Skip to content

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

7.8 KB
Raw Blame
general:
  heartbeat_interval: 3s
api:
  host: 0.0.0.0
  http_port: 9380
  # Proxies whose X-Forwarded-For / X-Real-IP headers the Go API server trusts
  # when resolving the client address (agent webhook ip_whitelist, login audit).
  # Defaults to loopback, i.e. the nginx bundled in this image. An explicit list
  # replaces that default, so keep the loopback entries when adding the address
  # or CIDR of a further proxy in front of nginx; an empty list trusts none.
  # trusted_proxies: ['127.0.0.0/8', '::1/128']

# Optional standalone MCP listener for the Go API server. Uncomment to enable.
# RAGFLOW_MCP_* environment variables override these settings.
# mcp:
#   enabled: true
#   host: 127.0.0.1
#   port: 9382
#   launch_mode: host # Each client supplies its own RAGFlow API key.
#   # self-host gives every client the same identity; use only on trusted networks.
#   # Set host_api_key below or, preferably, RAGFLOW_MCP_HOST_API_KEY.
#   host_api_key: ''
#   transport_sse_enabled: true # /sse and /messages/
#   transport_streamable_enabled: true # /mcp
#   json_response: true # false streams Streamable HTTP responses.

admin:
  host: 0.0.0.0
  http_port: 9381
mysql:
  name: 'rag_flow'
  user: 'root'
  password: 'infini_rag_flow'
  host: 'localhost'
  port: 3306
  max_connections: 900
  stale_timeout: 300
  max_allowed_packet: 1073741824
minio:
  user: 'rag_flow'
  password: 'infini_rag_flow'
  host: 'localhost:9000'
  bucket: ''
  prefix_path: ''
es:
  hosts: 'http://localhost:1200'
  username: 'elastic'
  password: 'infini_rag_flow'
os:
  hosts: 'http://localhost:1201'
  username: 'admin'
  password: 'infini_rag_flow_OS_01'
  # Optional hybrid (BM25 + KNN) search tuning. The connector self-provisions the
  # normalization search pipeline on start-up (requires OpenSearch >= 2.10).
  # hybrid_search_pipeline: 'ragflow_hybrid_pipeline'
  # hybrid_search_weights: [0.5, 0.5]   # [text/BM25 leg, vector/KNN leg]
infinity:
  uri: 'localhost:23817'
  postgres_port: 5432
  db_name: 'default_db'
oceanbase:
  scheme: 'oceanbase' # set 'mysql' to create connection using mysql config
  config:
    db_name: 'test'
    user: 'root@ragflow'
    password: 'infini_rag_flow'
    host: 'localhost'
    port: 2881
vastbase:
  host: 'localhost'
  port: 5432
  user: 'ragflow'
  password: 'Infini_Rag@123'
  db_name: 'ragflow'
  # Metadata database, read only when DB_TYPE=vastbase; host/port/user/
  # password/ssl_mode above are shared with the doc engine.
  name: 'rag_flow'
  max_connections: 50
  stale_timeout: 300
  dbcompatibility: 'B'
  ssl_mode: '' # empty: engine picks by host (disable on loopback, verify-full otherwise)
kvrocks:
  db: 1
  username: ''
  password: 'infini_rag_flow'
  host: 'localhost:6379'
nats:
  host: "localhost"
  port: 4222
clickhouse:
  host: 'localhost'
  port: 9900
  user: ragflow
  password: infini_rag_flow
  database: ragflow
otel:
  host: "localhost"
  port: 4318
  secure: false
  sample_ratio: 1.0
  stdout: false
  enable: false
ingestor:
  mq_type: 'nats'
  # NATS consumer worker count (K). How many ingestion tasks run in parallel.
  # Override with the env var RAGFLOW_INGESTOR_CONCURRENT_WORKERS, or the
  # CLI flag --ingestor-concurrent-workers (highest precedence).
  # Valid range [1, 256]. Default 1.
  concurrent_workers: 1
  # Process-wide page concurrency (N). Total PDF pages parsed concurrently across
  # the whole process (shared by all ingestor workers). Override with the env var
  # RAGFLOW_INGESTOR_PAGE_CONCURRENCY, or the CLI flag --ingestor-page-concurrency
  # (highest precedence). Valid range [1, 16]. Default 2.
  page_concurrency: 2
  compiler_pool_size: 0
  # DeepDoc in-process (Go) ONNX inference settings.
  # Max concurrent in-process DeepDoc ONNX inference Runs in this process (K).
  # Each Run opens with max(1, inference_cpu_cores / K) intra-op threads, so the
  # total CPU cores inference may occupy is bounded by inference_cpu_cores.
  # Override with env RAGFLOW_DEEPDOC_INFERENCE_CONCURRENCY or the CLI flag
  # --deepdoc-inference-concurrency (highest precedence). Default 1. K must be
  # a positive integer and must not exceed the machine's CPU core count.
  inference_concurrency: 1
