Skip to content

NiubiGEO v0.2.0

Latest
Latest

Choose a tag to compare

View all tags
@Albert-Weasker Albert-Weasker released this 08 Sep 10:01
· 2 commits to main since this release
v0.2.0
a4e778e

NiubiGEO v0.2.0

NiubiGEO is an open-source, self-hosted tool for observing how AI models describe a product and which competing products appear in their answers.

What Is Included

  • Independent domain projects, saved model selections and versioned monitoring configurations.
  • Per-model domain recognition, raw answers, competitor descriptions, brand keywords and competitor keywords.
  • Separate offline and Provider-native search settings. Provider citations stay separate from ordinary answer URLs.
  • Cross-model evidence views, keyword measurements, comparable repeat runs and a separate scheduling worker.
  • Synchronized English and Chinese READMEs, new logos and 20 public real-domain cases with original answers and screenshots.

English guide · 简体中文 · 20 real cases

Install

docker pull ghcr.io/albert-weasker/niubigeo:v0.2.0
docker run -d --name niubigeo \
  -p 127.0.0.1:8787:8787 \
  -e OPENROUTER_API_KEY \
  -v niubigeo-data:/app/data/product-v2 \
  ghcr.io/albert-weasker/niubigeo:v0.2.0

Set OPENROUTER_API_KEY in your shell before running the container. Open http://localhost:8787. Model inference and native search use your API balance. The image supports Linux amd64 and arm64. Scheduling requires the separate worker.

git clone --branch v0.2.0 --depth 1 https://github.com/Albert-Weasker/niubigeo.git
cd niubigeo
npm ci
cp .env.example .env
npm run server

Enter your key in .env before starting from source. Node.js 22 or later is required. This release distributes a Docker image, not an npm package.

Real-Case Coverage

The retained study covers 20 real domains, each with at least one analyzable domain answer. Ten cases completed and ten were partial; eleven cases also ran keyword measurements. Failures and uncertain results remain visible in the cases. This is not a claim that every model or every observation succeeded.

The study made 204 paid calls, including preflight, and recorded USD 1.00476325 in Provider costs. Publishing this release does not rerun those calls. Repeated runs taken seconds apart demonstrate repeatability workflows, not long-term growth.

Known Issues

  • Some model outputs claim more than one unique first position. The raw answers are retained unchanged; the current aggregate can still count these conflicts as adjudicable zero values. Do not interpret those aggregate values as proof of rank. See the affected cases in known issues.
  • A real all-models-failed export scenario has not completed full acceptance. Synthetic tests are not presented as real Provider evidence.
  • The product workbench still has Chinese-only interface areas. Bilingual documentation does not mean every product screen or error path has English acceptance.
  • The workbench is intended for trusted, local/self-hosted use. Do not expose it publicly without access controls. Back up data before upgrading.

The maintainer authorized direct publication of v0.2.0 with these disclosures. Earlier blocked candidate acceptance records have not been rewritten as passed. See limitations, backup and upgrade, and the release's image-verification attachments for the tested artifact.

中文说明

本版包含独立项目、多模型域名认知、原始证据、逐模型联网配置、关键词测量、复测与独立定时监测 Worker，以及同步的中英文 README、20 个真实案例和截图。

20 个案例均取得至少一条可分析的域名回答，但只有 10 例完整完成，另 10 例部分完成；11 例执行了关键词测试。模型的原话、分析失败和不确定结果均保留，不将其改写为成功。

上述排名冲突统计、真实全失败流程验收及产品英文覆盖问题仍存在。本次按维护者要求直接发布正式版本，不代表此前所有验收门禁已经通过。请先备份，并仅在受信任环境中部署。

Licensed under Apache-2.0. NiubiStar sponsors development; inclusion of other products in the cases does not imply affiliation or endorsement.

Published Image Verification

Source commit: a4e778e8ca9e3ffc691013a91cda950a8e98fa62.

Image index: sha256:5e7a01b07d246201114a4e0a769183184ef88f636ca609f3d0a9753012934588. v0.2.0, latest, and the source-SHA tag resolve to this same digest.

Both Linux platforms passed 36 container checks in total. The publicly pulled arm64 image passed 9 browser checks, including creation, refresh, persistence, deletion and mobile layout. This does not replace the disclosed product acceptance gaps.

Machine-readable verification and test counts. Container records and real browser screenshots are attached to this Release. The original workflow cache failure and the recovery using the unchanged image are recorded separately.

Assets 11
Loading