Module Reference

Complete reference for all 26 packages in semantic_code_intelligence/.

analysis

Code explanation and repository summarization.

Symbol	Type	Description
`LanguageStats`	class	Per-language statistics (files, lines, symbols)
`RepoSummary`	class	Structured repository summary
`CodeExplanation`	class	Explanation result for a symbol or file
`summarize_repo()`	function	Generate full repository summary
`explain_symbol()`	function	Produce structural explanation of a symbol
`explain_file()`	function	Explain all symbols in a file
`detect_languages()`	function	Detect languages in the codebase

bridge

HTTP bridge server for IDE extensions and AI agents.

Symbol	Type	Description
`BridgeServer`	class	HTTP server implementing the bridge protocol
`AgentRequest`	class	Incoming bridge request with kind and payload
`AgentResponse`	class	Response with correlation ID and result
`ContextProvider`	class	Provides context to bridge consumers
`VSCodeBridge`	class	VS Code-specific bridge adapter
`RequestRouter`	class	Routes requests by kind
`ResponseFormatter`	class	Formats results for agent consumption
`start_bridge()`	function	Start the bridge server
`handle_request()`	function	Process a single bridge request

ci

Quality analysis, metrics, hotspots, impact analysis, and CI gates.

Symbol	Type	Description
`QualityAnalyzer`	class	Runs full quality analysis pipeline
`QualitySnapshot`	class	Point-in-time quality metrics
`QualityTrend`	class	Trend analysis over snapshots
`QualityGate`	class	Configurable pass/fail gates
`Hotspot`	class	High-risk code location
`HotspotAnalyzer`	class	Detects hotspots via complexity, churn, fan-in
`ImpactReport`	class	Blast radius analysis result
`ImpactAnalyzer`	class	Predicts change impact via call graph BFS
`MetricsCollector`	class	Collects code metrics
`BanditIssue`	class	Security issue found by Bandit
`run_quality_analysis()`	function	Run full quality pipeline
`run_bandit_scan()`	function	Run Bandit security scan
`calculate_maintainability_index()`	function	Compute maintainability score
`detect_hotspots()`	function	Find high-risk code areas
`analyze_impact()`	function	Compute blast radius of changes

cli

36 Click commands organized by capability. All commands support --json for machine-readable output.

See CLI Reference for the complete command documentation.

config

Pydantic configuration models loaded from .codex/config.json.

Symbol	Type	Description
`AppConfig`	class	Root configuration aggregating all sub-configs
`EmbeddingConfig`	class	Model name, chunk size, overlap settings
`SearchConfig`	class	top_k, similarity threshold, hybrid weights
`IndexConfig`	class	Incremental indexing, file extensions
`LLMConfig`	class	Provider, model, API key, temperature
`QualityConfig`	class	Quality gate thresholds
`load_config()`	function	Load config from `.codex/config.json`

context

Context building for AI consumption — symbol resolution, call graphs, dependency maps.

Symbol	Type	Description
`ContextBuilder`	class	Builds rich context windows for AI tools
`ContextWindow`	class	Bounded token-aware context container
`CallGraph`	class	Function call graph (callers/callees)
`DependencyMap`	class	File-level dependency graph
`SessionMemory`	class	Multi-turn conversation memory with persistence
`SymbolResolver`	class	Resolves symbol names to definitions
`TokenCounter`	class	Estimates token count for context budgets
`ReferenceCollector`	class	Collects all references to a symbol

daemon

Background file watching and incremental indexing.

Symbol	Type	Description
`IndexingDaemon`	class	Long-running daemon for watch mode
`FileWatcher`	class	Filesystem event watcher
`AsyncIndexer`	class	Async incremental indexer
`DaemonConfig`	class	Daemon configuration

embeddings

Sentence-transformer model management and vector encoding.

Symbol	Type	Description
`ModelInfo`	class	Metadata about an embedding model
`load_model()`	function	Load a sentence-transformer model
`encode_text()`	function	Encode text to vector
`encode_batch()`	function	Batch encode multiple texts
`list_models()`	function	List available models
`download_model()`	function	Download a model by name
`switch_model()`	function	Switch active embedding model
`get_model_info()`	function	Get model metadata

evolution

Self-improving development loop with budget awareness.

Symbol	Type	Description
`EvolutionEngine`	class	Orchestrates the self-improvement loop
`BudgetGuard`	class	Enforces token/time/cost budgets
`TaskSelector`	class	Selects next improvement task
`PatchGenerator`	class	Generates code patches via LLM
`TestRunner`	class	Runs tests to validate patches
`EvolutionResult`	class	Result of an evolution cycle
`EvolutionTask`	class	A single improvement task
`EvolutionBudget`	class	Budget configuration
`evolve()`	function	Run one evolution cycle

indexing

File scanning, code chunking, and semantic chunk creation.

