Fix: Enforce strict uvicorn entrypoint configuration

Dockerfile

@@ -19,5 +19,4 @@ EXPOSE 7860
 # Set PYTHONPATH to /app
 ENV PYTHONPATH=/app
 
-# Use entrypoint.sh for proper startup sequence (cleans stale errors, starts services)
-ENTRYPOINT ["/bin/bash", "/app/entrypoint.sh"]
+CMD ["./entrypoint.sh"]

entrypoint.sh

@@ -1,234 +1,2 @@
 #!/bin/bash
-set -e
-# Entrypoint for HuggingClaw - Cain
-
-# Ensure frontend directory exists (defensive: created at build time, but verify at runtime)
-mkdir -p /app/frontend
-mkdir -p /data/frontend
-
-# Ensure /data is writable and exists
-mkdir -p /data
-chmod 777 /data 2>/dev/null || true
-
-# Configure git identity for commits from within the container
-# This ensures commits (like atomic fixes from Cain) succeed
-git config --global user.email "cain@huggingface.co"
-git config --global user.name "Cain"
-
-echo "=========================================="
-echo "Cain Starting..."
-echo "=========================================="
-echo "Timestamp: $(date -u +"%Y-%m-%dT%H:%M:%SZ")"
-echo "PORT: ${PORT:-7860}"
-echo "Working Directory: $(pwd)"
-echo "Python Version: $(python --version)"
-echo "Data directory: $(ls -la /data 2>&1 | head -5)"
-echo "=========================================="
-
-# Change to app directory
-cd /app
-
-# CRITICAL: Always regenerate .env from .env.example before starting services
-# This prevents split-brain state where uvicorn/worker use different env contexts
-# when .env exists but is outdated (e.g., HF secrets changed in Space settings)
-# FIX: Skip empty values to avoid overriding HF Space secrets with empty strings
-if [ -f "/app/.env.example" ]; then
-    echo "Regenerating .env from .env.example (skipping empty values to preserve HF secrets)..."
-    # Only include lines that have non-empty values (VAR=value, not VAR=)
-    # This prevents empty values from overriding HF-injected secrets
-    grep -E '^[A-Z_]+=.+[^[:space:]]' /app/.env.example > /app/.env 2>/dev/null || touch /app/.env
-else
-    echo "WARNING: .env.example not found, creating empty .env..."
-    touch /app/.env
-fi
-
-# Load .env file (for local development - ensures worker and uvicorn share env)
-echo "Loading .env file..."
-set -a  # Automatically export all variables
-source /app/.env
-set +a
-
-# Set data directory to dataset mount point
-export OPENCLAW_DATA_DIR=/data
-
-# CRITICAL: Explicitly export key environment variables for worker subprocess
-# This ensures the brain_minimal.py subprocess has access to all required variables
-export WORKER_MODE="${WORKER_MODE:-auto}"
-export AGENT_MODE="${AGENT_MODE:-active}"
-export SLEEP_INTERVAL="${SLEEP_INTERVAL:-5}"
-export WORKER_START_TIMEOUT="${WORKER_START_TIMEOUT:-30}"
-export OPENCLAW_DATA_DIR="/data"
-
-# CRITICAL: Export API keys for error_handlers.py used by worker
-# These are required by Claude/Anthropic SDK in the worker process
-# Preserves existing values (from HF secrets) or sets empty if not set
-export ANTHROPIC_API_KEY="${ANTHROPIC_API_KEY:-}"
-export CLAUDE_API_KEY="${CLAUDE_API_KEY:-}"
-
-# Log environment for debugging (exclude sensitive values)
-echo "Environment check:"
-echo "  WORKER_MODE: ${WORKER_MODE:-not set}"
-echo "  OPENCLAW_DATASET_REPO: ${OPENCLAW_DATASET_REPO:-not set}"
-echo "  OPENCLAW_LOG_LEVEL: ${OPENCLAW_LOG_LEVEL:-not set}"
-echo "  TZ: ${TZ:-not set}"
-echo ""
-
-# Verify app.py exists
-if [ ! -f "/app/app.py" ]; then
-    echo "ERROR: app.py not found in /app"
-    exit 1
-fi
-echo "✓ app.py found"
-
-# List key files for debugging
-echo ""
-echo "Files in /app:"
-ls -la /app/*.py 2>/dev/null || echo "  (no .py files in /app root)"
-
-# Create logs directory
-mkdir -p /app/logs
-
-# Ensure cain_status.json directory exists
-mkdir -p /data
-
-# CRITICAL: Clear stale error field from cain_status.json before startup
-# This prevents "unknown" errors from persisting across container restarts
-# Clean ALL possible status file locations to ensure consistency
-# NOTE: HF Spaces may cache error states - this cleaning runs on every container start
-echo "Cleaning stale error field from status files..."
-python3 -c "
-import json
-from pathlib import Path
-
-status_files = [
-    Path('/data/cain_status.json'),  # Primary (OPENCLAW_DATA_DIR)
-    Path('/app/openclaw/.openclaw/agents/cain_status.json'),  # Nested structure
-    Path('/app/.openclaw/agents/cain_status.json'),  # Legacy flat structure
-    Path('/app/cain_status.json'),  # App root
-    Path('/app/data/cain_status.json'),  # App data subdirectory
-    Path('/app/memory/cain_status.json'),  # Memory directory (CRITICAL: was missing!)
-    Path('/data/memory/cain_status.json'),  # Data memory directory
-]
-
-cleaned = 0
-for status_file in status_files:
