Review-compliance release: readable sources now ship inside the archive.\nNo behaviour change.<\/p>","0.3.12":"
Plugin Check clean: no warnings. No behaviour change.<\/p>","0.3.11":"
The GPU model library now ships with the plugin instead of being downloaded.<\/p>","0.3.10":"
Directory-compliance release: readme, GPL-2.0-or-later, filesystem helpers.\nNo behaviour change.<\/p>"},"ratings":[],"assets_icons":{"icon-128x128.png":{"filename":"icon-128x128.png","revision":3793498,"resolution":"128x128","location":"assets","locale":"","width":128,"height":128},"icon-256x256.png":{"filename":"icon-256x256.png","revision":3793498,"resolution":"256x256","location":"assets","locale":"","width":256,"height":256}},"assets_banners":{"banner-1544x500.png":{"filename":"banner-1544x500.png","revision":3793498,"resolution":"1544x500","location":"assets","locale":"","width":1544,"height":500},"banner-772x250.png":{"filename":"banner-772x250.png","revision":3793498,"resolution":"772x250","location":"assets","locale":"","width":772,"height":250}},"assets_blueprints":{},"all_blocks":[],"tagged_versions":["0.3.15"],"block_files":[],"assets_screenshots":[],"screenshots":[]},"plugin_section":[],"plugin_tags":[1320,2364,2323,396,944],"plugin_category":[54,59],"plugin_contributors":[291734],"plugin_business_model":[],"class_list":["post-398499","plugin","type-plugin","status-publish","hentry","plugin_tags-assistant","plugin_tags-chatbot","plugin_tags-offline","plugin_tags-privacy","plugin_tags-search","plugin_category-security-and-spam-protection","plugin_category-utilities-and-tools","plugin_contributors-00greyok00","plugin_committers-00greyok00"],"banners":{"banner":"https:\/\/ps.w.org\/grey-local-site-ai-assistant\/assets\/banner-772x250.png?rev=3793498","banner_2x":"https:\/\/ps.w.org\/grey-local-site-ai-assistant\/assets\/banner-1544x500.png?rev=3793498","banner_rtl":false,"banner_2x_rtl":false},"icons":{"svg":false,"icon":"https:\/\/ps.w.org\/grey-local-site-ai-assistant\/assets\/icon-128x128.png?rev=3793498","icon_2x":"https:\/\/ps.w.org\/grey-local-site-ai-assistant\/assets\/icon-256x256.png?rev=3793498","generated":false},"screenshots":[],"raw_content":"\n
Local Assistant adds a small chat panel to the site. The visitor asks a\nquestion, the plugin finds the most relevant passages in the site's own pages,\nand a small language model that runs in the visitor's browser writes the answer.\nThe answer cites the page it came from. If the pages do not answer the question,\nthe assistant says so instead of inventing one.<\/p>\n\n
The site owner builds the search index once from the dashboard: the plugin walks\nthe sitemap, extracts the readable text of each page, and saves a searchable\nindex plus its vectors into The plugin itself calls no external service, and the site server never sends\nvisitor data anywhere. On the visitor's device, the browser downloads model\nfiles the first time the assistant is used, after the visitor accepts the\nconsent card:<\/p>\n\n Those downloads are the only network traffic the assistant causes, they are\nstarted only after consent, and they go directly from the visitor's browser to\nHugging Face under Hugging Face's own terms. The site owner can point the\ndownloads at another host with the The assistant runs on the author's own site \u2014 ask it something and watch it\ncite the page it read:<\/p>\n\n The WordPress package is generated from readable sources, and the full\nbuild is public:<\/p>\n\n The two No. Retrieval and generation both run in the visitor's browser. The question is\nnot transmitted anywhere, and there is no analytics or telemetry call.