CLI tool
yaojingang/GEOFlow avatar
yaojingang/GEOFlow

GEOFlow's source is ahead of its newest tag

Open-source GEO content engineering and multi-site distribution system with AI tasks, RAG/semantic chunking, analytics, GEOFlow Agent and WordPress target publishing.

4,139 stars946 forksPHPApache-2.0

At a glance

What is it?
A GEO content engineering system on Laravel and PostgreSQL, where the working tree is 3.2.0-beta.2 while the newest tag is beta.1 and the stable line is 3.1.0, the example environment file hardcodes a 200 daily model call cap and three lease timeouts, and the upgrade guide explicitly forbids the git pull you would reach for first.
Who is it for?
GEOFlow is worth evaluating if your team already has reviewed source material and someone accountable for approving what ships, because the quality check is a publish barrier rather than a report and a person confirms the final publish. Two things to decide first.
Can I use it commercially?
Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
Is it still maintained?
Yes. The repository last received commits 2 days ago.
What is it written in?
Mainly PHP, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • yaojingang/GEOFlow4,139 ★ · +1,049
Star history
Repository2026-08-022026-11-15
yaojingang/GEOFlow3.1k4.1k

1,049 Stars in the last 105 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

The source is beta.2, the newest tag beta.1, the stable line 3.1.0

The version note under the badges says three different things at once. The current source is a `3.2.0-beta.2` preview, the newest stable release is `3.1.0`, and the exact source version is to be read from a `version.json` file rather than from the tag. The published tags do not agree with the working tree either: the newest tag is `v3.2.0-beta.1`, so the source is one beta ahead of anything downloadable. The page's own advice follows from that, recommending a formal Release for production or pinning to a commit that has been reviewed. The components table then shows why this matters, because the parts are versioned on separate streams. GEOFlow Core is the `3.2.0-beta.2` Laravel application. The GEOFlow CLI is `0.4.0-preview.2`. The Chrome operations assistant is `0.1.0`. The updater is a separate component in a different repository.

Three releases in thirteen days

The tag history is compressed. v3.0.0 was published on 2026-09-04, v3.1.0 on 2026-09-09 and v3.2.0-beta.1 on 2026-09-16, so a major, a minor and a beta in two weeks. That cadence explains the awkward position above, and it explains one more line in the components table, where the CLI is marked `0.4.0-preview.2` with the note that remote drafts are available but theme publishing is not yet open. So the newest part of the system is also the part with the smallest version number and an explicitly unfinished feature. Nothing on the page dates the last commit to the default branch beyond the tag list, which is dated 2026-10-01.

The example environment file carries the operating limits

The most useful document in the repository is the example environment file, because the policy is in it rather than in the prose. The AI workspace is off by default with `GEOFLOW_AI_WORKSPACE_RUNTIME_ENABLED=false` and a separate force-disable switch alongside it, and hosted channel sites are off too. When it is on, the numbers are hardcoded.

code
GEOFLOW_AI_WORKSPACE_ADMIN_DAILY_MODEL_CALLS=200
GEOFLOW_AI_WORKSPACE_GLOBAL_CONCURRENCY=10
GEOFLOW_AI_WORKSPACE_HISTORY_CHAR_BUDGET=16000
GEOFLOW_AI_WORKSPACE_RETENTION_DAYS=90

Three more settings are lease timeouts for generation, resolution and execution, at 180 seconds, 3 minutes and 15 minutes respectively, and `GEOFLOW_AI_WORKSPACE_REQUIRE_VERIFIED_MODEL=true` sits at the end. So an administrator gets two hundred model calls a day, retained history is capped by characters rather than turns, and the leases double as the crash recovery window.

An empty APP_KEY is the documented state and injecting it breaks things

The environment example ships `APP_KEY=` empty, with a comment saying the Docker entrypoint runs key generation and warning against injecting an empty APP_KEY environment variable into the container, because then a valid key cannot be written. That is a real deployment trap documented in the place someone will hit it. Two lines below, the proxy trust boundary is set the same careful way. `TRUSTED_PROXIES` is empty, and the comment says the init and app containers fall back to REMOTE_ADDR in that case, trusting only the internal Nginx caller, and that anyone running Artisan services directly or building their own proxy chain should either leave it empty or write explicit proxy addresses. A third line handles forwarded paths: `GEOFLOW_FORWARDED_PREFIX` clears any client-supplied X-Forwarded-Prefix when left empty.

The development compose file publishes a fallback database password

The development compose file gives every credential a default.

code
POSTGRES_DB: ${DB_DATABASE:-geo_flow}
POSTGRES_USER: ${DB_USERNAME:-geo_user}
POSTGRES_PASSWORD: ${DB_PASSWORD:-geo_password}

So a developer who does not set anything gets a database called geo_flow owned by geo_user with the password geo_password, written into a public repository. The mitigation is in the same file and is easy to miss: the port mapping binds to loopback rather than all interfaces, `"127.0.0.1:${DB_EXPOSE_PORT:-15432}:5432"`, and Redis is bound the same way on 16379. The Postgres service also has a healthcheck running `pg_isready` against the same defaults, so the check passes precisely when those defaults are in effect. The production path uses a separate file and a separate example.

