gitnexus wiki

Generate repository wiki from knowledge graph

Syntax

gitnexus wiki [path]

Description

Generates LLM-powered repository documentation from the knowledge graph. The wiki is built by analyzing clusters (functional communities) and using an LLM to describe each module’s purpose, architecture, and key components.

The wiki includes:

• Architecture overview — High-level structure and design patterns
• Module documentation — Each functional cluster gets a dedicated page
• Component diagrams — Auto-generated from the knowledge graph
• Interactive viewer — Single-file HTML with search and navigation

Options

path · string

Path to the repository. Defaults to current directory’s git root.

--force · boolean

Force full regeneration even if the wiki is up to date with the current index.

--model · string

LLM model name.

Default: minimax/minimax-m2.5 (via OpenRouter)

Examples:

• gpt-4o-mini (OpenAI)
• claude-3-5-sonnet-20241022 (OpenRouter)
• llama-3.1-70b-instruct (custom endpoint)

--base-url · string

LLM API base URL.

Default: https://openrouter.ai/api/v1

Examples:

• https://api.openai.com/v1 (OpenAI)
• http://localhost:11434/v1 (Ollama)

--api-key · string

LLM API key. Saved to ~/.gitnexus/config.json for future use.

Can also be set via environment variables:

• GITNEXUS_API_KEY
• OPENAI_API_KEY

--concurrency · string

Number of parallel LLM calls.

Default: 3

--gist · boolean

Publish wiki as a public GitHub Gist after generation.

Requires GitHub CLI (gh) installed and authenticated.

Usage Examples

Generate wiki (interactive setup)

gitnexus wiki

First-time usage prompts for LLM configuration:

GitNexus Wiki Generator

No LLM configured. Let's set it up.

Supports OpenAI, OpenRouter, or any OpenAI-compatible API.

[1] OpenAI (api.openai.com)
[2] OpenRouter (openrouter.ai)
[3] Custom endpoint

Select provider (1/2/3):

Use OpenAI

gitnexus wiki --model gpt-4o-mini --base-url https://api.openai.com/v1 --api-key sk-...

Use custom model

gitnexus wiki --model llama-3.1-70b-instruct --base-url http://localhost:11434/v1

Force regeneration

gitnexus wiki --force

Publish to GitHub Gist

gitnexus wiki --gist

High concurrency for faster generation

gitnexus wiki --concurrency 10

Output Example

GitNexus Wiki Generator

████████████████████████████████████████ 100% | Done

Wiki generated successfully (45.3s)

Mode: full
Pages: 18
Output: /Users/dev/projects/my-app/.gitnexus/wiki
Viewer: /Users/dev/projects/my-app/.gitnexus/wiki/index.html

Publish wiki as a GitHub Gist for easy viewing? (Y/n):

Interactive Setup

If no LLM configuration is found and you’re in an interactive terminal, the command walks you through setup:

1. Provider selection — OpenAI, OpenRouter, or custom endpoint
2. Model selection — With smart defaults per provider
3. API key — Optionally reuses existing env vars
4. Save to config — Stored in ~/.gitnexus/config.json

After setup, future runs skip the prompts.

GitHub Gist Publishing

The --gist flag (or interactive prompt) publishes the wiki as a public GitHub Gist:

gitnexus wiki --gist

Requirements:

• GitHub CLI installed
• Authenticated with gh auth login

Output:

Publishing to GitHub Gist...
Gist:   https://gist.github.com/username/abc123
Viewer: https://gistcdn.githack.com/username/abc123/raw/index.html

The viewer URL uses githack.com to serve the HTML with proper content-type headers.

Wiki Structure

Generated wiki:

.gitnexus/wiki/
├── index.html          # Interactive viewer (single file)
├── architecture.md     # High-level overview
├── module-auth.md      # Auth module docs
├── module-api.md       # API module docs
└── ...

The viewer (index.html) is a self-contained single-file app with:

• Sidebar navigation
• Full-text search
• Markdown rendering
• Syntax highlighting

When Wiki is Up-to-Date

The wiki generator tracks the last indexed commit. If the index hasn’t changed since the last wiki generation, it skips regeneration:

Wiki is already up to date.
Viewer: /Users/dev/projects/my-app/.gitnexus/wiki/index.html

Use --force to regenerate anyway.

Failed Modules

If some modules fail to generate (e.g., LLM errors, rate limits), the command reports them:

Failed modules (3):
  - module-auth
  - module-api
  - module-db

Re-run to retry failed modules (pages will be regenerated).

Failed modules are automatically retried on the next run.

LLM Cost Estimate

For a typical repository with 20-30 modules:

• OpenAI (gpt-4o-mini): ~$0.50-$1.00
• OpenRouter (minimax/minimax-m2.5): ~$0.10-$0.30
• Ollama (local): Free

Performance

Generation time depends on:

• Number of modules (clusters)
• LLM speed
• Concurrency setting

Typical times:

• Small repo (5-10 modules): 10-20 seconds
• Medium repo (20-30 modules): 30-60 seconds
• Large repo (50+ modules): 2-5 minutes

Increase concurrency for faster generation:

gitnexus wiki --concurrency 10

Related Commands

• gitnexus analyze — Index repository first
• gitnexus status — Check index status