Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

refactor: entrypoint.sh (#20314)

Jin Haic2f17b772e96
diff --git a/.github/workflows/sep-tests.yml b/.github/workflows/sep-tests.yml
index a9979b4705..acb383c546 100644
--- a/.github/workflows/sep-tests.yml
+++ b/.github/workflows/sep-tests.yml
@@ -687,10 +687,6 @@ jobs:
-e 's#"8222:8222"#"${NATS_MONITORING_PORT}:8222"#' \
docker-compose-base.yml
- # Patch entrypoint.sh for coverage
- sed -i '/"\$PY" api\/ragflow_server.py \${INIT_SUPERUSER_ARGS} &/c\ echo "Ensuring coverage is installed..."\n "$PY" -m pip install coverage -i https://mirrors.aliyun.com/pypi/simple\n export COVERAGE_FILE=/ragflow/logs/.coverage\n echo "Starting ragflow_server with coverage..."\n "$PY" -m coverage run --source=./api/apps --omit="*/tests/*,*/migrations/*" -a api/ragflow_server.py ${INIT_SUPERUSER_ARGS} &' ./entrypoint.sh
-
-
- name: Start ragflow:nightly for Infinity
run: |
set -euo pipefail
@@ -1254,10 +1250,6 @@ jobs:
-e 's#"8222:8222"#"${NATS_MONITORING_PORT}:8222"#' \
docker-compose-base.yml
- # Patch entrypoint.sh for coverage
- sed -i '/"\$PY" api\/ragflow_server.py \${INIT_SUPERUSER_ARGS} &/c\ echo "Ensuring coverage is installed..."\n "$PY" -m pip install coverage -i https://mirrors.aliyun.com/pypi/simple\n export COVERAGE_FILE=/ragflow/logs/.coverage\n echo "Starting ragflow_server with coverage..."\n "$PY" -m coverage run --source=./api/apps --omit="*/tests/*,*/migrations/*" -a api/ragflow_server.py ${INIT_SUPERUSER_ARGS} &' ./entrypoint.sh
-
-
- name: Start ragflow:nightly for Elasticsearch
run: |
COMPOSE_FILE=docker/docker-compose.yml
@@ -1755,10 +1747,6 @@ jobs:
-e 's#"8222:8222"#"${NATS_MONITORING_PORT}:8222"#' \
docker-compose-base.yml
- # Patch entrypoint.sh for coverage
- sed -i '/"\$PY" api\/ragflow_server.py \${INIT_SUPERUSER_ARGS} &/c\ echo "Ensuring coverage is installed..."\n "$PY" -m pip install coverage -i https://mirrors.aliyun.com/pypi/simple\n export COVERAGE_FILE=/ragflow/logs/.coverage\n echo "Starting ragflow_server with coverage..."\n "$PY" -m coverage run --source=./api/apps --omit="*/tests/*,*/migrations/*" -a api/ragflow_server.py ${INIT_SUPERUSER_ARGS} &' ./entrypoint.sh
-
-
- name: Start ragflow CI service stack (infinity)
# COMPOSE_PROFILES (set in .env by the prepare step) drops `cpu`, so the
# heavy ragflow-cpu image is NOT started — only the dependency services
diff --git a/docker/docker-compose.yml b/docker/docker-compose.yml
index 7efcfee773..849aad845a 100644
--- a/docker/docker-compose.yml
+++ b/docker/docker-compose.yml
@@ -56,86 +56,3 @@ services:
extra_hosts:
- "host.docker.internal:host-gateway"
- ragflow-gpu:
- depends_on:
- mysql:
- condition: service_healthy
- required: false
- profiles:
- - gpu
- image: ${RAGFLOW_IMAGE}
- ulimits:
- nofile:
- soft: 65535
- hard: 65535
- nproc:
- soft: 65535
- hard: 65535
- # MCP is served in-process by the API server (POST /mcp on the 9380 port).