Production is nine services and the docs forbid git pull

The production compose file is driven by an explicit env file, repeated on every command.

bash
docker compose --env-file .env.prod -f docker-compose.prod.yml build
docker compose --env-file .env.prod -f docker-compose.prod.yml up -d init
docker compose --env-file .env.prod -f docker-compose.prod.yml up -d --remove-orphans app web queue ai-quality-queue ai-quality-backfill-queue ai-optimization-queue knowledge-queue scheduler reverb

That last line names nine services in one container group, alongside postgres and redis started separately, and the repository carries four compose files at the root for development, production, a prebuilt variant and a UI variant. The upgrade section is the part to read twice, because it names the obvious mistake: back up the database, the env file, the upload directory and storage, stop the service and drain the old processes, then migrate and rebuild. The page says explicitly to avoid running `git pull` and rebuilding containers, and routes you to a shutdown and security migration protocol instead.

The quality check is a publish gate, not a report

The capability table says articles that are pending review, blocked, anomalous or expired stay in the draft stage. That single clause turns the AI check from a scorecard into a barrier, and the audit trail behind it is unusually complete: it records per-item scores, positions in the source text, the regulatory basis, revision suggestions and previous results, and it keeps records when a manual override happens or a policy changes. The same section publishes four generated addresses per site, `robots.txt`, `sitemap.xml`, `sitemap.txt` and `llms.txt`, produced in real time from the current site, publish state, index switches, canonical article links and site settings. The page is candid about the limit of the first one: those rules only manage crawler intent, and sensitive data still has to be protected with authentication, authorisation and server access control.

The extensions deliberately hold no credentials

The security table's browser collaboration row says the Chrome extension uses device pairing and a least-privilege token, and does not store external platform passwords, cookies or OAuth credentials. The feature description adds the shape of the workflow: the assistant picks up manual publish tickets, fills pending drafts, returns execution receipts, and the final publish is confirmed by an operator. The updater row is the strongest control on the page, with signed packages, a local Unix socket, environment acceptance checks, a full backup and restore points, and high-risk requests requiring the administrator password plus a 6-digit authenticator code. Telemetry is off by default and, when enabled, sends a fixed whitelist of fields with business content, accounts, emails, domains, cookies and keys excluded from the payload.

Editorial conclusion

GEOFlow is worth evaluating if your team already has reviewed source material and someone accountable for approving what ships, because the quality check is a publish barrier rather than a report and a person confirms the final publish. Two things to decide first. The source tree is a beta ahead of the newest tag while the stable line is a minor behind, so pin to a reviewed commit rather than tracking the branch. And the limits that shape day to day use, a daily model call cap for the admin, a character budget on retained history, and lease timeouts that double as crash recovery, live in an example environment file rather than in the documentation, so read that file before you size the deployment.

Frequently asked questions

What is GEOFlow used for?

GEOFlow is an open-source GEO content engineering and multi-site distribution system. It chains trusted knowledge, AI content production, quality gating, human review, multi-site distribution and analytics into one operating loop, aimed at teams running a corporate site, a GEO sub-channel, an industry source site or an internal content platform.

What are the disadvantages of subsurface irrigation?

The repository has nothing to do with irrigation. GEOFlow is a content engineering system written primarily in PHP on Laravel, licensed AGPL-3.0-only, with a public front end, an admin UI, an API v1, a CLI and a browser extension.

What is a subsurface drip septic system?

Not this repository's subject. GEOFlow publishes to a WordPress REST target, per-channel hosted site packages, a GEOFlow Agent target and a generic HTTP API, and its queue services include ai-quality, ai-optimization and knowledge queues alongside the main one.

What does GEOFlow require to run?

PHP 8.3 or newer, with PHP 8.4 in the Docker default. PostgreSQL, with a pgvector image or a compatible extension recommended. Redis for queues, cache and run state. Node.js for the front-end build, with Node 22 in CI. Docker Compose, and Nginx with php-fpm in production.

How do I upgrade GEOFlow from 2.x?

Back up the database, the env file, the upload directory and storage, then stop the service and drain the old processes before migrating, rebuilding the front end and restarting. The page says explicitly to avoid running git pull and rebuilding containers, and points to a shutdown and security migration protocol document plus a separate 3.0 upgrade tutorial.

Which languages does the GEOFlow admin UI support?

Six. The readme switcher lists Simplified Chinese, English, Japanese, Spanish, Russian and Brazilian Portuguese, and the capabilities table states that the Admin UI V3 supports six languages with a responsive layout, a PWA and illustrated help.

Official sources

  1. Official README
  2. Project repository
  3. Release notes
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/yaojingang-geoflow.svg)](https://hysenlabs.com/projects/yaojingang-geoflow)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/yaojingang-geoflow.svg)](https://hysenlabs.com/projects/yaojingang-geoflow)

Was this useful?