{"id":19958,"date":"2026-09-23T12:51:49","date_gmt":"2026-09-23T12:51:49","guid":{"rendered":"http:\/\/www.max-sperling.bplaced.net\/?p=19958"},"modified":"2026-09-24T12:10:10","modified_gmt":"2026-09-24T12:10:10","slug":"sw-architecture-around-an-llm-introduction","status":"publish","type":"post","link":"http:\/\/www.max-sperling.bplaced.net\/?p=19958","title":{"rendered":"SW-Architecture around LLMs (Introduction)"},"content":{"rendered":"<p><strong>Components<\/strong><\/p>\n<ul>\n<li>Frontend \/ UI<\/li>\n<li>Orchestrator <sup>1<\/sup><\/li>\n<li>Tools (e.g. code interpreter, image generator, web search)<\/li>\n<li>Backend \/ LLM<\/li>\n<\/ul>\n<p><u>Examples<\/u><\/p>\n<table>\n<tr>\n<th>Company<\/th>\n<th>Frontend<\/th>\n<th>Backend<\/th>\n<\/tr>\n<tr>\n<td>Google<\/td>\n<td>Bard <sup>2<\/sup><\/td>\n<td>LaMDA \/ PaLM 2 <sup>2<\/sup><\/td>\n<\/tr>\n<tr>\n<td>Meta<\/td>\n<td>Meta AI<\/td>\n<td>LLaMa<\/td>\n<\/tr>\n<tr>\n<td>OpenAI<\/td>\n<td>ChatGPT<\/td>\n<td>GPT<\/td>\n<\/tr>\n<\/table>\n<p><sup>1<\/sup> Routes the context to the backend and uses, if necessary, the right tool on its result.<br \/>\n<sup>2<\/sup> Those names are historic; these days they are combined under the name Gemini.<\/p>\n<hr>\n<p><strong>Workflow<\/strong><\/p>\n<pre>\r\nFrontend ---> Orchestrator ---> Backend ---> Orchestrator ---> Frontend\r\n                                                  |\r\n                                                Tools\r\n<\/pre>\n<p><u>Example<\/u> (for GitHub Copilot)<\/p>\n<ol>\n<li>Frontend: Gets the request &#8220;Change the output message to &#8216;Hello World!&#8217;.&#8221;<\/li>\n<li>Orchestrator: Provides the prompt plus additional context to a selected LLM. <sup>3<\/sup><\/li>\n<li>Backend: Generates an answer containing a tool call to edit a file. <sup>4<\/sup><\/li>\n<li>Orchestrator: Executes the tool call and modifies the answer accordingly.<\/li>\n<li>Frontend: Shows the modified answer as an output like &#8220;Modified main.cpp&#8221;.<\/li>\n<\/ol>\n<p><sup>3<\/sup> That contains system instructions, chat history, retrieved knowledge, and tool definitions.<br \/>\n<sup>4<\/sup> The LLM learns to handle general tool calling in post-training on specific training data.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Components Frontend \/ UI Orchestrator 1 Tools (e.g. code interpreter, image generator, web search) Backend \/ LLM Examples Company Frontend<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false},"categories":[30],"tags":[],"_links":{"self":[{"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/posts\/19958"}],"collection":[{"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=19958"}],"version-history":[{"count":23,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/posts\/19958\/revisions"}],"predecessor-version":[{"id":19983,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=\/wp\/v2\/posts\/19958\/revisions\/19983"}],"wp:attachment":[{"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=19958"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=19958"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.max-sperling.bplaced.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=19958"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}