feat: consolidate backend and docker-compose setup

2026-06-30 21:09:08 +07:00
commit ff3753a745
@@ -0,0 +1,44 @@
+import os
+from PIL import Image
+
+def generate_thumbnails():
+    src_base = os.path.join("pfm-web-app", "public", "produk-pfm", "foto-kemasan-v2")
+    dest_base = os.path.join("pfm-web-app", "public", "produk-pfm", "thumbs")
+    
+    if not os.path.exists(src_base):
+        print(f"Source directory {src_base} does not exist.")
+        return
+
+    # Supported image extensions
+    valid_exts = (".jpg", ".jpeg", ".png", ".webp", ".bmp")
+    
+    count = 0
+    # Walk through the source directory
+    for root, dirs, files in os.walk(src_base):
+        for file in files:
+            if file.lower().endswith(valid_exts):
+                src_path = os.path.join(root, file)
+                
+                # Get the relative path from the source base
+                rel_path = os.path.relpath(src_path, src_base)
+                dest_path = os.path.join(dest_base, rel_path)
+                
+                # Ensure destination directory exists
+                dest_dir = os.path.dirname(dest_path)
+                os.makedirs(dest_dir, exist_ok=True)
+                
+                try:
+                    with Image.open(src_path) as img:
+                        # Resize to maximum dimension of 120px while preserving aspect ratio
+                        img.thumbnail((120, 120))
+                        # Save the thumbnail
+                        img.save(dest_path)
+                        print(f"Generated thumbnail: {dest_path}")
+                        count += 1
+                except Exception as e:
+                    print(f"Failed to generate thumbnail for {src_path}: {e}")
+                    
+    print(f"Successfully generated {count} thumbnails.")
+
+if __name__ == "__main__":
+    generate_thumbnails()
@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT"
+
+unset VIRTUAL_ENV
+
+PYTHON="${PYTHON:-3.12}"
+VENV=".venv-api"
+
+echo "==> Creating PaddleOCR-VL pipeline API env (${VENV}) with Python ${PYTHON}"
+uv venv "${VENV}" --python "${PYTHON}"
+
+echo "==> Installing PaddlePaddle GPU (CUDA 12.6)"
+uv pip install --python "${VENV}" \
+  paddlepaddle-gpu \
+  -i https://www.paddlepaddle.org.cn/packages/stable/cu126/
+
+echo "==> Installing PaddleOCR doc-parser stack"
+uv pip install --python "${VENV}" "paddleocr[doc-parser]>=3.3.0"
+
+echo "==> Installing PaddleX serving dependencies"
+uv pip install --python "${VENV}" \
+  "aiohttp>=3.9" "filetype>=1.2" "fastapi>=0.110" "starlette>=0.36" "uvicorn>=0.16"
+
+echo "==> Done. Start pipeline API with: ./scripts/serve-pipeline.sh"
@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT"
+
+# Avoid picking up another project's active venv
+unset VIRTUAL_ENV
+
+PYTHON="${PYTHON:-3.12}"
+FLASH_ATTN_WHEEL="${FLASH_ATTN_WHEEL:-}"
+
+echo "==> Using uv with Python ${PYTHON}"
+uv python pin "${PYTHON}"
+uv sync
+
+if [[ -n "${FLASH_ATTN_WHEEL}" ]]; then
+  echo "==> Installing prebuilt flash-attn wheel"
+  uv pip install --python .venv "${FLASH_ATTN_WHEEL}"
+else
+  echo "==> Installing flash-attn (may require a prebuilt wheel; set FLASH_ATTN_WHEEL)"
+  uv pip install --python .venv "flash-attn==2.8.2"
+fi
+
+echo "==> Verifying paddleocr genai_server entrypoint"
+uv run paddleocr genai_server --help >/dev/null
+
+echo "==> Done. Start the service with: ./scripts/serve.sh"
@@ -0,0 +1,44 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+DEMO="${ROOT}/PaddleOCR-VL-1.6_Online_Demo"
+cd "$DEMO"
+
+unset VIRTUAL_ENV
+
+if [[ -f "${ROOT}/.env" ]]; then
+  set -a
+  # shellcheck disable=SC1091
+  source "${ROOT}/.env"
+  set +a
+fi
+
+PIPELINE_PORT="${PIPELINE_PORT:-8090}"
+GRADIO_PORT="${GRADIO_PORT:-7870}"
+export API_URL="${API_URL:-http://127.0.0.1:${PIPELINE_PORT}/layout-parsing}"
+
+echo "==> Starting PaddleOCR-VL-1.6 Online Demo"