-    try:
-        if status_file.exists():
-            with open(status_file, 'r') as f:
-                data = json.load(f)
-            error = data.get('error')
-            # Fix: if error is string 'unknown' or contains 'unknown', set to null
-            if isinstance(error, str) and error.strip().lower() in ('unknown', 'none', 'null', ''):
-                data['error'] = None
-                data['_cleaned_at'] = 'entrypoint'
-                with open(status_file, 'w') as f:
-                    json.dump(data, f, indent=2)
-                cleaned += 1
-                print(f'  Cleaned: {status_file}')
-            elif error is None:
-                print(f'  OK: {status_file} (error already null)')
-    except Exception as e:
-        print(f'  Could not clean {status_file}: {e}')
-
-print(f'Cleaned {cleaned} status file(s)')
-" 2>/dev/null || echo "(Could not clean status files - continuing)"
-echo ""
-
-echo ""
-echo "=========================================="
-echo "Starting A2A proxy..."
-echo "=========================================="
-
-# Start a2a-proxy on port 7860 in background (EXTERNAL port)
-# Routes A2A traffic to FastAPI backend on internal port
-if [ -f "/app/scripts/a2a-proxy.cjs" ]; then
-    echo "✓ Starting a2a-proxy.cjs on port 7860 (external)..."
-    OPENCLAW_PORT=7862 LISTEN_PORT=7860 node /app/scripts/a2a-proxy.cjs > /app/logs/a2a-proxy.log 2>&1 &
-    A2A_PID=$!
-    echo "  A2A proxy PID: $A2A_PID"
-    # Record PID for cleanup on shutdown
-    echo $A2A_PID > /tmp/a2a-proxy.pid
-else
-    echo "  WARNING: a2a-proxy.cjs not found - A2A routing disabled"
-fi
-
-echo ""
-echo "=========================================="
-echo "Starting uvicorn..."
-echo "=========================================="
-
-# Start uvicorn on INTERNAL port 7862 (a2a-proxy faces external on 7860)
-echo "✓ Starting uvicorn on internal port 7862..."
-uvicorn app:app --host 0.0.0.0 --port 7862 > /app/logs/uvicorn.log 2>&1 &
-UVICORN_PID=$!
-echo "  Uvicorn PID: $UVICORN_PID"
-echo $UVICORN_PID > /tmp/uvicorn.pid
-
-# Verify processes are running AND uvicorn is accepting connections
-echo ""
-echo "=========================================="
-echo "Verifying processes..."
-echo "=========================================="
-sleep 3
-
-# Check if processes exist
-ps aux | grep -E "(uvicorn|node.*a2a-proxy)" | grep -v grep || echo "  WARNING: Some processes may not have started"
-
-# CRITICAL: Verify uvicorn is actually accepting connections on internal port 7862
-echo ""
-echo "Verifying uvicorn is accepting connections on internal port 7862..."
-for i in {1..10}; do
-    if curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:7862/health | grep -q "200\|401"; then
-        echo "  ✓ Uvicorn is accepting connections on port 7862"
-        break
-    else
-        if [ $i -eq 10 ]; then
-            echo "  ✗ ERROR: Uvicorn not responding after 10 attempts"
-            echo "  Checking uvicorn log:"
-            tail -20 /app/logs/uvicorn.log 2>&1 || echo "  (no uvicorn log found)"
-        else
-            echo "  Attempt $i/10: uvicorn not ready yet, waiting..."
-            sleep 2
-        fi
-    fi
-done
-
-echo ""
-echo "=========================================="
-echo "Cain is running!"
-echo "=========================================="
-echo "A2A Proxy (external): http://0.0.0.0:7860"
-echo "FastAPI (internal): http://0.0.0.0:7862"
-echo "Worker: Integrated (FastAPI background thread)"
-echo ""
-echo "PIDs:"
-echo "  - Uvicorn: $UVICORN_PID"
-echo "  - A2A Proxy: $A2A_PID"
-echo ""
-
-# Keep container running - monitor background processes and restart if needed
-# This prevents the container from exiting when uvicorn crashes
-while true; do
-    # Check if uvicorn is still running
-    if ! kill -0 $UVICORN_PID 2>/dev/null; then
-        echo "WARNING: Uvicorn died (PID $UVICORN_PID), restarting..."
-        uvicorn app:app --host 0.0.0.0 --port 7862 > /app/logs/uvicorn.log 2>&1 &
-        UVICORN_PID=$!
-        echo "  Uvicorn restarted with PID: $UVICORN_PID"
-        echo $UVICORN_PID > /tmp/uvicorn.pid
-    fi
-
-    # Check if a2a-proxy is still running (if it was started)
-    if [ -n "$A2A_PID" ] && ! kill -0 $A2A_PID 2>/dev/null; then
-        echo "WARNING: A2A proxy died (PID $A2A_PID), restarting..."
-        if [ -f "/app/scripts/a2a-proxy.cjs" ]; then
-            node /app/scripts/a2a-proxy.cjs > /app/logs/a2a-proxy.log 2>&1 &
-            A2A_PID=$!
-            echo "  A2A proxy restarted with PID: $A2A_PID"
-            echo $A2A_PID > /tmp/a2a-proxy.pid
-        fi
-    fi
-
-    # Sleep before next check
-    sleep 5
-done
+exec uvicorn app:app --host 0.0.0.0 --port 7860