Add a query entry

Answer a question about a document with one function call This is the simplest way to use GraphRAG. Just provide text content and a question, and get

Build the final registry

Calculate confidence score for an entity

Classify the intent of a query. Analyzes the query to determine its primary intent, detect ambiguity, identify inappropriate content, and decide whet

Calculate completeness score (0.0 to 1.0) based on populated fields Higher scores indicate more complete metadata

Compute delta between last snapshot and current state Returns the minimal set of changes since the last snapshot.

Access the underlying Ollama configuration

Create a test knowledge graph with varying sizes for benchmarking

Detect browser and version

Embed multiple texts (batched)

Enrich multiple chunks with metadata

Extract entities from text

Extract approximate Unix timestamp from temporal marker if possible This is a best-effort extraction and may not work for all formats. Returns None i

Force propagate all pending lazy updates Should be called before queries to ensure fresh results. Automatically called when `propagate_on_query` is e

Generate extractive summary using simple sentence ranking

Get all entities in a specific community at a given level # Arguments `level` - Hierarchical level (0 = finest) `community_id` - Community identifier

Get functions by category

Compute content hash for a node

Enhanced check if a word could be part of a person's name

Compute Jaro-Winkler similarity

Calculate lexical cohesion between two texts (word overlap)

Helper function to load fixture files

Build an OllamaConfig pointing at localhost with KV-cache settings. `OLLAMA_MODEL` env var overrides the default model — useful when you want to test

Match a query against the knowledge graph using fuzzy semantic similarity. Executes a multi-stage matching process: 1. Direct entity name matching (e

Merge another index into this one Useful for combining indices from multiple documents

Parse inline markdown markers (**bold**, *italic*, `code`) from a string. Returns `Span<'static>` values using owned strings.

Perform POS tagging on text

Propagate all pending updates

Put a value into the cache, evicting the oldest entry if at capacity

Put a value into the cache

Queue a node update for lazy propagation

Record a failed request

Run the application

Search for similar vectors

Search for k nearest neighbors using cosine similarity

Get snapshot for time range

Split text into sentences (simple heuristic)

Split text into sentences (simple implementation) This is a basic sentence splitter. For production, consider using a dedicated NLP library or more s

Start the batch processor This spawns worker tasks that continuously process incoming operations. Call this once during initialization.

Get structured suggestions for recovering from this error Returns actionable suggestions with optional code examples. # Example ```rust use graphrag

Get style for title text

Tokenize text into normalized tokens

Update graph statistics in info panel

Set the chunk size for text processing (optional)

Add mentions to the entity

Add metadata to the document

Set number of hierarchy levels

Add a query template # Arguments `template` - Query template

Analyze query and determine complexity

Ask a question and get an answer

Generate embeddings for multiple texts in batch with parallel processing

Increase weight for a specific objective

Build a streaming response from subquery results Synthesizes multiple subquery results into a coherent response using the configured synthesis strate

Calculate adaptive threshold based on content characteristics

Calculate F1 score based on token overlap

Calculate slope of a metric over recent window

Check rate limit for a user

Chunk text and enrich with semantic metadata

Calculate cosine similarity between two vectors

Create scored chunks from text and split positions

Create a collection with the specified dimension # Arguments `dimension` - Embedding dimension (e.g., 384 for MiniLM, 768 for BERT)

Create hybrid results from combined scores

Create a new workspace

End span and record duration

Execute a query with caching

Extract version number from user agent string

Extract entities and relationships from text

Extract relationships between entities in the same chunk

Fill template with provided values

Find a safe character boundary at or before the given position

Generate evaluation prompt for a query result

Generate a cache key with generation parameters