AELSHA™ NEHEMVIX

Understand an unfamiliar codebase before you change it. Nehemvix turns any Git repository into a grounded guide for beginners, senior engineers or architects. Every guide includes architecture diagrams and citations to exact lines, and every file and code name in it is checked against the code. It runs with your own AI model key; the key is never stored.

This package is the command-line tool. It runs the same pipeline as the web app at nehemvix.aelsha.com, on your own machine.

Install

You need Python 3.12 or newer, git 2.50.1 or newer, Graphviz (diagrams) and Pango (PDF export):

pipx install aelsha-nehemvix
nehemvix --version

What's new in 2.1.0

• nehemvix finish <job-id>: writes only the sections a finished guide is missing (for example after its cap ran out), on any model, without paying twice for finished work
• For AI agents: --json (one JSON result on stdout, with a published schema), --events jsonl, and exit code 4 when a person must approve the cost first
• Read commands: nehemvix list, status, outline, section, search, readiness
• A framework-neutral Python API: from nehemvix.agentcli.api import Nehemvix (estimate, start with a cap, wait, pause, resume, read; typed errors)
• nehemvix mcp: a local MCP server for your agents, read-only unless you set a spending ceiling. Tested with LangChain, LangGraph, Google ADK, OpenAI Agents SDK, Claude Agent SDK, Microsoft Agent Framework and CrewAI (guide)

What's new in 2.0.0

• A new name: AELSHA™ NEHEMVIX. The command is nehemvix and the package is aelsha-nehemvix; your progress and remembered file summaries move to ~/.nehemvix the first time you run it

What's new in 1.1.0

• nehemvix ask <job-id> "question": answers from a finished guide, with its sources
• --format html,pdf,docx,pptx,drawio: the same exports as the web app
• --git-token-env NAME: private repositories, with a token from a variable (never stored)
• nehemvix doctor: checks your machine before a run
• A readable estimate that says which --cap is enough, one-line errors with stable exit codes, and Ctrl-C that tells you how to resume

Every release: changelog.

Run

Put your key in an environment variable, never on the command line:

export OPENAI_API_KEY=...
nehemvix run https://github.com/octocat/Hello-World --provider openai --model gpt-6-luna --cap 0.50

The run continues only if the estimate's upper end is within --cap (US dollars). Otherwise it stops before spending anything. The guide, an assessment and a cost report are written to ./out.

Nehemvix works with Google Gemini, Vertex AI, Anthropic Claude, OpenAI, AWS Bedrock, Microsoft Foundry and any OpenAI-compatible service, including free models on OpenRouter at $0. If a model's limit runs out, the run pauses; nehemvix resume <job-id> continues later on the same or another model, without paying twice for finished work. If a guide finished with sections missing (its cap ran out while writing), nehemvix finish <job-id> writes only those.

Learn more

• Command-line guide: every provider, option, output and exit code
• Integrations: GitHub Actions, GitLab CI, Docker, make and a grounding-score gate
• Set up your AI provider
• Changelog: what changed in each release
• How every guide is checked and security
• Security reports: security@aelsha.com

Built with AELSHA™ ASTRA by AELSHA INNOVATIONS OPC PRIVATE LIMITED. Licensed under the Apache License 2.0.