-
- # Example configuration to start Admin server:
- command:
- - --enable-adminserver
- ports:
- - ${SVR_WEB_HTTP_PORT}:80
- - ${SVR_WEB_HTTPS_PORT}:443
- - ${SVR_HTTP_PORT}:9380
- - ${ADMIN_SVR_HTTP_PORT}:9381
- volumes:
- - ./ragflow-logs:/ragflow/logs
- # - ./nginx/ragflow.conf:/etc/nginx/conf.d/ragflow.conf
- # - ./nginx/proxy.conf:/etc/nginx/proxy.conf
- # - ./nginx/nginx.conf:/etc/nginx/nginx.conf
- - ./service_conf.yaml.template:/ragflow/conf/service_conf.yaml.template
- - ./entrypoint-go.sh:/ragflow/entrypoint.sh
- - ${RAGFLOW_SSH_KNOWN_HOSTS:-${HOME}/.ssh/known_hosts}:${SSH_KNOWN_HOSTS:-/etc/ragflow/ssh_known_hosts}:ro
- env_file: .env
- # Route the Go services' cache/queue backend to the separate Kvrocks service.
- # The Python services use the Valkey `redis` service, which the Go deployment
- # disables (see the `redis` profile override above). KVROCKS_HOST/KVROCKS_PORT
- # are the IN-NETWORK coordinates the Go process uses to reach `kvrocks:6379`; the
- # host-published port is KVROCKS_PORT (see docker-compose-base.yml), which CI
- # offsets per job and keeps distinct from REDIS_PORT.
- environment:
- - KVROCKS_HOST=kvrocks
- - KVROCKS_PORT=6379
- networks:
- - ragflow
- restart: unless-stopped
- # https://docs.docker.com/engine/daemon/prometheus/#create-a-prometheus-configuration
- # If you use Docker Desktop, the --add-host flag is optional. This flag ensures that the host's internal IP is exposed to the Prometheus container.
- extra_hosts:
- - "host.docker.internal:host-gateway"
- deploy:
- resources:
- reservations:
- devices:
- - driver: nvidia
- count: all
- capabilities: [gpu]
-
-
- # executor:
- # depends_on:
- # mysql:
- # condition: service_healthy
- # image: ${RAGFLOW_IMAGE}
- # volumes:
- # - ./ragflow-logs:/ragflow/logs
- # - ./nginx/ragflow.conf:/etc/nginx/conf.d/ragflow.conf
- # env_file: .env
- # entrypoint: "/ragflow/entrypoint_task_executor.sh 1 3"
- # networks:
- # - ragflow
- # restart: unless-stopped
- # # https://docs.docker.com/engine/daemon/prometheus/#create-a-prometheus-configuration
- # # If you're using Docker Desktop, the --add-host flag is optional. This flag makes sure that the host's internal IP gets exposed to the Prometheus container.
- # extra_hosts:
- # - "host.docker.internal:host-gateway"
- # deploy:
- # resources:
- # reservations:
- # devices:
- # - driver: nvidia
- # count: all
- # capabilities: [gpu]
diff --git a/docker/entrypoint-go.sh b/docker/entrypoint-go.sh
index aeb147d1da..e76ab4adc2 100755
--- a/docker/entrypoint-go.sh
+++ b/docker/entrypoint-go.sh
@@ -21,8 +21,7 @@ function usage() {
echo " --consumer-no-end=<num> End range for consumers (if using range-based)."
echo " --workers=<num> Number of task executors to run (if range is not used)."
echo " --host-id=<string> Unique ID for the host (defaults to \`hostname\`)."
- echo
- echo "MCP: configure service_conf.yaml or RAGFLOW_MCP_* environment variables."
+
echo
echo "Examples:"
echo " $0 --disable-taskexecutor"
diff --git a/docker/entrypoint.sh b/docker/entrypoint.sh
deleted file mode 100755
index 569ffd965e..0000000000
--- a/docker/entrypoint.sh
+++ /dev/null
@@ -1,400 +0,0 @@
-#!/usr/bin/env bash
-
-set -e
-
-echo "Start RAGFlow cluster, version: "
-cat /ragflow/VERSION
-
-# -----------------------------------------------------------------------------
-# Usage and command-line argument parsing
-# -----------------------------------------------------------------------------
-function usage() {
- echo "Usage: $0 [OPTIONS]"
- echo
- echo " --disable-webserver Disables the web server (nginx + ragflow_server)."
