Skip to main content
GraphRAG provides sensible defaults for all configuration options. This reference documents all default values.

​
Model defaults

​
Completion models

​
DEFAULT_COMPLETION_MODEL_ID
str
default:"default_completion_model"
Default identifier for completion model configurations.
​
DEFAULT_COMPLETION_MODEL
str
default:"gpt-4.1"
Default completion model name.
​
DEFAULT_COMPLETION_MODEL_AUTH_TYPE
AuthMethod
default:"ApiKey"
Default authentication method for completion models.
​
DEFAULT_MODEL_PROVIDER
str
default:"openai"
Default model provider.

​
Embedding models

​
DEFAULT_EMBEDDING_MODEL_ID
str
default:"default_embedding_model"
Default identifier for embedding model configurations.
​
DEFAULT_EMBEDDING_MODEL
str
default:"text-embedding-3-large"
Default embedding model name.
​
DEFAULT_EMBEDDING_MODEL_AUTH_TYPE
AuthMethod
default:"ApiKey"
Default authentication method for embedding models.

​
Encoding

​
ENCODING_MODEL
str
default:"o200k_base"
Default encoding model for tokenization.

​
Directory defaults

​
DEFAULT_INPUT_BASE_DIR
str
default:"input"
Default base directory for input files.
​
DEFAULT_OUTPUT_BASE_DIR
str
default:"output"
Default base directory for output files.
​
DEFAULT_CACHE_BASE_DIR
str
default:"cache"
Default base directory for cache storage.
​
DEFAULT_UPDATE_OUTPUT_BASE_DIR
str
default:"update_output"
Default base directory for incremental update output.

​
Entity types

​
DEFAULT_ENTITY_TYPES
list[str]
Default entity types to extract during graph construction.

​
Configuration class defaults

The following sections document default values for each configuration class.

​
BasicSearchDefaults

​
prompt
None
default:"None"
Basic search prompt template.
​
k
int
default:"10"
Number of results to return.
​
max_context_tokens
int
default:"12000"
Maximum context tokens.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
embedding_model_id
str
default:"default_embedding_model"
Embedding model ID.

​
ChunkingDefaults

​
type
str
default:"tokens"
Chunking strategy type (from ChunkerType enum).
​
size
int
default:"1200"
Chunk size in tokens.
​
overlap
int
default:"100"
Overlap between chunks in tokens.
​
encoding_model
str
default:"o200k_base"
Encoding model for tokenization.
​
prepend_metadata
None
default:"None"
Metadata to prepend to chunks.

​
ClusterGraphDefaults

​
max_cluster_size
int
default:"10"
Maximum size of clusters.
​
use_lcc
bool
default:"True"
Whether to use the largest connected component.
​
seed
int
default:"0xDEADBEEF"
Random seed for clustering (3735928559 in decimal).

​
CommunityReportDefaults

​
graph_prompt
None
default:"None"
Prompt for graph-based community reports.
​
text_prompt
None
default:"None"
Prompt for text-based community reports.
​
max_length
int
default:"2000"
Maximum report length in tokens.
​
max_input_length
int
default:"8000"
Maximum input length in tokens.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
model_instance_name
str
default:"community_reporting"
Model instance name for caching.

​
DriftSearchDefaults

​
prompt
None
default:"None"
DRIFT search prompt.
​
reduce_prompt
None
default:"None"
Reduce step prompt.
​
data_max_tokens
int
default:"12000"
Maximum data tokens.
​
reduce_max_tokens
None
default:"None"
Maximum reduce tokens.
​
reduce_temperature
float
default:"0"
Temperature for reduce step.
​
reduce_max_completion_tokens
None
default:"None"
Maximum completion tokens for reduce.
​
concurrency
int
default:"32"
Concurrency level for DRIFT operations.
​
drift_k_followups
int
default:"20"
Number of followup queries.
​
primer_folds
int
default:"5"
Number of primer folds.
​
primer_llm_max_tokens
int
default:"12000"
Maximum tokens for primer LLM.
​
n_depth
int
default:"3"
Search depth.
​
local_search_text_unit_prop
float
default:"0.9"
Text unit proportion for local search component.
​
local_search_community_prop
float
default:"0.1"
Community proportion for local search component.
​
local_search_top_k_mapped_entities
int
default:"10"
Top k entities for local search.
​
local_search_top_k_relationships
int
default:"10"
Top k relationships for local search.
​
local_search_max_data_tokens
int
default:"12000"
Maximum data tokens for local search.
​
local_search_temperature
float
default:"0"
Temperature for local search.
​
local_search_top_p
float
default:"1"
Top p for local search.
​
local_search_n
int
default:"1"
Number of completions for local search.
​
local_search_llm_max_gen_tokens
int | None
default:"None"
Maximum generation tokens for local search.
​
local_search_llm_max_gen_completion_tokens
int | None
default:"None"
Maximum completion tokens for local search.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
embedding_model_id
str
default:"default_embedding_model"
Embedding model ID.

