MODEL_ID=${MODEL_ID}
     command: uvicorn backend.main:app --host 0.0.0.0 --port 8000
-    volumes:
-      - ./backend:/app/backend    # <-- mount under /app/backend
     ports:
       - "8000:8000"
-    networks: [appnet]
+    healthcheck:
+      test: [ "CMD", "curl", "-f", "http://api:8000/healthz" ]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+    models:
+      llm:
+        endpoint_var: DMR_BASE_URL
 
   ui:
-    build: ./frontend
-    container_name: smolchat-ui
+    build: .
     env_file: .env.docker
+    environment:
+      - MODEL_ID=${MODEL_ID}
+    command: streamlit run frontend/app.py --server.address=0.0.0.0 --server.port=8501
     ports:
       - "8501:8501"
     depends_on:
       - api
-    networks: [appnet]
-    volumes:
-      - ./frontend:/app
 
-networks:
-  appnet:
-    driver: bridge
+models:
+  llm:
+    model: ${MODEL_ID}
+    context_size: 8192

compose.yaml

@@ -168,7 +168,7 @@ def _on_n_change() -> None:
     payload["stop"] = None if not s else [w.strip() for w in s.split(",") if w.strip()]
 
     if st.session_state.stream:
-        url = f"{API_BASE_URL}chat/stream"
+        url = f"{API_BASE_URL}/chat/stream"
         running_text = ""
         try:
             with httpx.stream("POST", url, json=payload, timeout=None) as r:
@@ -183,7 +183,7 @@ def _on_n_change() -> None:
         except Exception as e:
             st.error(f"Streaming error: {e}")
     else:
-        url = f"{API_BASE_URL}chat"
+        url = f"{API_BASE_URL}/chat"
         try:
             with httpx.Client(timeout=60.0) as client:
                 r = client.post(url=url, json=payload)