- echo " --disable-taskexecutor Disables task executor workers."
- echo " --disable-datasync Disables synchronization of datasource workers."
- echo " --enable-mcpserver Enables the MCP server."
- echo " --enable-adminserver Enables the Admin server."
- echo " --init-model-provider-tables Run model provider table migrations and exit."
- echo " --init-superuser Initializes the superuser."
- echo " --consumer-no-beg=<num> Start range for consumers (if using range-based)."
- echo " --consumer-no-end=<num> End range for consumers (if using range-based)."
- echo " --workers=<num> Number of task executors to run (if range is not used)."
- echo " --host-id=<string> Unique ID for the host (defaults to \`hostname\`)."
- echo " --mcp-host=<string> Address the MCP server binds to (default: 127.0.0.1)."
- echo " --mcp-port=<num> Port the MCP server listens on (default: 9382)."
- echo " --mcp-base-url=<string> RAGFlow base URL the MCP server calls (default: http://127.0.0.1:9380)."
- echo " --mcp-script-path=<path> MCP server entry script (default: /ragflow/mcp/server/server.py)."
- echo " --mcp-mode=<self-host|host> MCP server mode (default: self-host)."
- echo " --mcp-host-api-key=<string> API key required when --mcp-mode=self-host."
- echo " --no-transport-sse-enabled Disables the MCP SSE transport."
- echo " --no-transport-streamable-http-enabled Disables the MCP streamable HTTP transport. Disabling"
- echo " both transports re-enables this one, since the server"
- echo " requires at least one."
- echo " --no-json-response Disables JSON responses from the MCP server."
- echo
- echo "Examples:"
- echo " $0 --disable-taskexecutor"
- echo " $0 --disable-webserver --consumer-no-beg=0 --consumer-no-end=5"
- echo " $0 --disable-webserver --workers=2 --host-id=myhost123"
- echo " $0 --enable-mcpserver"
- echo " $0 --enable-adminserver"
- echo " $0 --init-superuser"
- exit 1
-}
-
-ENABLE_WEBSERVER=1 # Default to enable web server
-ENABLE_TASKEXECUTOR=1 # Default to enable task executor
-ENABLE_DATASYNC=1
-ENABLE_MCP_SERVER=0
-ENABLE_ADMIN_SERVER=0 # Default close admin server
-INIT_SUPERUSER_ARGS="" # Default to not initialize superuser
-INIT_MODEL_PROVIDER_TABLES=0
-CONSUMER_NO_BEG=0
-CONSUMER_NO_END=0
-WORKERS=1
-
-MCP_HOST="127.0.0.1"
-MCP_PORT=9382
-MCP_BASE_URL="http://127.0.0.1:9380"
-MCP_SCRIPT_PATH="/ragflow/mcp/server/server.py"
-MCP_MODE="self-host"
-MCP_HOST_API_KEY=""
-MCP_TRANSPORT_SSE_FLAG="--transport-sse-enabled"
-MCP_TRANSPORT_STREAMABLE_HTTP_FLAG="--transport-streamable-http-enabled"
-MCP_JSON_RESPONSE_FLAG="--json-response"
-
-# -----------------------------------------------------------------------------
-# Host ID logic:
-# 1. By default, use the system hostname if length <= 32
-# 2. Otherwise, use the full MD5 hash of the hostname (32 hex chars)
-# -----------------------------------------------------------------------------
-CURRENT_HOSTNAME="$(hostname)"
-if [ ${#CURRENT_HOSTNAME} -le 32 ]; then
- DEFAULT_HOST_ID="$CURRENT_HOSTNAME"
-else
