Call Graph Database Reference

Purpose: Code call graph analysis for impact assessment and code navigation Storage: org.db (knowledge graph) + sessions.db (legacy tables) MCP Servers: coditect-call-graph, coditect-impact-analysis Task: J.21.5.2

---

Overview

The call graph system indexes function definitions and call relationships from source files. It is accessed through MCP (Model Context Protocol) servers rather than direct database queries. The data is stored across two databases:

1. org.db — Knowledge graph nodes/edges (primary, ADR-151)
2. sessions.db — Legacy call_graph_* tables (backward compatibility)

Note: call_graph.db exists as an empty file at ~/.coditect-data/context-storage/call_graph.db (0 bytes). It was intended as a standalone database but the call graph data was integrated into the knowledge graph (org.db) and legacy tables (sessions.db) instead.

---

Statistics (as of 2026-02-19)

Knowledge Graph (org.db — Primary)

Metric	Value
Functions Indexed	7,066
Call Edges	2,565
Files Indexed	577
Python Functions	7,011
TypeScript Functions	25
JavaScript Functions	30

Legacy Tables (sessions.db)

Metric	Value
Functions	13,379
Edges	91,700
Memory Links	0

---

Data Location

org.db Knowledge Graph (Primary)

Function and call data stored as knowledge graph nodes and edges:

Entity	KG Node Type	Count
Functions	function	7,066
Files	file	577

Relationship	KG Edge Type	Count
Function calls	CALLS	2,565

sessions.db Legacy Tables

-- Function definitions
CREATE TABLE call_graph_functions (
    node_id TEXT PRIMARY KEY,
    function_name TEXT NOT NULL,
    file_path TEXT NOT NULL,
    line_number INTEGER,
    language TEXT
);

-- Call relationships
CREATE TABLE call_graph_edges (
    id INTEGER PRIMARY KEY,
    caller_id TEXT NOT NULL,       -- References call_graph_functions.node_id
    callee_id TEXT NOT NULL,       -- References call_graph_functions.node_id
    call_count INTEGER DEFAULT 1
);

-- Session memory links (context association)
CREATE TABLE call_graph_memory (
    id INTEGER PRIMARY KEY,
    node_id TEXT NOT NULL,          -- References call_graph_functions.node_id
    message_id INTEGER,             -- References messages.id
    context TEXT                    -- Association context
);

---

MCP Server API

The call graph is accessed through two MCP servers:

coditect-call-graph

Tool	Purpose
index_file	Index a source file into the call graph
index_directory	Index all files in a directory
get_callers	Find all functions that call a given function
get_callees	Find all functions called by a given function
call_chain	Find call paths between two functions
memory_linked_search	Search with session history context
call_graph_stats	Database statistics

coditect-impact-analysis

Tool	Purpose
analyze_impact	Blast radius analysis for a function change
analyze_file_impact	Impact analysis for an entire file
find_decisions	Find ADRs related to code
assess_risk	Calculate change risk score
impact_stats	System statistics

Usage Examples

# Via MCP (preferred)
# These are called as tool invocations, not direct Python

# Index a directory
mcp__coditect-call-graph__index_directory(dir_path="scripts/core/")

# Find callers of a function
mcp__coditect-call-graph__get_callers(function_name="get_org_db_path")

# Assess change risk
mcp__coditect-impact-analysis__assess_risk(function_name="publish")

---

Indexing

Files are indexed on demand or during /cx post-processing:

# Triggered by /cx (automatic)
# Indexes ~/.coditect/scripts by default

# Manual reindex via /cx
/cx --reindex-path /path/to/code

# Skip reindex
/cx --no-reindex

Supported languages: Python (.py), JavaScript (.js), TypeScript (.ts)

---

Related Documentation

• DATABASE-SCHEMA.md — Overview of all databases
• ER-DIAGRAMS.md — Visual ER diagrams (sessions.db section)
• DATA-DICTIONARY.md — Field-level documentation
• ADR-151 — Knowledge graph architecture
• CONTEXT-GRAPH-GUIDE.md — Context graph usage guide

---

Generated: 2026-02-19 Task: J.21.5.2