​
EmbedTextDefaults

​
embedding_model_id
str
default:"default_embedding_model"
Embedding model ID.
​
model_instance_name
str
default:"text_embedding"
Model instance name for caching.
​
batch_size
int
default:"16"
Batch size for embedding operations.
​
batch_max_tokens
int
default:"8191"
Maximum tokens per batch.
​
names
list[str]
List of embeddings to generate (uses default_embeddings).

​
ExtractClaimsDefaults

​
enabled
bool
default:"False"
Whether claim extraction is enabled.
​
prompt
None
default:"None"
Claim extraction prompt.
​
description
str
Description of claims to extract.
​
max_gleanings
int
default:"1"
Maximum number of gleaning iterations.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
model_instance_name
str
default:"extract_claims"
Model instance name for caching.

​
ExtractGraphDefaults

​
prompt
None
default:"None"
Graph extraction prompt.
​
entity_types
list[str]
Entity types to extract.
​
max_gleanings
int
default:"1"
Maximum number of gleaning iterations.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
model_instance_name
str
default:"extract_graph"
Model instance name for caching.

​
TextAnalyzerDefaults

Used for NLP-based graph extraction.
​
extractor_type
NounPhraseExtractorType
default:"RegexEnglish"
Noun phrase extractor type.
​
model_name
str
default:"en_core_web_md"
SpaCy model name.
​
max_word_length
int
default:"15"
Maximum word length to consider.
​
word_delimiter
str
default:"' '"
Delimiter between words.
​
include_named_entities
bool
default:"True"
Whether to include named entities.
​
exclude_nouns
list[str]
List of nouns to exclude (uses EN_STOP_WORDS).
​
exclude_entity_tags
list[str]
default:"[\"DATE\"]"
Entity tags to exclude.
​
exclude_pos_tags
list[str]
default:"[\"DET\", \"PRON\", \"INTJ\", \"X\"]"
Part-of-speech tags to exclude.
​
noun_phrase_tags
list[str]
default:"[\"PROPN\", \"NOUNS\"]"
Tags for noun phrases.
​
noun_phrase_grammars
dict[str, str]
Grammar rules for noun phrase combination.Default:
{
    "PROPN,PROPN": "PROPN",
    "NOUN,NOUN": "NOUNS",
    "NOUNS,NOUN": "NOUNS",
    "ADJ,ADJ": "ADJ",
    "ADJ,NOUN": "NOUNS"
}

​
ExtractGraphNLPDefaults

​
normalize_edge_weights
bool
default:"True"
Whether to normalize edge weights.
​
text_analyzer
TextAnalyzerDefaults
Text analyzer configuration.
​
concurrent_requests
int
default:"25"
Number of concurrent requests.
​
async_mode
AsyncType
default:"Threaded"
Async mode to use.

​
GlobalSearchDefaults

​
map_prompt
None
default:"None"
Map step prompt.
​
reduce_prompt
None
default:"None"
Reduce step prompt.
​
knowledge_prompt
None
default:"None"
Knowledge generation prompt.
​
max_context_tokens
int
default:"12000"
Maximum context tokens.
​
data_max_tokens
int
default:"12000"
Maximum data tokens.
​
map_max_length
int
default:"1000"
Maximum map response length in words.
​
reduce_max_length
int
default:"2000"
Maximum reduce response length in words.
​
dynamic_search_threshold
int
default:"1"
Community rating threshold for inclusion.
​
dynamic_search_keep_parent
bool
default:"False"
Keep parent community if children are relevant.
​
dynamic_search_num_repeats
int
default:"1"
Number of times to rate each community.
​
dynamic_search_use_summary
bool
default:"False"
Use community summary instead of full context.
​
dynamic_search_max_level
int
default:"2"
Maximum community hierarchy level.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.

​
StorageDefaults

​
type
str
default:"file"
Storage type (from StorageType enum).
​
encoding
str | None
default:"None"
Text encoding.
​
base_dir
str | None
default:"None"
Base directory for file storage.
​
azure_connection_string
None
default:"None"
Azure connection string.
​
azure_container_name
None
default:"None"
Azure container name.
​
azure_account_url
None
default:"None"
Azure account URL.
​
azure_cosmosdb_account_url
None
default:"None"
Azure CosmosDB account URL.

​
InputDefaults

​
type
InputType
default:"Text"
Input type.
​
encoding
str | None
default:"None"
Text encoding.
​
file_pattern
None
default:"None"
File pattern for matching input files.
​
id_column
None
default:"None"
Column name for document IDs.
​
title_column
None
default:"None"
Column name for document titles.
​
text_column
None
default:"None"
Column name for document text.

​
InputStorageDefaults

Extends StorageDefaults.
​
base_dir
str
default:"input"
Base directory for input storage.

​
CacheStorageDefaults

Extends StorageDefaults.
​
base_dir
str
default:"cache"
Base directory for cache storage.