<\/p><\/dd>\n The first use downloads the model files (roughly 100-400 MB depending on the\nengine the browser supports) and caches them. Later visits start from the cache.<\/p><\/dd>\n No. WebGPU makes answers faster where it is available; otherwise the plugin\nloads a small CPU model, and if that cannot run it shows the closest matching\npage instead.<\/p><\/dd>\n Press \"Ingest site\" again. The index is rebuilt from the sitemap; there is no\nschedule and no background job.<\/p><\/dd>\n It is written to wp-content\/uploads\/local-assistant\/<\/code>. Nothing is\nsent to a server at any point.<\/p>\n\n\n
External services<\/h4>\n\n
\n
HuggingFaceTB\/SmolLM2-135M-Instruct<\/code>\nfrom Hugging Face.<\/li>\ndist\/vendor\/web-llm\/SmolLM2-360M-Instruct-q4f16_1_cs1k-webgpu.wasm<\/code>) and is\nloaded from the plugin folder, so no executable code is fetched from a third\nparty at runtime.<\/li>\n<\/ul>\n\nmodelBaseUrl<\/code>, modelLibUrl<\/code>, and\n modelMirror settings.<\/p>\n\nSee it live<\/h4>\n\n
\n
Source Code<\/h3>\n\n
\n
src\/<\/code> holds the plugin's JavaScript modules, scripts\/<\/code> the build tools.<\/li>\nsource\/<\/code> folder.<\/li>\nnpm ci && npm run vendor && npm run\nbuild<\/code> (Node.js 20+), then npm run release<\/code> for versioned zips.<\/li>\ndist\/vendor\/<\/code> are Transformers.js\n(Apache-2.0, sources<\/a>) and\nWebLLM (Apache-2.0, sources<\/a>); exact\nversions are listed in the included THIRD-PARTY-NOTICES.txt<\/code>.<\/li>\n<\/ul>\n\n.wasm<\/code> files flagged in review are required, and both are loaded\nonly from the plugin's own folder \u2014 the plugin never fetches executable\ncode from a third-party host:<\/p>\n\n\n
dist\/vendor\/transformers\/ort-wasm-simd-threaded.jsep.wasm<\/code> is the ONNX\nRuntime Web engine that runs the embedding model and the CPU fallback\nmodel inside Transformers.js; the plugin cannot function without it.<\/li>\ndist\/vendor\/web-llm\/SmolLM2-360M-Instruct-q4f16_1_cs1k-webgpu.wasm<\/code> is\nWebLLM's compiled GPU program for the SmolLM2-360M-Instruct-q4f16_1<\/code>\nchat model; it is the WebGPU conversation engine itself. The fetched copy\nis tracked in the repository under scripts\/vendor-extra\/<\/code>.<\/li>\n<\/ul>\n\n\n\n
\/wp-content\/plugins\/<\/code>, or install the zip from\nPlugins > Add New > Upload Plugin. The zip already contains the built\n dist\/ assets; no build step is needed.<\/li>\nwp-content\/uploads\/local-assistant\/<\/code>.<\/li>\n\n
Does this send my visitors' questions to a server?<\/h3><\/dt>\n
Why does the first answer take longer?<\/h3><\/dt>\n
Do I need a GPU?<\/h3><\/dt>\n
What happens if a page changes?<\/h3><\/dt>\n
Is the index public?<\/h3><\/dt>\n
wp-content\/uploads\/local-assistant\/<\/code>, which is normally\nreadable by visitors \u2014 that is what makes offline answers possible. It contains\ntext already published on the site, not private content.<\/p><\/dd>\n\n<\/dl>\n\n\n0.3.15<\/h4>\n\n
\n
source\/<\/code> folder, and\nthe repository now carries the sources and\nbuild scripts (src\/<\/code>, scripts\/<\/code>, package files) so the package can be\nrebuilt from source. The readme documents the source location and the\nthird-party library origins. No runtime behaviour changed.<\/li>\n<\/ul>\n\n0.3.12<\/h4>\n\n
\n
0.3.11<\/h4>\n\n
\n
0.3.10<\/h4>\n\n
\n
0.3.9<\/h4>\n\n
\n
0.3.8<\/h4>\n\n
\n