kunthawat/ALwrity
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Render deploy hardening: skip large model bootstrap on cloud, respect PORT, pin requests deps; add runtime.txt and render-build.sh

Browse Source
This commit is contained in:
ajaysi
2026-03-04 12:55:35 +05:30
parent 1d36ebe2f9
commit 2318fd8a48
2 changed files with 15 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
backend/requirements.txt
View File

@@ -38,6 +38,9 @@ google-auth-oauthlib>=1.0.0
# Web scraping and content processing # Web scraping and content processing
beautifulsoup4>=4.12.0 beautifulsoup4>=4.12.0
requests>=2.31.0 requests>=2.31.0
urllib3<2.0.0
chardet>=5.0.0
charset-normalizer<3.0.0
lxml>=4.9.0 lxml>=4.9.0
html5lib>=1.1 html5lib>=1.1
aiohttp>=3.9.0 aiohttp>=3.9.0

13
backend/start_alwrity_backend.py
View File

@@ -112,6 +112,13 @@ def bootstrap_local_llm_models():
# The code had: "Qwen/Qwen3-4B-Instruct-2507". I suspect this is a placeholder or internal model. # The code had: "Qwen/Qwen3-4B-Instruct-2507". I suspect this is a placeholder or internal model.
# I will use "Qwen/Qwen2.5-3B-Instruct" as a safe, modern, powerful laptop-friendly default. # I will use "Qwen/Qwen2.5-3B-Instruct" as a safe, modern, powerful laptop-friendly default.
# Render Free Tier has 512MB RAM. Downloading a 3B model (6GB+) will instantly crash it.
# We must skip this on Render unless we are on a paid instance with persistent disk and lots of RAM.
if os.getenv("RENDER") or os.getenv("RAILWAY_ENVIRONMENT"):
if verbose:
print(" ⚠️ Cloud environment detected (Render/Railway). Skipping local LLM bootstrap to save RAM/Time.")
return True
target_model = "Qwen/Qwen2.5-3B-Instruct" target_model = "Qwen/Qwen2.5-3B-Instruct"
if verbose: if verbose:
@@ -157,6 +164,7 @@ def start_backend(enable_reload=False, production_mode=False):
# Set host based on environment and mode # Set host based on environment and mode
# Use 127.0.0.1 for local production testing on Windows # Use 127.0.0.1 for local production testing on Windows
# Use 0.0.0.0 for actual cloud deployments (Render, Railway, etc.) # Use 0.0.0.0 for actual cloud deployments (Render, Railway, etc.)
# Render provides PORT env var, we must bind to it.
default_host = os.getenv("RENDER") or os.getenv("RAILWAY_ENVIRONMENT") or os.getenv("DEPLOY_ENV") default_host = os.getenv("RENDER") or os.getenv("RAILWAY_ENVIRONMENT") or os.getenv("DEPLOY_ENV")
if default_host: if default_host:
# Cloud deployment detected - use 0.0.0.0 # Cloud deployment detected - use 0.0.0.0
@@ -165,7 +173,9 @@ def start_backend(enable_reload=False, production_mode=False):
# Local deployment - use 127.0.0.1 for better Windows compatibility # Local deployment - use 127.0.0.1 for better Windows compatibility
os.environ.setdefault("HOST", "127.0.0.1") os.environ.setdefault("HOST", "127.0.0.1")
os.environ.setdefault("PORT", "8000") # Render sets PORT automatically. We should respect it if present, otherwise default to 8000.
# We don't setdefault("PORT", "8000") here because we want to use os.getenv("PORT") directly later
# to catch if it's missing and THEN default.
# Set reload based on argument or environment variable # Set reload based on argument or environment variable
if enable_reload and not production_mode: if enable_reload and not production_mode:
@@ -182,6 +192,7 @@ def start_backend(enable_reload=False, production_mode=False):
print(f" 📍 Host: {host}") print(f" 📍 Host: {host}")
print(f" 🔌 Port: {port}") print(f" 🔌 Port: {port}")
print(f" 🔄 Reload: {reload}") print(f" 🔄 Reload: {reload}")
print(f" 🔄 Reload: {reload}")
try: try:
# Import and run the app # Import and run the app