privateGPT/settings.yaml

# The default configuration file.
# More information about configuration can be found in the documentation: https://docs.privategpt.dev/
# Syntax in `private_pgt/settings/settings.py`
server:
  env_name: ${APP_ENV:prod}
  root_path: ${PGPT_ROOT_PATH:}
  port: ${PORT:8080}
  cors:
    enabled: ${PGPT_CORS_ENABLED:true}
    allow_origins: ["*"]
    allow_methods: ["*"]
    allow_headers: ["*"]
  auth:
    enabled: false
    # python -c 'import base64; print("Basic " + base64.b64encode("secret:key".encode()).decode())'
    # 'secret' is the username and 'key' is the password for basic auth by default
    # If the auth is enabled, this value must be set in the "Authorization" header of the request.
    secret: "Basic c2VjcmV0OmtleQ=="
  network:
    offline_mode: ${PGPT_OFFLINE_MODE,HF_HUB_OFFLINE:false}
    proxy:
      enabled: ${PROXY_ENABLED:false}
      http_proxy: ${HTTP_PROXY,http_proxy:}
      https_server: ${HTTPS_PROXY,https_proxy:}
      username: ${PROXY_USERNAME:}
      password: ${PROXY_PASSWORD:}
      bypass: ${NO_PROXY:localhost,127.0.0.1}
    ssl:
      cert_file: ${SSL_CERT_FILE:}
      cert_dir: ${SSL_CERT_DIR:}
      verify_ssl: ${SSL_VERIFY:true}
  api_doc:
    enabled: ${PGPT_SWAGGER_ENABLED:true}
    swagger_url: ${PGPT_SWAGGER_URL:/docs}
    redoc_url: ${PGPT_REDOC_URL:/redoc}
    openapi_url: ${PGPT_OPENAPI_URL:/openapi.json}
  ui:
    enabled: ${PGPT_UI_ENABLED:true}
    path: ${PGPT_UI_PATH:/ui}
  debug_mode: ${DEBUG_MODE:false}
  max_workers: ${PGPT_MAX_WORKERS:0}

data:
  local_ingestion:
    enabled: ${LOCAL_INGESTION_ENABLED:false}
    allow_ingest_from: ["*"]
  local_data_folder: ${PGPT_LOCAL_DATA_FOLDER:local_data/private_gpt}
  reader: ${PGPT_DATA_READER:auto}
  enable_fake_progress: ${PGPT_ENABLE_FAKE_PROGRESS:false}
  enable_reuse_generated_nodes_before: ${PGPT_ENABLE_REUSE_GENERATED_NODES_BEFORE:false}
  enable_term_extractor: ${PGPT_ENABLE_TERM_EXTRACTOR:false}
  max_num_nodes: ${PGPT_MAX_NUM_NODES:0}
  use_async: ${PGPT_USE_ASYNC:false}

observability:
  mode: ${PGPT_OBSERVABILITY:none}

retrieval:
  top_k: ${PGPT_RETRIEVAL_TOP_K:32}
  maximize_top_k: ${PGPT_MAXIMIZE_TOP_K:true}
  max_merging_recalculations: ${PGPT_MAX_MERGING_RECALCULATIONS:3}