Symbol	Type	Description
`CodeChunk`	class	A chunk of source code with metadata
`ScannedFile`	class	A scanned source file
`SemanticChunk`	class	Chunk with embedding vector attached
`scan_directory()`	function	Scan a directory for source files
`chunk_file()`	function	Split a file into semantic chunks
`run_indexing()`	function	Full indexing pipeline
`incremental_index()`	function	Index only changed files

llm

LLM provider abstraction with caching, streaming, and investigation chains.

Symbol	Type	Description
`LLMProvider`	class	Abstract base for LLM providers
`OpenAIProvider`	class	OpenAI API provider
`OllamaProvider`	class	Ollama local inference provider
`MockProvider`	class	Mock provider for testing
`CachedProvider`	class	Caching wrapper around any provider
`LLMCache`	class	Disk-based LLM response cache
`InvestigationChain`	class	ReAct-style autonomous investigation
`StreamHandler`	class	Handles streaming LLM responses
`create_provider()`	function	Factory for LLM providers

lsp

Language Server Protocol implementation.

Symbol	Type	Description
`LSPServer`	class	LSP server for editor integration
`start_lsp()`	function	Start the LSP server

mcp

Model Context Protocol server using the official MCP SDK.

Symbol	Type	Description
`run_mcp_server()`	function	Start the MCP stdio server
`_create_server()`	function	Create and configure MCP server instance
`_dispatch_tool()`	function	Route tool calls to the executor

parsing

Tree-sitter AST parsing for 12 programming languages.

Symbol	Type	Description
`Symbol`	class	Extracted symbol (function, class, method) with location
`parse_file()`	function	Parse a file into an AST
`extract_symbols()`	function	Extract symbols from AST
`get_language()`	function	Get tree-sitter language for file extension

Supported languages: Python, JavaScript, TypeScript, Java, Go, Rust, Ruby, C, C++, C#, PHP, Swift

plugins

Extensible plugin system with 22 hook points.

Symbol	Type	Description
`PluginManager`	class	Discovery, loading, lifecycle management
`PluginBase`	class	Abstract base class for plugins
`PluginHook`	enum	22 hook point definitions
`PluginMetadata`	class	Plugin name, version, author, hooks
`PluginRegistry`	class	Registry of loaded plugins
`load_plugins()`	function	Discover and load plugins from directory

scalability

Performance optimizations for large codebases.

Symbol	Type	Description
`BatchProcessor`	class	Process items in configurable batches
`MemoryAwareEmbedder`	class	Embedder that respects memory limits
`ParallelScanner`	class	Multi-threaded file scanning
`StreamingIndexer`	class	Streaming indexer for large repos

search

FAISS vector search, BM25 keyword search, and hybrid RRF fusion.

Symbol	Type	Description
`HybridResult`	class	Combined search result with score
`BM25Index`	class	BM25 keyword index
`FaissSearcher`	class	FAISS vector similarity search
`hybrid_search()`	function	Combined vector + keyword search with RRF
`vector_search()`	function	Pure vector similarity search
`keyword_search()`	function	Pure BM25 keyword search

services

Service-layer abstractions for indexing and search results.

Symbol	Type	Description
`IndexingResult`	class	Result of an indexing operation
`SearchResult`	class	Normalized search result

storage

Persistent storage for vectors, symbols, and metadata.

Symbol	Type	Description
`VectorStore`	class	FAISS-backed vector storage
`SymbolRegistry`	class	Symbol definition and reference store
`IndexStats`	class	Index size, file count, chunk count
`QueryHistory`	class	Search query history
`ChunkStore`	class	Stores code chunks with metadata
`CacheManager`	class	Manages disk caches

tools

AI Agent Tool Protocol — structured tool invocation for LLM agents.

Symbol	Type	Description
`ToolExecutor`	class	Executes tool invocations
`ToolInvocation`	class	Typed tool call request
`ToolExecutionResult`	class	Typed tool call response
`ToolError`	class	Tool execution error
`ToolRegistry`	class	Registry of available tools
`ToolSchema`	class	JSON Schema for a tool

tui

Textual-based interactive terminal UI.

Symbol	Type	Description
`CodexaTUI`	class	Main TUI application
`start_tui()`	function	Launch the interactive terminal

utils

Logging and utility functions.

Symbol	Type	Description
`setup_logging()`	function	Configure rich logging
`get_logger()`	function	Get a named logger

web

Web UI and REST API server.

Symbol	Type	Description
`WebServer`	class	HTTP server combining API and UI
`APIHandler`	class	REST API request handler
`UIHandler`	class	Static file and UI handler
`start_web()`	function	Start the web server

workspace

Multi-repo workspace management.

Symbol	Type	Description
`Workspace`	class	Multi-repo workspace container
`WorkspaceManifest`	class	Workspace configuration manifest
`create_workspace()`	function	Create a new workspace

Module Reference ​

analysis ​

bridge ​

ci ​

cli ​

config ​

context ​

daemon ​

embeddings ​

evolution ​

indexing ​

llm ​

lsp ​

mcp ​

parsing ​

plugins ​

scalability ​

search ​

services ​

storage ​

tools ​

tui ​

utils ​

web ​

workspace ​