+echo "    API_URL: ${API_URL}"
+echo "    Gradio:  http://127.0.0.1:${GRADIO_PORT}"
+
+if command -v ss >/dev/null 2>&1 && ss -tln | grep -q ":${GRADIO_PORT} "; then
+  echo "ERROR: Port ${GRADIO_PORT} is already in use." >&2
+  echo "  The demo may already be running — open http://127.0.0.1:${GRADIO_PORT}" >&2
+  echo "  To restart: stop the old process (e.g. fuser -k ${GRADIO_PORT}/tcp) then run this script again." >&2
+  echo "  Or use another port: GRADIO_PORT=7871 ./scripts/run-demo.sh" >&2
+  exit 1
+fi
+
+if [[ ! -d .venv ]]; then
+  echo "==> Creating demo venv and installing requirements"
+  uv venv .venv --python 3.12
+  uv pip install --python .venv -r requirements.txt
+fi
+
+export GRADIO_SERVER_PORT="${GRADIO_PORT}"
+export GRADIO_SERVER_NAME="127.0.0.1"
+export GRADIO_MCP_SERVER="True"
+
+exec uv run --no-project --python .venv/bin/python python app.py
+
@@ -0,0 +1,61 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT"
+
+unset VIRTUAL_ENV
+
+if [[ -f .env ]]; then
+  set -a
+  # shellcheck disable=SC1091
+  source .env
+  set +a
+fi
+
+PIPELINE_CONFIG="${PIPELINE_CONFIG:-config/pipeline_config_vllm.yaml}"
+PIPELINE_HOST="${PIPELINE_HOST:-0.0.0.0}"
+PIPELINE_PORT="${PIPELINE_PORT:-8090}"
+PIPELINE_DEVICE="${PIPELINE_DEVICE:-gpu:0}"
+VENV=".venv-api"
+
+# Check if VLLM_SERVER_URL is provided to override the url in the config
+if [[ -n "${VLLM_SERVER_URL:-}" ]]; then
+  echo "==> Overriding VLLM server URL with ${VLLM_SERVER_URL}"
+  TEMP_CONFIG=$(mktemp /tmp/pipeline_config_XXXXXX.yaml)
+  cp "${PIPELINE_CONFIG}" "${TEMP_CONFIG}"
+  sed -i "s|server_url:.*|server_url: ${VLLM_SERVER_URL}|g" "${TEMP_CONFIG}"
+  PIPELINE_CONFIG="${TEMP_CONFIG}"
+  trap 'rm -f "${TEMP_CONFIG}"' EXIT
+fi
+
+if [[ ! -d "${VENV}" ]]; then
+  echo "Missing ${VENV}. Run ./scripts/install-pipeline.sh first." >&2
+  exit 1
+fi
+
+if [[ ! -f "${PIPELINE_CONFIG}" ]]; then
+  echo "Missing pipeline config: ${PIPELINE_CONFIG}" >&2
+  exit 1
+fi
+
+echo "==> Starting PaddleOCR-VL pipeline API"
+echo "    config:  ${PIPELINE_CONFIG}"
+echo "    device:  ${PIPELINE_DEVICE}"
+echo "    url:     http://${PIPELINE_HOST}:${PIPELINE_PORT}/layout-parsing"
+echo "    vllm:    ${VLLM_SERVER_URL:-http://127.0.0.1:8118/v1}"
+
+echo "==> Starting secondary PFM classifier & OCR server"
+"${ROOT}/${VENV}/bin/python" "${ROOT}/config/classify_ocr_server.py" &
+# Patch default max_new_tokens in paddlex to avoid context length overflow in vllm
+PIPELINE_PY="${ROOT}/${VENV}/lib/python3.12/site-packages/paddlex/inference/pipelines/paddleocr_vl/pipeline.py"
+if [[ -f "${PIPELINE_PY}" ]]; then
+  echo "==> Patching default max_new_tokens in paddlex to 1024"
+  sed -i 's/vlm_kwargs\["max_new_tokens"\] = 4096/vlm_kwargs\["max_new_tokens"\] = 1024/g' "${PIPELINE_PY}"
+fi
+
+exec "${ROOT}/${VENV}/bin/paddlex" --serve \
+  --pipeline "${PIPELINE_CONFIG}" \
+  --host "${PIPELINE_HOST}" \
+  --port "${PIPELINE_PORT}" \
+  --device "${PIPELINE_DEVICE}"
@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+cd "$ROOT"
+
+unset VIRTUAL_ENV
+
+if [[ -f .env ]]; then
+  set -a
+  # shellcheck disable=SC1091
+  source .env
+  set +a
+fi
+
+GENAI_HOST="${GENAI_HOST:-0.0.0.0}"
+GENAI_PORT="${GENAI_PORT:-8118}"
+GENAI_MODEL="${GENAI_MODEL:-PaddleOCR-VL-1.6-0.9B}"
+GENAI_BACKEND="${GENAI_BACKEND:-vllm}"
+VLLM_CONFIG="${VLLM_CONFIG:-config/vllm_config.yaml}"
+
+if [[ ! -f "${VLLM_CONFIG}" ]]; then
+  echo "Missing vLLM config: ${VLLM_CONFIG}" >&2
+  exit 1
+fi
+
+echo "==> Starting PaddleOCR-VL vLLM service"
+echo "    model:   ${GENAI_MODEL}"
+echo "    backend: ${GENAI_BACKEND}"
+echo "    url:     http://${GENAI_HOST}:${GENAI_PORT}/v1"
+echo "    config:  ${VLLM_CONFIG}"
+
+exec uv run paddleocr genai_server \
+  --model_name "${GENAI_MODEL}" \
+  --host "${GENAI_HOST}" \
+  --port "${GENAI_PORT}" \
+  --backend "${GENAI_BACKEND}" \
+  --backend_config "${VLLM_CONFIG}"