  # CPU-core budget (N) for DeepDoc in-process inference. A value of 0 means
  # "use all available cores" (runtime.NumCPU()). The per-Run intra-op thread
  # count is max(1, N/K); N may not exceed the machine's CPU core count.
  # Override with env RAGFLOW_DEEPDOC_INFERENCE_CPU_CORES or the CLI flag
  # --deepdoc-inference-cpu-cores (highest precedence). When this key is left
  # unset (recommended), N is derived from K so each Run stays single-threaded
  # by default (the prior behaviour); an explicit 0 opts into "all cores".
  # inference_cpu_cores: 0
  log:
    type: file              # file | database
    format: json            # text | json (reserved; text only today)
    path: ingest_log.log    # empty/missing = binary default (server_main.log / admin_server.log / ingestion_server.log)
    max_size: 1000          # MB before rotation
    max_backups: 10         # retained rotated files
    max_age: 30             # days
file_syncer:
  max_concurrent_syncs: 5
user_default_llm:
  default_models:
    embedding_model:
      name: 'bge-m3'
      factory: 'xxxx'
      api_key: 'xxx'
      base_url: 'http://localhost:6380'
gcs:
  bucket: '${GCS_BUCKET:-ragflow}'
  prefix_path: '${GCS_PREFIX_PATH:-}'
  endpoint_url: '${GCS_ENDPOINT:-}'
audit:
  type: file              # file | database
  format: json            # text | json (reserved; text only today)
  path: audit_log.log     # empty/missing = binary default (server_main.log / admin_server.log / ingestion_server.log)
  max_size: 100          # MB before rotation
  max_backups: 10         # retained rotated files
  max_age: 30             # days
# postgres:
#   name: 'rag_flow'
#   user: 'rag_flow'
#   password: 'infini_rag_flow'
#   host: 'postgres'
#   port: 5432
#   max_connections: 100
#   stale_timeout: 30
# s3:
#   access_key: 'access_key'
#   secret_key: 'secret_key'
#   region: 'region'
#gcs:
#  bucket: 'bridgtl-edm-d-bucket-ragflow'
# oss:
#   access_key: 'access_key'
#   secret_key: 'secret_key'
#   endpoint_url: 'https://s3.oss-cn-hangzhou.aliyuncs.com'
#   region: 'cn-hangzhou'
#   bucket: 'bucket_name'
#   signature_version: 's3'
#   addressing_style: 'virtual'
# azure:
#   auth_type: 'sas'
#   container_url: 'container_url'
#   sas_token: 'sas_token'
# azure:
#   auth_type: 'spn'
#   account_url: 'account_url'
#   client_id: 'client_id'
#   secret: 'secret'
#   tenant_id: 'tenant_id'
#   container_name: 'container_name'
# The OSS object storage uses the MySQL configuration above by default. If you need to switch to another object storage service, please uncomment and configure the following parameters.
# opendal:
#   scheme: 'mysql'  # Storage type, such as s3, oss, azure, etc.
#   config:
#     oss_table: 'opendal_storage'
# user_default_llm:
#   factory: 'BAAI'
#   api_key: 'backup'
#   base_url: 'backup_base_url'
#   default_models:
#     chat_model:
#       name: 'qwen2.5-7b-instruct'
#       factory: 'xxxx'
#       api_key: 'xxxx'
#       base_url: 'https://api.xx.com'
#     embedding_model:
#       api_key: 'xxx'
#       base_url: 'http://localhost:6380'
#     rerank_model: 'bge-reranker-v2'
#     asr_model:
#       model: 'whisper-large-v3' # alias of name
#     image2text_model: ''
# authentication:
#   client:
#     switch: false
#     http_app_key:
#     http_secret_key:
#   site:
#     switch: false
# permission:
#   switch: false
#   component: false
#   dataset: false
# smtp:
#   mail_server: ""
#   mail_port: 465
#   mail_use_ssl: true
#   mail_use_tls: false
#   mail_username: ""
#   mail_password: ""
#   mail_default_sender:
#     - "RAGFlow" # display name
#     - "" # sender email address
#   mail_frontend_url: "https://your-frontend.example.com"
# tcadp_config:
#  secret_id: 'tencent_secret_id'
#  secret_key: 'tencent_secret_key'
#  region: 'tencent_region'