chat:
  allow_use_default_prompt: ${PGPT_ALLOW_USE_DEFAULT_PROMPT:true}
  allow_use_context: ${PGPT_ALLOW_USE_CONTEXT:true}
  allow_generate_citations: ${PGPT_ALLOW_GENERATE_CITATIONS:true}
  allow_use_reasoning: ${PGPT_ALLOW_USE_REASONING:true}
  return_missing_citations: ${PGPT_RETURN_MISSING_CITATIONS:false}
  deduplicate_context_in_history: ${PGPT_DEDUPLICATE_CONTEXT_IN_HISTORY:true}
  add_context_to_system_prompt: ${PGPT_ADD_CONTEXT_TO_SYSTEM_PROMPT:false}
  force_to_return_citations: ${PGPT_FORCE_TO_RETURN_CITATIONS:false}
  numerical_shorter_citations: ${PGPT_NUMERICAL_SHORTER_CITATIONS:false}
  condense_strategy: ${PGPT_CONDENSE_CHAT_HISTORY:condenser}
  format_context_strategy: ${PGPT_FORMAT_CONTEXT:list}
  maximum_context_length: ${PGPT_MAXIMUM_CONTEXT_LENGTH:0}
  assistant_name: ${PGPT_ASSISTANT_NAME:PrivateGPT}
  assistant_description: ${PGPT_ASSISTANT_DESCRIPTION:}
  maximum_history_length: ${PGPT_MAXIMUM_HISTORY_LENGTH:0}
  tldr_timeout: ${PGPT_TLDR_TIMEOUT:300}
  tldr_minimum_threshold_seconds: ${PGPT_TLDR_MINIMUM_THRESHOLD_SECONDS:1.5}
  multiplexing_threshold: ${PGPT_MULTIPLEXING_THRESHOLD:10}
  maximum_concurrent_requests: ${PGPT_MAXIMUM_CONCURRENT_REQUESTS:10}
  maximum_blob_size: ${PGPT_MAXIMUM_BLOB_SIZE:26214400}

stream:
  broker: ${PGPT_ASYNC_BROKER:memory}
  stream_expiration: ${PGPT_ASYNC_STREAM_EXPIRATION:3600}
  maximum_stream_length: ${PGPT_ASYNC_MAXIMUM_STREAM_LENGTH:10000}
  stream_prefix: ${PGPT_ASYNC_STREAM_PREFIX:stream}
  status_prefix: ${PGPT_ASYNC_STATUS_PREFIX:status}
  minimum_connections: ${PGPT_ASYNC_MINIMUM_CONNECTIONS:50}

llm:
  default_model: ${PGPT_LLM_DEFAULT:}
  auto_discover_models: ${PGPT_LLM_AUTO_DISCOVER_MODELS:true}

# We need to have at least one model defined here, even if it's just a placeholder for the default model defined above.
models: []

embedding:
  default_model: ${PGPT_EMBEDDING_DEFAULT:}
  auto_discover_models: ${PGPT_EMBEDDING_AUTO_DISCOVER_MODELS:true}
  ingest_mode: simple

huggingface:
  username: ${PGPT_HUGGINGFACE_USERNAME:zylon-ai}
  access_token: ${PGPT_HUGGINGFACE_TOKEN:}
  # Warning: Enabling this option will allow the model to download and execute code from the internet.
  # Nomic AI requires this option to be enabled to use the model, be aware if you are using a different model.
  trust_remote_code: true

vectorstore:
  database: ${PGPT_VECTORSTORE:qdrant}
  # Needs to match the selected embeddings model
  embed_dim: ${PGPT_EMBED_DIM:1024}
  # Determine multi-tenancy type
  multitenancy: ${PGPT_MULTITENANCY:logical}
  # Default collection name when logical multitenancy is enabled
  default_collection: ${PGPT_DEFAULT_COLLECTION:zgptvector}

node_store:
  index_store: ${PGPT_INDEX_STORE:simple}
  doc_store: ${PGPT_DOC_STORE:simple}

openai:
  api_base: ${OPENAI_API_BASE:https://api.openai.com/v1}
  api_key: ${OPENAI_API_KEY:}
  embedding_api_base: ${OPENAI_EMBEDDING_API_BASE:}
  embedding_api_key: ${OPENAI_EMBEDDING_API_KEY:}
  request_timeout: ${OPENAI_REQUEST_TIMEOUT:600.0}