​
CacheDefaults

​
type
CacheType
default:"Json"
Cache type.
​
storage
CacheStorageDefaults
Cache storage configuration.

​
LocalSearchDefaults

​
prompt
None
default:"None"
Local search prompt.
​
text_unit_prop
float
default:"0.5"
Text unit proportion.
​
community_prop
float
default:"0.15"
Community proportion.
​
conversation_history_max_turns
int
default:"5"
Maximum conversation history turns.
​
top_k_entities
int
default:"10"
Top k entities to retrieve.
​
top_k_relationships
int
default:"10"
Top k relationships to retrieve.
​
max_context_tokens
int
default:"12000"
Maximum context tokens.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
embedding_model_id
str
default:"default_embedding_model"
Embedding model ID.

​
OutputStorageDefaults

Extends StorageDefaults.
​
base_dir
str
default:"output"
Base directory for output storage.

​
PruneGraphDefaults

​
min_node_freq
int
default:"2"
Minimum node frequency.
​
max_node_freq_std
None
default:"None"
Maximum node frequency standard deviation.
​
min_node_degree
int
default:"1"
Minimum node degree.
​
max_node_degree_std
None
default:"None"
Maximum node degree standard deviation.
​
min_edge_weight_pct
float
default:"40.0"
Minimum edge weight percentage.
​
remove_ego_nodes
bool
default:"True"
Whether to remove ego nodes.
​
lcc_only
bool
default:"False"
Keep only largest connected component.

​
ReportingDefaults

​
type
ReportingType
default:"file"
Reporting type.
​
base_dir
str
default:"logs"
Base directory for reporting.
​
connection_string
None
default:"None"
Connection string for blob reporting.
​
container_name
None
default:"None"
Container name for blob reporting.
​
storage_account_blob_url
None
default:"None"
Storage account blob URL.

​
SnapshotsDefaults

​
embeddings
bool
default:"False"
Whether to save embedding snapshots.
​
graphml
bool
default:"False"
Whether to save GraphML snapshots.
​
raw_graph
bool
default:"False"
Whether to save raw graph snapshots.

​
SummarizeDescriptionsDefaults

​
prompt
None
default:"None"
Summarization prompt.
​
max_length
int
default:"500"
Maximum summary length in tokens.
​
max_input_tokens
int
default:"4000"
Maximum input tokens.
​
completion_model_id
str
default:"default_completion_model"
Completion model ID.
​
model_instance_name
str
default:"summarize_descriptions"
Model instance name for caching.

​
UpdateOutputStorageDefaults

Extends StorageDefaults.
​
base_dir
str
default:"update_output"
Base directory for update output storage.

​
VectorStoreDefaults

​
type
str
default:"lancedb"
Vector store type (from VectorStoreType enum).
​
db_uri
str
default:"output/lancedb"
Database URI for vector store.

​
GraphRagConfigDefaults

Root configuration defaults.
​
models
dict
default:"{}"
Legacy model configurations.
​
completion_models
dict
default:"{}"
Completion model configurations.
​
embedding_models
dict
default:"{}"
Embedding model configurations.
​
concurrent_requests
int
default:"25"
Default concurrent requests.
​
async_mode
AsyncType
default:"Threaded"
Default async mode.
​
reporting
ReportingDefaults
Reporting configuration defaults.
​
input_storage
InputStorageDefaults
Input storage configuration defaults.
​
output_storage
OutputStorageDefaults
Output storage configuration defaults.
​
update_output_storage
UpdateOutputStorageDefaults
Update output storage configuration defaults.
​
cache
CacheDefaults
Cache configuration defaults.
​
input
InputDefaults
Input configuration defaults.
​
embed_text
EmbedTextDefaults
Text embedding configuration defaults.
​
chunking
ChunkingDefaults
Chunking configuration defaults.
​
snapshots
SnapshotsDefaults
Snapshots configuration defaults.
​
extract_graph
ExtractGraphDefaults
Graph extraction configuration defaults.
​
extract_graph_nlp
ExtractGraphNLPDefaults
NLP graph extraction configuration defaults.
​
summarize_descriptions
SummarizeDescriptionsDefaults
Description summarization configuration defaults.
​
community_reports
CommunityReportDefaults
Community reports configuration defaults.
​
extract_claims
ExtractClaimsDefaults
Claims extraction configuration defaults.
​
prune_graph
PruneGraphDefaults
Graph pruning configuration defaults.
​
cluster_graph
ClusterGraphDefaults
Graph clustering configuration defaults.
​
local_search
LocalSearchDefaults
Local search configuration defaults.
​
global_search
GlobalSearchDefaults
Global search configuration defaults.
​
drift_search
DriftSearchDefaults
DRIFT search configuration defaults.
​
basic_search
BasicSearchDefaults
Basic search configuration defaults.
​
vector_store
VectorStoreDefaults
Vector store configuration defaults.
​
workflows
None
default:"None"
Workflows list.

Build docs developers (and LLMs) love

Get started for freeTalk to us