How is the text readability score computed?

The local engine uses Flesch-Kincaid algorithms, counting words, sentences, and syllables to calculate grade difficulty levels.

Will using this analyzer upload my content draft anywhere?

No. Your written content is analyzed completely in-browser, protecting unpublished manuscripts and campaign copies.

What is the purpose of OpenGraph tag previews?

OpenGraph previews show you how search engines and social apps present your page, helping you fix broken metadata before indexing.

Can I compose and format rich text here?

Yes. The editor provides real-time markdown parsing, displaying fully formatted HTML in split-screen panels locally.

Browser AI Lab · WebLLM MVP

Local Prompt Tester

Run a small local model in your browser before spending API money. No API key, no account, optional WebLLM local mode, streaming output and shareable result card.

Local mode uses WebGPU and may download model files to your browser cache. Your prompt is not sent to Bluesky Labs servers during local runs. Performance varies by device, browser, GPU, memory and model.

Model

Max output tokens

Where is the local model stored? (모델 저장 위치 / 삭제 방법)

Browser site storage cache

WebLLM model files are not saved to your normal Downloads folder. They are cached inside your browser’s site storage for this domain. Depending on the browser and WebLLM cache settings, the files may be stored through Cache Storage, IndexedDB, or OPFS. If you clear site data for bluesky-labs.com, the downloaded model cache may be removed and the model may need to be downloaded again next time. Model files are loaded into your device's RAM/VRAM during execution, which is separate from their local storage location.

WebLLM 모델 파일은 일반 다운로드 폴더에 저장되지 않습니다. bluesky-labs.com 사이트 전용 브라우저 저장공간에 캐시됩니다. 브라우저와 WebLLM 설정에 따라 Cache Storage, IndexedDB, OPFS 같은 저장소가 사용될 수 있습니다. bluesky-labs.com의 사이트 데이터를 삭제하면 모델 캐시도 삭제될 수 있으며, 다음 실행 시 다시 다운로드가 필요할 수 있습니다. 실행 중에는 기기의 RAM/VRAM에 임시 로드되며, 저장 장치 내 파일 보관소와 실행 메모리는 다르게 관리됩니다.

How to remove downloaded models (삭제 방법)

To remove downloaded model files, clear site data for bluesky-labs.com in your browser settings. In Chrome or Edge, you can also open DevTools → Application → Storage → Clear site data.

다운로드된 모델 파일을 지우려면 브라우저 설정에서 bluesky-labs.com 사이트 데이터를 삭제하세요. Chrome 또는 Edge에서는 개발자도구 → Application → Storage → Clear site data에서도 삭제할 수 있습니다.

⚠️ Caution: Clearing the cache does not delete anything from your normal file folders. It only removes browser-stored data for this site. After clearing, the model may need to download again.

System prompt

User prompt

Model availability depends on WebLLM's supported MLC model list and your browser/device memory. First run can take several minutes.

* WebLLM models are saved to browser site cache, not normal downloads. Where is the model stored?

Local runtime status

WebGPUNot checked

AdapterNot checked

WebLLM packageNot loaded

Model stateIdle

Download / init progress

Click “Check browser” first. The page will not download a model until you choose “Load local model”.

manual load onlyno server prompt upload during local run

How to clear downloaded model cache:

To remove browser-cached model files, clear site data for bluesky-labs.com in your browser settings, or open DevTools → Application → Storage → Clear site data. Normal file folders are unaffected.

(다운로드 캐시 삭제) 브라우저 설정에서 bluesky-labs.com 사이트 데이터를 삭제하거나, 크롬 개발자도구(F12) → Application → Storage → Clear site data를 진행하십시오. 로컬 파일 시스템에는 영향을 주지 않습니다.

Streaming output

Load a supported local model, then run your prompt.

load time --generation --tokens/sec --

Rule-based preflight

Use this quick checklist even before the local model loads. It catches vague prompts, missing output formats and risky guarantee language.

Open Prompt Optimizer

Share result card

Support local-first AI tools

If this browser AI lab is useful, you can support maintenance and more local-first tools. Support never controls scores, rankings or editorial recommendations.

Support via PayPal

Coming later

Side-by-side model comparison.

Local prompt history with IndexedDB.

Batch testing and export reports.

Downloaded model management guidance.

Launch resources

Browser Local LLM guide

WebLLM vs Ollama vs LM Studio

No-API Prompt Testing workflow

Safe wording

This tool does not claim GPT-4 quality, guaranteed privacy, or guaranteed prompt performance. It is a local test lab with device-dependent results.

Effective digital content relies on readability, engagement patterns, and metadata formatting. Readability estimators process written text by calculating syllable-to-word ratios, utilizing standard algorithms such as the Flesch-Kincaid index to determine text grade level difficulty. Similarly, social media previewers render OpenGraph and Twitter Meta tags inside simulated sandboxes, rendering exactly how links will look on platforms like Google, Slack, and Discord. For video hooks, optimization models grade content based on pacing, hook timing, and subtitle duration offsets. Because all textual analysis and social layouts are processed locally via browser APIs, your content drafts, SEO keywords, and campaign plans remain secure before publication.

Specifically, this Local Prompt Tester utility is designed to run a small local llm in your browser with webllm. test prompts with no api key, streaming output, webgpu checks, local metrics and shareable result cards. It runs with zero server dependencies, ensuring complete security and zero latency.