- DEFAULT_HOST_ID="$(echo -n "$CURRENT_HOSTNAME" | md5sum | cut -d ' ' -f 1)"
-fi
-
-HOST_ID="$DEFAULT_HOST_ID"
-
-# Parse arguments
-for arg in "$@"; do
- case $arg in
- --disable-webserver)
- ENABLE_WEBSERVER=0
- shift
- ;;
- --disable-taskexecutor)
- ENABLE_TASKEXECUTOR=0
- shift
- ;;
- --disable-datasync)
- ENABLE_DATASYNC=0
- shift
- ;;
- --enable-mcpserver)
- ENABLE_MCP_SERVER=1
- shift
- ;;
- --enable-adminserver)
- ENABLE_ADMIN_SERVER=1
- shift
- ;;
- --init-model-provider-tables)
- INIT_MODEL_PROVIDER_TABLES=1
- shift
- ;;
- --init-superuser)
- INIT_SUPERUSER_ARGS="--init-superuser"
- shift
- ;;
- --mcp-host=*)
- MCP_HOST="${arg#*=}"
- shift
- ;;
- --mcp-port=*)
- MCP_PORT="${arg#*=}"
- shift
- ;;
- --mcp-base-url=*)
- MCP_BASE_URL="${arg#*=}"
- shift
- ;;
- --mcp-mode=*)
- MCP_MODE="${arg#*=}"
- shift
- ;;
- --mcp-host-api-key=*)
- MCP_HOST_API_KEY="${arg#*=}"
- shift
- ;;
- --mcp-script-path=*)
- MCP_SCRIPT_PATH="${arg#*=}"
- shift
- ;;
- --no-transport-sse-enabled)
- MCP_TRANSPORT_SSE_FLAG="--no-transport-sse-enabled"
- shift
- ;;
- --no-transport-streamable-http-enabled)
- MCP_TRANSPORT_STREAMABLE_HTTP_FLAG="--no-transport-streamable-http-enabled"
- shift
- ;;
- --no-json-response)
- MCP_JSON_RESPONSE_FLAG="--no-json-response"
- shift
- ;;
- --consumer-no-beg=*)
- CONSUMER_NO_BEG="${arg#*=}"
- shift
- ;;
- --consumer-no-end=*)
- CONSUMER_NO_END="${arg#*=}"
- shift
- ;;
- --workers=*)
- WORKERS="${arg#*=}"
- shift
- ;;
- --host-id=*)
- HOST_ID="${arg#*=}"
- shift
- ;;
- *)
- usage
- ;;
- esac
-done
-
-# -----------------------------------------------------------------------------
-# Replace env variables in the service_conf.yaml file
-# -----------------------------------------------------------------------------
-CONF_DIR="/ragflow/conf"
-TEMPLATE_FILE="${CONF_DIR}/service_conf.yaml.template"
-CONF_FILE="${CONF_DIR}/service_conf.yaml"
-
-rm -f "${CONF_FILE}"
-DEF_ENV_VALUE_PATTERN="\$\{([^:]+):-([^}]+)\}"
-while IFS= read -r line || [[ -n "$line" ]]; do
- if [[ "$line" =~ DEF_ENV_VALUE_PATTERN ]]; then
- varname="${BASH_REMATCH[1]}"
- default="${BASH_REMATCH[2]}"
-
- if [ -n "${!varname}" ]; then
- eval "echo \"$line"\" >> "${CONF_FILE}"
- else
- echo "$line" | sed -E "s/\\\$\{[^:]+:-([^}]+)\}/\1/g" >> "${CONF_FILE}"
- fi
- else
- eval "echo \"$line\"" >> "${CONF_FILE}"
- fi
-done < "${TEMPLATE_FILE}"
-
-export LD_LIBRARY_PATH="/usr/lib/x86_64-linux-gnu/"
-PY=python3
-
-# -----------------------------------------------------------------------------
-# Select Nginx Configuration based on API_PROXY_SCHEME
-# -----------------------------------------------------------------------------
-NGINX_CONF_DIR="/etc/nginx/conf.d"
-if [ -n "$API_PROXY_SCHEME" ]; then
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]]; then
- cp -f "$NGINX_CONF_DIR/ragflow.conf.hybrid" "$NGINX_CONF_DIR/ragflow.conf"
- echo "Applied nginx config: ragflow.conf.hybrid"
- elif [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- cp -f "$NGINX_CONF_DIR/ragflow.conf.golang" "$NGINX_CONF_DIR/ragflow.conf"
