Getting Started

Getting Started

This page is not a full deployment manual. It is the shortest stable path to getting GEOFlow running and proving the core workflow.

Step 1: Define the Goal First

Clarify three things:

• whether you are building a single-site content system or a multi-site distribution system
• who the audience is
• which knowledge assets you want to build first

If this part is unclear, models, prompts, tasks, and distribution channels will drift.

Step 2: Get the System Running

Docker is the recommended starting point.

The basic path is:

1. clone the repository
2. copy .env.example to .env
3. adjust port, site URL, and secret key
4. start postgres + redis + init + app + queue + scheduler + reverb

The first goal is not perfect configuration. The first goal is to get a reachable frontend and admin.

Step 3: Sign in to the Admin

Admin path:

• /geo_admin/

Default credentials:

• username: admin
• password: password

Immediately after login:

• change the admin password
• confirm Laravel APP_KEY is generated
• read the first-deployment guide for security, queue, and backup checks

Step 4: Configure Models

Go to:

AI Configurator -> AI Model Settings

Start with:

• one stable, reasonably fast chat model
• one embedding model if you need RAG

GEOFlow supports OpenAI-compatible providers and native Gemini chat / embedding routes. Do not start with too many models; prove the workflow first.

Step 5: Configure Knowledge Chunking

If you use knowledge bases, confirm the chunking strategy on the AI Models page:

• use structured rule chunking by default
• enable LLM semantic planning for long or structurally complex documents
• semantic planning falls back to rule chunking when it fails

If the knowledge preview shows chunks but zero vectors, the embedding model is usually missing or not selected as the default embedding model.

Step 6: Prepare Minimum Materials

At minimum, prepare:

• one title library
• one knowledge base
• one body-generation prompt
• one author
• one category
• optionally one image library

If you do not yet have a real knowledge base, do not rush into large-scale task creation.

Step 7: Create the First Task

Recommended minimum task setup:

• title library: valid titles available
• model: one stable chat model
• prompt: body-generation prompt
• category: a clear content category
• generation count and publishing interval: start small
• review mode: start with draft / review
• publication scope: start with local site only

The first task is for workflow validation, not for scale.

Step 8: Verify the First Article

At minimum, confirm these things:

1. the task is queued correctly
2. the worker executes correctly
3. the article lands in draft
4. the review page shows the generated content
5. the frontend renders the published article correctly
6. Markdown, images, SEO, Open Graph, and structured data are correct

Once these points work, expansion becomes much safer.

Step 9: Configure Distribution If Needed

If content must sync to target sites, go to:

Distribution Management

Available channel types:

• GEOFlow Agent: best for new target sites. Download a target-site package; static mode is recommended by default.
• WordPress REST: best for existing WordPress sites. Use an Application Password to publish and manage posts.

Task publication scope can be:

• local site plus channels
• channels only
• local site only

Local-only mode disables channel selection and never creates remote distribution jobs.

Step 10: Review with Analytics

Go to:

Analytics

Start with:

• System Overview: articles, tasks, materials, and AI/API state
• Single-site Operations: publishing trends, content funnel, and category distribution
• Multi-site Distribution: channel status, failed jobs, and remote URLs
• Logs: visit trends, top articles, and AI crawlers

Optimize Last

Recommended order:

1. knowledge base
2. models and prompts
3. tasks and review
4. frontend themes and templates
5. multi-site distribution
6. CLI / Skill / API automation

Do not start with complex themes, heavy automation, large task volume, or many channels. First prove the real content workflow.

Recommended Next Reading

If your first article workflow is already working, continue with:

• What Is GEOFlow
• GEOFlow Methodology
• Knowledge Chunking and RAG
• Distribution Management and Target Sites
• Analytics and Logs