[app]
# app settings are global available like `r2r_config.agent.app`
# project_name = "r2r_default" # optional, can also set with `R2R_PROJECT_NAME` env var
default_max_documents_per_user = 10_000
default_max_chunks_per_user = 10_000_000
default_max_collections_per_user = 5_000

# Set the default max upload size to 2 GB for local testing
default_max_upload_size = 2147483648  # 2 GB for anything not explicitly listed

  [app.max_upload_size_by_type]
    # Common text-based formats
    txt   = 2147483648  # 2 GB
    md    = 2147483648
    tsv   = 2147483648
    csv   = 2147483648
    xml   = 2147483648
    html  = 2147483648

    # Office docs
    doc   = 2147483648
    docx  = 2147483648
    ppt   = 2147483648
    pptx  = 2147483648
    xls   = 2147483648
    xlsx  = 2147483648
    odt   = 2147483648

    # PDFs
    pdf   = 2147483648

    # E-mail
    eml   = 2147483648
    msg   = 2147483648
    p7s   = 2147483648

    # Images
    bmp   = 2147483648
    heic  = 2147483648
    jpeg  = 2147483648
    jpg   = 2147483648
    png   = 2147483648
    tiff  = 2147483648

    # E-books and other formats
    epub  = 2147483648
    rtf   = 2147483648
    rst   = 2147483648
    org   = 2147483648

[agent]
system_instruction_name = "rag_agent"
# tool_names = ["local_search", "web_search"] # uncomment to enable web search
tool_names = ["local_search"]

  [agent.generation_config]
  model = "openai/gpt-4o"

[auth]
provider = "r2r"
access_token_lifetime_in_minutes = 60000 # set a very high default value, for easier testing
refresh_token_lifetime_in_days = 7
require_authentication = false
require_email_verification = false
default_admin_email = "xujiawei@cocorobo.cc"
default_admin_password = "usestudio-1"

[completion]
provider = "r2r"
concurrent_request_limit = 256
fast_llm = "openai/gpt-4o-mini"

  [completion.generation_config]
  model = "openai/gpt-4o"
  temperature = 0.1
  top_p = 1
  max_tokens_to_sample = 1_024
  stream = false
  add_generation_kwargs = { }

[crypto]
provider = "bcrypt"

[database]
provider = "postgres"
default_collection_name = "Default"
default_collection_description = "Your default collection."
# collection_summary_system_prompt = 'default_system'
# collection_summary_task_prompt = 'default_collection_summary'

# KG settings
batch_size = 256

  [database.graph_creation_settings]
    clustering_mode = "local"
    graph_entity_description_prompt = "graphrag_entity_description"
    entity_types = [] # if empty, all entities are extracted
    relation_types = [] # if empty, all relations are extracted
    fragment_merge_count = 1 # number of fragments to merge into a single extraction
    max_knowledge_relationships = 100
    max_description_input_length = 65536
    generation_config = { model = "openai/gpt-4o-mini", max_tokens_to_sample = 4_096 } # and other params, model used for relationshipt extraction
    automatic_deduplication = true # enable automatic deduplication of entities

  [database.graph_enrichment_settings]
    community_reports_prompt = "graphrag_community_reports"
    max_summary_input_length = 65536
    generation_config = { model = "openai/gpt-4o-mini", max_tokens_to_sample = 4_096 } # and other params, model used for node description and graph clustering
    leiden_params = {}

  [database.graph_search_settings]
    generation_config = { model = "openai/gpt-4o-mini" }

  [database.limits]
    # Default fallback limits if no route or user-level overrides are found
    global_per_min = 300
    monthly_limit = 10000

  [database.route_limits]
    # Set the `v3/retrieval/search` route to have a maximum of 5 requests per minute
    "/v3/retrieval/search" = { route_per_min = 120 }
    "/v3/retrieval/rag" = { route_per_min = 30 }

[embedding]
provider = "litellm"

# For basic applications, use `openai/text-embedding-3-small` with `base_dimension = 512`

# RECOMMENDED - For advanced applications,
# use `openai/text-embedding-3-large` with `base_dimension = 3072` and binary quantization
#base_model = "openai/text-embedding-3-small"
#base_dimension = 512

base_model = "openai/text-embedding-3-large"
base_dimension = 256


# rerank_model = "huggingface/mixedbread-ai/mxbai-rerank-large-v1" # reranking model

batch_size = 128
add_title_as_prefix = false
concurrent_request_limit = 256
quantization_settings = { quantization_type = "FP32" }

[completion_embedding]
# Generally this should be the same as the embedding config, but advanced users may want to run with a different provider to reduce latency
provider = "litellm"
base_model = "openai/text-embedding-3-large"
base_dimension = 256
batch_size = 128
add_title_as_prefix = false
concurrent_request_limit = 256

[file]
provider = "postgres"

[ingestion]
provider = "r2r"
chunking_strategy = "recursive"
chunk_size = 800
chunk_overlap = 400
excluded_parsers = ["mp4"]

# Ingestion-time document summary parameters
# skip_document_summary = False
# document_summary_system_prompt = 'default_system'
# document_summary_task_prompt = 'default_summary'
# chunks_for_document_summary = 128
document_summary_model = "openai/gpt-4o-mini"
vision_img_model = "openai/gpt-4o"
vision_pdf_model = "openai/gpt-4o"
automatic_extraction = true # enable automatic extraction of entities and relations

  [ingestion.chunk_enrichment_settings]
    enable_chunk_enrichment = false # disabled by default
    n_chunks = 2 # the number of chunks (both preceeding and succeeding) to use in enrichment
    generation_config = { model = "openai/gpt-4o-mini" }

  [ingestion.extra_parsers]
    pdf = "zerox"

[logging]
provider = "r2r"
log_table = "logs"
log_info_table = "log_info"

[orchestration]
provider = "simple"


[prompt]
provider = "r2r"

[email]
provider = "console_mock"