Extract type information from Jedi analysis.

Extract JSDoc comment from JavaScript function node.

Extract dynamic JSON loading patterns like fetch(), require(), json.load().

Extract link relations from href attributes and form actions.

Extract metadata from messages.

Extract named entity from Tree-sitter node with enhanced observations.

Extract content from nested arrays (replies, comments).

Extract entities and relations from JSON object structure.

Extract IndexerConfig-compatible overrides from project config.

Extract dependencies from package.json.

Extract parameter patterns from function signature.

Extract Python docstring from function or class node.

Extract return patterns from function.

Extract return type annotation from function signature.

Extract section content with intelligent chunking and semantic boundaries.

Extract entities from Tree-sitter AST.

Extract relations from Tree-sitter AST (inheritance + imports for debugging).

Extract TypeScript configuration info.

ENHANCED variable extraction with comprehensive fixes: 1. Destructuring pattern support (object_pattern, array_pattern) 2. En

Extract multiple variables from complex assignment patterns (tuple unpacking, etc.).

Extract variables from walrus operator (:=) expressions.

Filter out relations pointing to non-existent entities - handles CALLS and IMPORTS.

Find the function that contains the given node.

Find what the decorator applies to (class, method, property).

Fallback implementation using search_similar.

Find files modified since timestamp using state cache.

Find the first object node in the tree.

Format message content with proper HTML rendering.

Generate full conversation display section.

Generate conversation-specific memory entries based on conversation content.

Generate debugging information section.

Generate header section with metadata.

Generate complete HTML content.

Generate output path for HTML report.

Generate related memory entries section.

Generate GPT analysis summary section.

Lazy initialize BM25 embedder for sparse vectors.

Generate cache key for model state.

Get content hash from item.

Generate hash of corpus for caching.

Get CSS styles for the HTML report.

Get the current debug log file to use.

Get detailed health analysis for troubleshooting.

Get effective token limit for batching.

Calculate SHA256 hash of file contents.

Calculate SHA256 hash of file contents.

Calculate SHA256 hash of file contents.

Calculate SHA256 hash of file contents.

Get actionable recommendations based on health status.

Get health status using direct API call - example implementation.

Get JavaScript for interactive features.

Get enhanced docstring from Jedi analysis.

Extract latest mtime from state for timestamp filtering.

Extract property name from property definition.

Generate cache key for search query.

Get state directory (configurable for test isolation).

Extract text content from style element.

Get text count limit for batching.

Get count of files actually tracked in indexer state files for this collection.

Extract type hints from Jedi definition.

Handle Docker Compose files.

Handle a single file system event.

Handle GitHub Actions workflow files.

Handle Kubernetes manifest files.

Special handling for known JSON file types.

Hybrid search using Reciprocal Rank Fusion (RRF). Args: collection_name: Name of the collection dense_vector:

Initialize tiktoken encoder for the model.

Initialize TypeScript language server (stub for future implementation).

Initialize Tree-sitter and Jedi parsers.

Check if function name is a built-in or common library function.

Check if a call relation is phantom (entities exist but call doesn't). Args: all_points: All points from the collection for effic

Enhanced logic for v2.4 chunk format.

Load conversation from ID or file path.

Load trained model from cache.

Load project-specific configuration.

Parse markdown into hierarchical sections with token counts.

Parse individual message from JSONL data.

Parse OpenAI response into SummaryResult.

Parse file with Tree-sitter.

Prepare conversation text for OpenAI processing.

Print database overview statistics with improved layout.

Main event processing loop.

Process a file deletion using shared deletion logic.

Process Jedi analysis results - IMPORTS ONLY (entities handled by Tree-sitter).

Process a single streamed item.

Redistribute sections from undersized chunks to meet minimum token requirements.

Register default embedder implementations.

Register default parsers.

Register default storage implementations.

Convert relation to text for embedding.

Reliable batch upsert with splitting, timeout handling, and retry logic.

Perform reliable batch upsert with verification.

Run initial indexing to ensure collection exists and project is indexed.

Save trained model to cache.

Check if a file should be processed.

Determine if file entities should be replaced.

Determine if an error should trigger a retry.

Split points into batches of specified size.

Split sections that exceed token limit at semantic boundaries.

Start background timer thread to automatically process files.