triton:
  api_base: ${PGPT_TRITON_API_BASE:http://localhost:8000}
  ssl: ${PGPT_TRITON_SSL:false}
  grpc_keep_alive_time: ${PGPT_TRITON_GRPC_KEEP_ALIVE_TIME:30000}
  grpc_keep_alive_timeout: ${PGPT_TRITON_GRPC_KEEP_ALIVE_TIMEOUT:20000}
  grpc_permit_keep_alive_without_calls: ${PGPT_TRITON_GRPC_PERMIT_KEEP_ALIVE_WITHOUT_CALLS:true}
  grpc_max_pings_without_data: ${PGPT_TRITON_GRPC_MAX_PINGS_WITHOUT_DATA:0}

docling:
  mode: ${PGPT_DOCLING_MODE:api}
  api_base: ${PGPT_DOCLING_API_BASE:http://localhost:5001}
  api_version: ${PGPT_DOCLING_API_VERSION:v1alpha}
  api_key: ${PGPT_DOCLING_API_KEY:}
  tenant_id: ${PGPT_DOCLING_TENANT_ID:}
  num_threads: ${PGPT_DOCLING_NUM_WORKERS:4}
  use_ocr: ${PGPT_DOCLING_USE_OCR:true}
  ocr_model: ${PGPT_DOCLING_OCR_MODEL:easyocr}
  use_gpu: ${PGPT_DOCLING_USE_GPU:false}
  force_full_page_ocr: ${PGPT_DOCLING_FORCE_FULL_PAGE_OCR:false}
  bitmap_area_threshold: ${PGPT_DOCLING_BITMAP_AREA_THRESHOLD:0.2}
  do_cell_matching: ${PGPT_DOCLING_DO_CELL_MATCHING:true}
  image_mode: ${PGPT_DOCLING_IMAGE_MODE:placeholder}
  table_mode: ${PGPT_DOCLING_TABLE_MODE:accurate}
  image_classifier: ${PGPT_DOCLING_IMAGE_CLASSIFIER:none}
  image_descriptor: ${PGPT_DOCLING_IMAGE_DESCRIPTOR:none}
  langs: ${PGPT_DOCLING_LANGS:["en-US", "fr-FR", "de-DE", "es-ES"]}
  use_async: ${PGPT_DOCLING_USE_ASYNC:true}

celery:
  use_workers: true
  acks_late: ${PGPT_CELERY_ACKS_LATE:true}
  soft_time_limit: ${PGPT_CELERY_SOFT_TIME_LIMIT:}
  hard_time_limit: ${PGPT_CELERY_HARD_TIME_LIMIT:}
  visibility_timeout: ${PGPT_CELERY_VISIBILITY_TIMEOUT:21600}

# Broker to which workers will notify the result of tasks
tasks_results_broker:
  mode: ${PGPT_TASKS_RESULTS_BROKER_MODE:rabbitmq}

rabbitmq:
  ssl: ${PGPT_RABBITMQ_SSL:false}
  host: ${PGPT_RABBITMQ_HOST:localhost:5672}
  username: ${PGPT_RABBITMQ_USERNAME:guest}
  password: ${PGPT_RABBITMQ_PASSWORD:guest}

database:
  host: ${PGPT_DATABASE_HOST:localhost:5432}
  database: ${PGPT_DATABASE_NAME:backend-main}
  username: ${PGPT_DATABASE_USERNAME:postgres}
  password: ${PGPT_DATABASE_PASSWORD:postgres}

redis:
  host: ${PGPT_REDIS_HOST:localhost:6379}
  username: ${PGPT_REDIS_USERNAME:}
  password: ${PGPT_REDIS_PASSWORD:}
  database: ${PGPT_REDIS_DATABASE:0}

unstructured:
  url: ${PGPT_UNSTRUCTURED_URL:}

qdrant:
  url: ${PGPT_QDRANT_URL:}
  path: ${PGPT_QDRANT_PATH:local_data/qdrant}
  hybrid_search: ${PGPT_QDRANT_HYBRID_SEARCH:false}
  distance_metric: ${PGPT_QDRANT_DISTANCE_METRIC:cosine}
  hnsw_m: ${PGPT_QDRANT_HSNW_M:0}
  hnsw_payload_m: ${PGPT_QDRANT_HSNW_PAYLOAD_M:16}
  prefer_grpc: ${PGPT_QDRANT_PREFER_GRPC:true}
  timeout: ${PGPT_QDRANT_TIMEOUT:120000}