- echo "Applied nginx config: ragflow.conf.golang (default)"
- else
- cp -f "$NGINX_CONF_DIR/ragflow.conf.python" "$NGINX_CONF_DIR/ragflow.conf"
- echo "Applied nginx config: ragflow.conf.python"
- fi
-else
- # Default to python backend
- cp -f "$NGINX_CONF_DIR/ragflow.conf.python" "$NGINX_CONF_DIR/ragflow.conf"
- echo "Default: applied nginx config: ragflow.conf.python"
-fi
-
-# -----------------------------------------------------------------------------
-# Function(s)
-# -----------------------------------------------------------------------------
-
-function task_exe() {
- local consumer_id="$1"
- local host_id="$2"
-
- JEMALLOC_PATH="$(pkg-config --variable=libdir jemalloc)/libjemalloc.so"
- while true; do
- LD_PRELOAD="$JEMALLOC_PATH" \
- "$PY" rag/svr/task_executor.py -i "${host_id}_${consumer_id}" -t "common" &
- wait;
- sleep 1;
- done
-}
-
-function start_mcp_server() {
- echo "Starting MCP Server on ${MCP_HOST}:${MCP_PORT} with base URL ${MCP_BASE_URL}..."
- "$PY" "${MCP_SCRIPT_PATH}" \
- --host="${MCP_HOST}" \
- --port="${MCP_PORT}" \
- --base-url="${MCP_BASE_URL}" \
- --mode="${MCP_MODE}" \
- --api-key="${MCP_HOST_API_KEY}" \
- "${MCP_TRANSPORT_SSE_FLAG}" \
- "${MCP_TRANSPORT_STREAMABLE_HTTP_FLAG}" \
- "${MCP_JSON_RESPONSE_FLAG}" &
-}
-
-function ensure_docling() {
- [[ "${USE_DOCLING}" == "true" ]] || { echo "[docling] disabled by USE_DOCLING"; return 0; }
- DOCLING_PIN="${DOCLING_VERSION:-==2.71.0}"
- "$PY" -c "import importlib.util,sys; sys.exit(0 if importlib.util.find_spec('docling') else 1)" \
- || uv pip install -i https://pypi.tuna.tsinghua.edu.cn/simple --extra-index-url https://pypi.org/simple --no-cache-dir "docling${DOCLING_PIN}"
-}
-
-function ensure_db_init() {
- echo "Initializing database tables..."
- "$PY" -c "from api.db.db_models import init_database_tables as init_web_db; init_web_db()"
- echo "Database tables initialized."
-}
-
-# Run the model provider table migrations. The Go backend owns those tables when
-# it serves the API, so it runs its own --migrate action instead of the Python
-# script.
-function run_model_provider_migrations() {
- DB_TYPE_NORMALIZED="${DB_TYPE:-mysql}"
- DB_TYPE_NORMALIZED="${DB_TYPE_NORMALIZED,,}"
- if [[ "${DB_TYPE_NORMALIZED}" == "gaussdb" || "${DB_TYPE_NORMALIZED}" == "gauss" ]]; then
- # These migrations contain MySQL-only SQL and cannot run against a
- # GaussDB metadata database.
- echo "Skipping MySQL-specific model provider table migrations for DB_TYPE=${DB_TYPE:-mysql}."
- return 0
- fi
-
- if [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Running model provider table migrations (go)..."
- bin/ragflow_server --migrate
- else
- tools/scripts/run_migrations.sh
- fi
-}
-
-# -----------------------------------------------------------------------------
-# Start components based on flags
-# -----------------------------------------------------------------------------
-
-run_with_restart() {
- local process_name="$1"
- shift
-
- while true; do
- echo "Attempt to start ${process_name}..."
