settings.yaml

# /home/wangxinxin/my_graphrag/settings.yaml

### This config file contains required core defaults that must be set, along with a handful of common optional settings.
### For a full list of available settings, see https://microsoft.github.io/graphrag/config/yaml/

### LLM settings ###
## There are a number of settings to tune the threading and token limits for LLM calls - check the docs.

models:
  default_chat_model:
    type: chat
    model_provider: openai
    auth_type: api_key
    api_key: "sk-dummy"
    model: sensenova
    # Points to Bridge Server (LLM)
    api_base: http://localhost:8900/v1
    # Critical: Set to false to allow GraphRAG to accept plain text returned by the Bridge Server
    model_supports_json: false
    # Recommended to keep concurrency low for local models
    concurrent_requests: 4
    async_mode: threaded
    retry_strategy: exponential_backoff
    max_retries: 3
    # Must be null
    tokens_per_minute: null
    requests_per_minute: null

  default_embedding_model:
    type: embedding
    model_provider: openai
    auth_type: api_key
    api_key: "sk-dummy"
    model: bge-m3
    # 2. Points to fix_tei_proxy
    api_base: http://localhost:8102/v1
    concurrent_requests: 20
    async_mode: threaded
    retry_strategy: exponential_backoff
    max_retries: 3
    # Must be null
    tokens_per_minute: null
    requests_per_minute: null

### Input settings ###

input:
  storage:
    type: file # or blob
    base_dir: "input"
  file_type: text # [csv, text, json]

chunks:
  # BGE-M3 supports 8k; 1200 is a good balance point
  size: 1200
  overlap: 100
  group_by_columns: [id]

### Output/storage settings ###
## If blob storage is specified in the following four sections,
## connection_string and container_name must be provided

output:
  type: file # [file, blob, cosmosdb]
  base_dir: "output"
    
cache:
  type: file # [file, blob, cosmosdb]
  base_dir: "cache"

reporting:
  type: file # [file, blob]
  base_dir: "logs"

vector_store:
  default_vector_store:
    type: lancedb
    db_uri: output/lancedb
    container_name: default

### Workflow settings ###

embed_text:
  model_id: default_embedding_model
  vector_store_id: default_vector_store

extract_graph:
  model_id: default_chat_model
  prompt: "prompts/extract_graph.txt"
  entity_types: [organization,person,geo,event]
  max_gleanings: 1

summarize_descriptions:
  model_id: default_chat_model
  prompt: "prompts/summarize_descriptions.txt"
  max_length: 500

extract_graph_nlp:
  text_analyzer:
    extractor_type: regex_english # [regex_english, syntactic_parser, cfg]
  async_mode: threaded # or asyncio

cluster_graph:
  max_cluster_size: 10

extract_claims:
  enabled: false
  model_id: default_chat_model
  prompt: "prompts/extract_claims.txt"
  description: "Any claims or facts that could be relevant to information discovery."
  max_gleanings: 1

community_reports:
  model_id: default_chat_model
  graph_prompt: "prompts/community_report_graph.txt"
  text_prompt: "prompts/community_report_text.txt"
  max_length: 2000
  # Leverage the 64K window to significantly increase input length and generate more comprehensive reports
  max_input_length: 32000

embed_graph:
  enabled: false # if true, will generate node2vec embeddings for nodes

umap:
  enabled: false # if true, will generate UMAP embeddings for nodes (embed_graph must also be enabled)

snapshots:
  graphml: false
  embeddings: false

### Query settings ###
## The prompt locations are required here, but each search method has a number of optional knobs that can be tuned.
## See the config docs: https://microsoft.github.io/graphrag/config/yaml/#query

local_search:
  chat_model_id: default_chat_model
  embedding_model_id: default_embedding_model
  prompt: "prompts/local_search_system_prompt.txt"
  
  # =========== Optimization for 64K models ===========
  # Allow using 48k context (approx. 75% of the window)
  max_context_tokens: 48000
  # Reserve 2000 tokens for LLM answer generation
  llm_max_tokens: 2000
  # ===================================================
  
  # 1. Increase recall quantity (default is 10, changed to 50)
  top_k_mapped_entities: 50
  
  # 2. Lower similarity threshold (default might be 0.2 or higher; BGE-M3 may require lower)
  # Force set to 0, meaning "pull in anything even slightly relevant" and let the LLM decide
  min_score: 0.0
  
  # 3. Increase context window fill ratio (to include more text in the prompt)
  local_search_text_unit_prop: 0.8
  

global_search:
  chat_model_id: default_chat_model
  map_prompt: "prompts/global_search_map_system_prompt.txt"
  reduce_prompt: "prompts/global_search_reduce_system_prompt.txt"
  knowledge_prompt: "prompts/global_search_knowledge_system_prompt.txt"
  
  # =========== Optimization for 64K models ===========
  max_context_tokens: 48000
  llm_max_tokens: 2000
  # Allow processing more data
  data_max_tokens: 48000
  map_max_length: 1000
  reduce_max_length: 2000
  dynamic_search_threshold: 1
  dynamic_search_keep_parent: False
  dynamic_search_num_repeats: 1
  dynamic_search_use_summary: False
  dynamic_search_max_level: 2
  # ===================================================

drift_search:
  chat_model_id: default_chat_model
  embedding_model_id: default_embedding_model
  prompt: "prompts/drift_search_system_prompt.txt"
  reduce_prompt: "prompts/drift_search_reduce_prompt.txt"
  
  # =========== Optimization for 64K models ===========
  data_max_tokens: 48000
  # ===================================================
  
  reduce_max_tokens: 2000
  reduce_temperature: 0
  reduce_max_completion_tokens: null
  concurrency: 32
  drift_k_followups: 20
  primer_folds: 5
  primer_llm_max_tokens: 12000
  n_depth: 3
  local_search_text_unit_prop: 0.9
  local_search_community_prop: 0.1
  local_search_top_k_mapped_entities: 10
  local_search_top_k_relationships: 10
  local_search_max_data_tokens: 12000
  local_search_temperature: 0
  local_search_top_p: 1
  local_search_n: 1
  local_search_llm_max_gen_tokens: null
  local_search_llm_max_gen_completion_tokens: null

basic_search:
  chat_model_id: default_chat_model
  embedding_model_id: default_embedding_model
  prompt: "prompts/basic_search_system_prompt.txt"
  k: 10
  # =========== Optimization for 64K models ===========
  max_context_tokens: 48000