s3:
  endpoint_url: ${PGPT_S3_ENDPOINT_URL:}
  public_endpoint_url: ${PGPT_S3_PUBLIC_ENDPOINT_URL:}
  path_prefix: ${PGPT_S3_PATH_PREFIX:}
  access_key_id: ${PGPT_S3_ACCESS_KEY_ID:}
  secret_access_key: ${PGPT_S3_SECRET_ACCESS_KEY:}
  durable_bucket_name: ${PGPT_S3_DURABLE_BUCKET_NAME:zylon-storage}
  temporary_bucket_name: ${PGPT_S3_TEMPORARY_BUCKET_NAME:zylon-temporary}

phoenix:
  url: ${PGPT_PHOENIX_URL:http://localhost:6006}

opik:
  host: ${PGPT_OPIK_HOST:http://localhost:5173/}
  project_name: ${PGPT_OPIK_PROJECT_NAME:}
  api_key: ${PGPT_OPIK_API_KEY:}
  workspace: ${PGPT_OPIK_WORKSPACE:default}
  task_threads: ${PGPT_OPIK_TASK_THREADS:16}

sandbox:
  provider: ${PGPT_SANDBOX_PROVIDER:local}
  base_url: ${PGPT_SANDBOX_URL:}
  timeout: ${PGPT_SANDBOX_TIMEOUT:60}


code_execution:
  provider: ${PGPT_CODE_EXECUTION_PROVIDER:local}
  workspace_path: ${PGPT_CODE_EXECUTION_WORKSPACE_PATH:}
  timeout: ${PGPT_CODE_EXECUTION_TIMEOUT:60}
  max_output_bytes: ${PGPT_CODE_EXECUTION_MAX_OUTPUT_BYTES:1048576}

web_fetch:
  enabled: ${PGPT_WEBFETCH_ENABLED:false}
  timeout_seconds: ${PGPT_WEBFETCH_TIMEOUT_SECONDS:15}
  pool_size: ${PGPT_WEBFETCH_POOL_SIZE:8}
  pool_idle_timeout_seconds: ${PGPT_WEBFETCH_POOL_IDLE_TIMEOUT:300}
  pool_max_pages_per_browser: ${PGPT_WEBFETCH_POOL_MAX_PAGES:4}

database_query:
  timeout_seconds: ${PGPT_DATABASE_QUERY_TIMEOUT_SECONDS:1000}
  batch_size: ${PGPT_DATABASE_QUERY_BATCH_SIZE:1000}
  max_mb_result: ${PGPT_DATABASE_QUERY_MAX_MB_RESULT:150}

web_search:
  enabled: ${PGPT_WEBSEARCH_ENABLED:false}
  provider: ${PGPT_WEBSEARCH_PROVIDER:mock}
  processor: ${PGPT_WEBSEARCH_PROCESSOR:best_links}
  num_links: ${PGPT_WEBSEARCH_NUM_LINKS:20}
  mode_quality: ${PGPT_WEBSEARCH_MODE_QUALITY:fast}
  cached: ${PGPT_WEBSEARCH_CACHED:false}
  context_token: ${PGPT_WEBSEARCH_CONTEXT_TOKEN:0}