- set +e
- "$@"
- local exit_code=$?
- set -e
- echo "${process_name} exited with code ${exit_code}. Restarting in 1 second..."
- sleep 1
- done
-}
-
-# Initialize the database schema before any of the services below start.
-# This used to run inside the web server block only, so hosts that ran e.g.
-# just the admin server, data sync, MCP server, or task executors never
-# initialized the schema.
-#
-# Under the go scheme the Go backend owns the schema: its --migrate action and
-# dao.InitDB create and converge the tables. Skip the Python initialization so a
-# single side owns the schema.
-if [[ "${API_PROXY_SCHEME}" != "go" ]]; then
- ensure_db_init
-fi
-
-if [[ "${INIT_MODEL_PROVIDER_TABLES}" -eq 1 ]]; then
- run_model_provider_migrations
-fi
-
-if [[ "${ENABLE_ADMIN_SERVER}" -eq 1 ]]; then
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "python" ]]; then
- echo "Attempt to start Admin python server..."
- run_with_restart "Admin python server" "$PY" admin/server/admin_server.py &
- fi
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Starting Admin go server..."
- run_with_restart "Admin go server" bin/ragflow_server --admin &
- fi
-fi
-
-if [[ "${ENABLE_WEBSERVER}" -eq 1 ]]; then
- ensure_docling
-
- echo "Starting nginx..."
- /usr/sbin/nginx -c /etc/nginx/nginx.conf
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "python" ]]; then
- echo "Attempt to start RAGFlow python server..."
- run_with_restart "RAGFlow python server" "$PY" api/ragflow_server.py ${INIT_SUPERUSER_ARGS} &
- fi
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Starting RAGFlow go server..."
- run_with_restart "RAGFlow go server" bin/ragflow_server --api &
- fi
-fi
-
-if [[ "${ENABLE_DATASYNC}" -eq 1 ]]; then
- if [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Starting data sync (go)..."
- run_with_restart "RAGFlow go server" bin/ragflow_server --syncer &
- else
- echo "Starting data sync..."
- run_with_restart "Data sync" "$PY" rag/svr/sync_data_source.py &
- fi
-fi
-
-# The Go backend serves MCP in-process from --api (POST /mcp on the main API
-# port) and exposes the same ragflow_retrieval, ragflow_list_datasets and
-# ragflow_list_chats tools. The standalone Python MCP server would only
-# duplicate it, so the go scheme relies on the built-in endpoint instead.
-if [[ "${ENABLE_MCP_SERVER}" -eq 1 ]] && [[ "${API_PROXY_SCHEME}" != "go" ]]; then
- start_mcp_server
-fi
-
-
-if [[ "${ENABLE_TASKEXECUTOR}" -eq 1 ]]; then
- if [[ "${CONSUMER_NO_END}" -gt "${CONSUMER_NO_BEG}" ]]; then
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "python" ]]; then
- echo "Starting python task executors on host '${HOST_ID}' for IDs in [${CONSUMER_NO_BEG}, ${CONSUMER_NO_END})..."
- for (( i=CONSUMER_NO_BEG; i<CONSUMER_NO_END; i++ ))
- do
- task_exe "${i}" "${HOST_ID}" &
- done
- fi
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Starting ingestor..."
- run_with_restart "ingestor" bin/ragflow_server --ingestor &
- fi
- else
- # Otherwise, start a fixed number of workers
- echo "Starting ${WORKERS} task executor(s) on host '${HOST_ID}'..."
- for (( i=0; i<WORKERS; i++ ))
- do
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "python" ]]; then
- echo "Starting python task executor..."
- task_exe "${i}" "${HOST_ID}" &
- sleep 1;
- fi
-
- if [[ "${API_PROXY_SCHEME}" == "hybrid" ]] || [[ "${API_PROXY_SCHEME}" == "go" ]]; then
- echo "Starting ingestor..."
- run_with_restart "ingestor" bin/ragflow_server --ingestor &
- fi
- done
- fi
-fi
-
-wait