r2r/r2r.toml

[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 = 100
default_max_chunks_per_user = 100_000
default_max_collections_per_user = 10

[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 = 60
refresh_token_lifetime_in_days = 7
require_authentication = false
require_email_verification = false
default_admin_email = "xujiawei@cocorobo.cc"
default_admin_password = "usestudio-1"

#default_admin_email = "admin@example.com"
#default_admin_password = "change_me_immediately"

[completion]
provider = "litellm"
concurrent_request_limit = 64

  [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" } # and other params, model used for relationshipt extraction

  [database.graph_entity_deduplication_settings]
    graph_entity_deduplication_type = "by_name"
    graph_entity_deduplication_prompt = "graphrag_entity_deduplication"
    max_description_input_length = 65536
    generation_config = { model = "openai/gpt-4o-mini" } # and other params, model used for deduplication

  [database.graph_enrichment_settings]
    community_reports_prompt = "graphrag_community_reports"
    max_summary_input_length = 65536
    generation_config = { model = "openai/gpt-4o-mini" } # 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-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" }

[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"

  [ingestion.chunk_enrichment_settings]
    enable_chunk_enrichment = false # disabled by default
    strategies = ["semantic", "neighborhood"]
    forward_chunks = 3
    backward_chunks = 3
    semantic_neighbors = 10
    semantic_similarity_threshold = 0.7
    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"