  fast_params:
    max_timeout_seconds: ${PGPT_WEBSEARCH_FAST_MAX_TIMEOUT_SECONDS:100}
    max_summary_timeout_seconds: ${PGPT_WEBSEARCH_FAST_MAX_SUMMARY_TIMEOUT_SECONDS:40}
    num_references: ${PGPT_WEBSEARCH_FAST_NUM_REFERENCES:3}
    num_concurrent_consumers: ${PGPT_WEBSEARCH_FAST_NUM_CONCURRENT_CONSUMERS:6}
    max_parallel_summary: ${PGPT_WEBSEARCH_FAST_MAX_PARALLEL_SUMMARY:2}
    index_weight: ${PGPT_WEBSEARCH_FAST_INDEX_WEIGHT:0}
    token_weight: ${PGPT_WEBSEARCH_FAST_TOKEN_WEIGHT:1}
    token_exponent_penalty: ${PGPT_WEBSEARCH_FAST_TOKEN_EXPONENT_PENALTY:4}
  accurate_params:
    max_timeout_seconds: ${PGPT_WEBSEARCH_ACCURATE_MAX_TIMEOUT_SECONDS:500}
    max_summary_timeout_seconds: ${PGPT_WEBSEARCH_ACCURATE_MAX_SUMMARY_TIMEOUT_SECONDS:100}
    num_references: ${PGPT_WEBSEARCH_ACCURATE_NUM_REFERENCES:5}
    num_concurrent_consumers: ${PGPT_WEBSEARCH_ACCURATE_NUM_CONCURRENT_CONSUMERS:4}
    max_parallel_summary: ${PGPT_WEBSEARCH_ACCURATE_MAX_PARALLEL_SUMMARY:2}
    index_weight: ${PGPT_WEBSEARCH_ACCURATE_INDEX_WEIGHT:1}
    token_weight: ${PGPT_WEBSEARCH_ACCURATE_TOKEN_WEIGHT:1}
    token_exponent_penalty: ${PGPT_WEBSEARCH_ACCURATE_TOKEN_EXPONENT_PENALTY:4}

brave:
  api_key: ${PGPT_BRAVE_API_KEY:}
  rate_limit: ${PGPT_BRAVE_RATE_LIMIT:1}

skills:
  database: ${SKILLS_DATABASE:sqlite}
  storage_provider: ${SKILLS_STORAGE_PROVIDER:local}
  skill_injection_mode: ${SKILL_INJECTION_MODE:system_prompt}
  maximum_loaded_skills: ${SKILL_MAXIMUM_LOADED_SKILLS:1}

semaphore:
  mode: ${PGPT_SEMAPHORE_MODE:memory}

transformation:
  pptx:
    vision:
      retry_number: ${PGPT_TRANSFORMATION_PPTX_VISION_RETRY_NUMBER:3}
      enable_evaluation: ${PGPT_TRANSFORMATION_PPTX_VISION_ENABLE_EVALUATION:true}
      max_concurrent: ${PGPT_TRANSFORMATION_PPTX_VISION_MAX_CONCURRENT:4}
      batch_size: ${PGPT_TRANSFORMATION_PPTX_VISION_BATCH_SIZE:8}
      max_iterations: ${PGPT_TRANSFORMATION_PPTX_VISION_MAX_ITERATIONS:2}
      enable_semaphore: ${PGPT_TRANSFORMATION_PPTX_ENABLE_SEMAPHORE:false}
      mode: ${PGPT_TRANSFORMATION_PPTX_VISION_MODE:none}

  docling:
    vision:
      retry_number: ${PGPT_TRANSFORMATION_DOCLING_VISION_RETRY_NUMBER:3}
      enable_evaluation: ${PGPT_TRANSFORMATION_DOCLING_VISION_ENABLE_EVALUATION:true}
      max_concurrent: ${PGPT_TRANSFORMATION_DOCLING_VISION_MAX_CONCURRENT:4}
      batch_size: ${PGPT_TRANSFORMATION_DOCLING_VISION_BATCH_SIZE:8}
      max_iterations: ${PGPT_TRANSFORMATION_DOCLING_VISION_MAX_ITERATIONS:2}
      enable_semaphore: ${PGPT_TRANSFORMATION_DOCLING_ENABLE_SEMAPHORE:false}
      mode: ${PGPT_TRANSFORMATION_DOCLING_VISION_MODE:none}