codexray-analyser 0.1.4

Offline-first static analyser for Python files and notebooks.

Codexray

Latest stable package: 0.1.4 (tag v0.1.4)

codexray "C:\path\to\database.py" --summary --findings-only --min-severity high

Codexray is an offline-first Python static analyser for .py and .ipynb files. It helps teams inspect security risks, code quality problems, and dependency patterns without sending source code outside the local machine.

Why Codexray?

You are about to send a notebook or script to a client and want confidence that nothing risky is hidden inside. Run Codexray first. It checks for risky patterns such as shell execution and dangerous builtins, highlights dependency concerns, and gives you a graph of call and import relationships while keeping all source analysis local to your machine.

What Codexray does

• Scans Python files and notebooks.
• Supports project-level scans, full file scans, and targeted line-range checks.
• Produces a structured JSON report.
• Builds a dependency graph view with nodes and edges.
• Flags common security risks such as shell execution and dangerous builtins.

Privacy and security behavior

• No telemetry.
• No source upload.
• Network features are disabled by default.
• Strict input limits for file size, notebook size, snippet length, and AST depth.

This tool is designed for local analysis workflows where proprietary code must stay on the client system.

Installation

pip install codexray-analyser

Quick start with CLI

Analyse a folder:

codexray ./my_project

Analyse a single file:

codexray ./my_project/app.py

Analyse a code snippet:

codexray --snippet "import os; os.system('whoami')"

Analyse only specific lines from one file:

codexray ./my_project/app.py --start-line 40 --end-line 80

Save output to JSON:

codexray ./my_project --output codexray-report.json

Print only a compact summary:

codexray ./my_project --summary

Print findings only (no graph payload):

codexray ./my_project --findings-only

Show only high and critical findings:

codexray ./my_project --findings-only --min-severity high

Export graph for Graphviz:

codexray ./my_project --graph-dot codexray-graph.dot

Export graph to an offline HTML visual:

codexray ./my_project --graph-html codexray-graph.html

Dependency graph preview

Sample graph JSON snippet:

{
  "nodes": [
    { "node_id": "file:database.py", "kind": "file", "label": "database.py" },
    { "node_id": "function:database.py:get_conn", "kind": "function", "label": "get_conn" },
    { "node_id": "import:psycopg2.connect", "kind": "import", "label": "psycopg2.connect" }
  ],
  "edges": [
    { "source": "file:database.py", "target": "function:database.py:get_conn", "relation": "contains" },
    { "source": "function:database.py:get_conn", "target": "import:psycopg2.connect", "relation": "calls" }
  ]
}

Python API usage

from codexray import analyse_file_snippet, analyse_path, analyse_snippet

project_result = analyse_path("./my_project")
snippet_result = analyse_snippet("import os\nos.system('whoami')")
range_result = analyse_file_snippet("./my_project/app.py", 20, 50)

Understanding the report

Each result returns:

• findings: list of detected issues.
• graph: nodes and edges representing imports, files, and function relationships.
• metadata: run information such as analysed path and offline mode state.

Example finding shape:

{
  "rule_id": "SEC002",
  "title": "Dangerous builtin eval",
  "severity": "critical",
  "message": "Avoid eval on untrusted content.",
  "file_path": "src/app.py",
  "line": 18,
  "column": 4
}

Reading graph output

Graph output contains:

• nodes: entities such as files, imports, and functions
• edges: relationships such as imports, contains, and calls
• DOT export for Graphviz via --graph-dot
• offline HTML graph visual via --graph-html

Typical use:

1. Run Codexray and save JSON output.
2. Load graph.nodes and graph.edges into your graph viewer.
3. Track dependency hotspots and risky call paths.

How to use findings to make code changes

Recommended workflow:

1. Sort findings by severity.
2. Fix critical and high findings first.
3. Re-run Codexray after each fix batch.
4. Keep evidence by committing report diffs in your internal workflow.

Examples:

• SEC001 shell execution:
  • Replace dynamic shell calls with safe Python APIs.
  • Avoid passing untrusted input to command execution.
• SEC002 dangerous builtin:
  • Replace eval or exec with safe parsing and strict allow-lists.
• DEP001 unpinned dependency:
  • Pin versions in requirements files with == where practical.

Contributing

Read CONTRIBUTING.md before opening a pull request. Security reports should follow SECURITY.md.

License

MIT. See LICENSE.