graph

name	graph
description	Query the CodeGraphContext code graph for cross-cutting analysis (callers, callees, dead code, dependency chains)
argument-hint	[callers <func>|calls <func>|dead-code|complexity|chain <from> <to>|tree <class>|deps <module>|find <pattern>|reindex|stats]
user-invocable	true
allowed-tools	Bash, Read

CodeGraphContext Graph Query

Query the CGC code graph for structural analysis that complements Serena's symbol-level intelligence.

Prerequisites

• Neo4j Docker container must be running: docker ps | grep cgc-neo4j
• If not running: docker start cgc-neo4j
• CGC venv at .venv-cgc/ with codegraphcontext installed

CRITICAL: Command Prefix

The project .env overrides CGC config. ALL commands MUST use this prefix:

cd "C:/Users/Anirudh Mohan/Desktop/FloodSafe"
source .venv-cgc/Scripts/activate
PYTHONIOENCODING=utf-8 NEO4J_URI=bolt://localhost:7687 NEO4J_USERNAME=neo4j NEO4J_PASSWORD=floodsafe123 cgc -db neo4j <command>

When to Use CGC vs Serena

Need	Use	Why
Find symbol definition + body	Serena find_symbol	Precise, includes body
Find direct references to a symbol	Serena find_referencing_symbols	Language-server powered
Transitive call chains (N-hop)	CGC analyze callers / analyze calls	Graph traversal
Dead code detection	CGC analyze dead-code	Cross-language, whole-repo
Cyclomatic complexity report	CGC analyze complexity	Threshold-based filtering
Call chain between two symbols	CGC analyze chain	Graph shortest-path
Class inheritance tree	CGC analyze tree	Hierarchy visualization
Module dependencies	CGC analyze deps	Import-level relationships
Safe refactoring (rename/replace)	Serena editing tools	LSP-powered
Impact analysis before refactoring	CGC then Serena	CGC for scope, Serena for execution

Commands

Based on $ARGUMENTS, run the appropriate query:

callers <function_name> — Who Calls This Function?

cgc -db neo4j analyze callers "$FUNCTION_NAME"

Shows the full call chain leading to a function. Useful for impact analysis before refactoring.

calls <function_name> — What Does This Function Call? (Callees)

cgc -db neo4j analyze calls "$FUNCTION_NAME"

Shows everything a function depends on. Useful for understanding complexity and blast radius.

dead-code — Unreferenced Symbols

cgc -db neo4j analyze dead-code
# Exclude FastAPI route decorators (they're entry points, not dead code):
cgc -db neo4j analyze dead-code --exclude route,router,get,post,put,delete,api

Finds unreferenced functions/classes. Review before deleting — some may be:

• FastAPI router entry points (decorated with @router.get/post)
• React components (imported in JSX)
• Event handlers (referenced in HTML/templates)

complexity — Cyclomatic Complexity

cgc -db neo4j analyze complexity --threshold 10

Lists functions exceeding the complexity threshold. Candidates for refactoring.

chain <from> <to> — Call Chain Between Two Symbols

cgc -db neo4j analyze chain "$SYMBOL_A" "$SYMBOL_B"

Finds the shortest call chain connecting two symbols. Useful for tracing data flow.

tree <class_name> — Class Inheritance Hierarchy

cgc -db neo4j analyze tree "$CLASS_NAME"

Shows the inheritance hierarchy for a class.

deps <module_name> — Module Dependencies

cgc -db neo4j analyze deps "$MODULE_NAME"

Shows imports and dependencies for a module.

find <pattern> — Pattern Search in Graph

cgc -db neo4j find pattern "$PATTERN"

Search for symbols matching a pattern in the graph database. Returns up to 50 matches with type, location, and source.

reindex — Re-index Codebase

cgc -db neo4j index apps/backend/src
cgc -db neo4j index apps/frontend/src
cgc -db neo4j index apps/ml-service/src
cgc -db neo4j index apps/iot-ingestion

NOTE: Root cgc index . doesn't work well due to .cgcignore exclusions. Index subdirectories explicitly.

stats — Indexing Statistics

cgc -db neo4j stats

Shows files, functions, classes, modules, and repositories in the graph.

Visualization

Add --viz or -V to any analyze/find command to generate an interactive HTML graph:

cgc -db neo4j analyze callers "FHICalculator" --viz

Indexed Directories (as of 2026-03-08)

Directory	Files	Content
apps/backend/src	~120	FastAPI routers, services, models
apps/frontend/src	~180	React components, hooks, contexts
apps/ml-service/src	~60	ML models, FHI calculator
apps/iot-ingestion	~6	IoT data ingestion
Total	366 files, 2082 functions, 339 classes

Output

Present results as:

1. The call chain, dependency list, or analysis results
2. File paths for each node
3. Actionable insight (e.g., "safe to delete", "high fan-in — refactor carefully", "circular dependency detected")

Troubleshooting

Issue	Fix
"Database not configured" / fallback to FalkorDB	Ensure env vars are set: NEO4J_URI, NEO4J_USERNAME, NEO4J_PASSWORD + -db neo4j flag
Neo4j container not running	docker start cgc-neo4j
Unicode/encoding errors	Ensure PYTHONIOENCODING=utf-8 is set
"No indexed repos" / 0 files	Index subdirectories explicitly (see reindex section)
Stale results after refactor	Re-index affected directories