{"id":7464,"date":"2026-08-19T19:10:06","date_gmt":"2026-08-19T19:10:06","guid":{"rendered":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/"},"modified":"2026-08-19T19:10:06","modified_gmt":"2026-08-19T19:10:06","slug":"conversations","status":"publish","type":"wvn_doc","link":"https:\/\/www.wiven.ai\/en\/wiven-llm\/docs\/utilisation\/conversations\/","title":{"rendered":"Conversations"},"content":{"rendered":"<h1>Conversations<\/h1>\n<p>This page describes the settings that determine how the assistant responds.,<br \/>\nand the concrete effect of each.<\/p>\n<hr \/>\n<h2>Conversation modes<\/h2>\n<p>The mode is selected in the space settings, tab <strong>Conversation<\/strong>.<\/p>\n<h3>Cat <em>(by default)<\/em><\/h3>\n<p>The template uses your documents <strong>And<\/strong> his general knowledge. If he doesn&#039;t<br \/>\nHe finds nothing relevant in the space, yet he still responds from this<br \/>\nthat he knows.<\/p>\n<p><strong>For :<\/strong> General assistance, writing, brainstorming, questions mixing<br \/>\ndocuments and general knowledge.<\/p>\n<h3>Request<\/h3>\n<p>The model responds <strong>exclusively<\/strong> from documents in space. In<br \/>\nIn the absence of a relevant passage, it returns a rejection message instead of<br \/>\nanswer.<\/p>\n<p><strong>For :<\/strong> Client files, regulatory framework, technical documentation \u2014<br \/>\nany use where a made-up answer would be worse than no answer.<\/p>\n<p>THE <strong>rejection message<\/strong> is customizable. By default, it indicates that none<br \/>\nNo relevant information was found; you can replace it with a<br \/>\nwording adapted to your users (\u00abThis information is not included in the<br \/>\nfile. Contact the manager.&quot;).<\/p>\n<h3>Automatic<\/h3>\n<p>Each message is entrusted to an agent, who alone decides whether or not to use...<br \/>\ntools.<\/p>\n<p><strong>Reserved for spaces dedicated to automation.<\/strong> This mode is slower and<br \/>\nmore unpredictable than the other two, and assumes a model capable of piloting<br \/>\ntools correctly. For occasional use by agents, the prefix is preferred.<br \/>\n<code>@agent<\/code> in a chat-mode space.<\/p>\n<hr \/>\n<h2>The prompt system<\/h2>\n<p>This is the application&#039;s most powerful lever: permanent instructions.<br \/>\ntransmitted to the model before each message.<\/p>\n<p>A good prompt system specifies:<\/p>\n<ul>\n<li><strong>the role<\/strong> \u2014 \u00abYou assist the employees of a Swiss trust company.\u00bb ;<\/li>\n<li><strong>the reference frame<\/strong> \u2014 what rules, what law, what method apply; ;<\/li>\n<li><strong>the expected behavior in the face of uncertainty<\/strong> \u2014 \u00abIf the information is not included<br \/>\n  &quot;Not in the documents, state it explicitly. Do not deduct any amounts.&quot; ;<\/li>\n<li><strong>the format<\/strong> \u2014 length, structure, response language; ;<\/li>\n<li><strong>the prohibitions<\/strong> \u2014 something the assistant should never do.<\/li>\n<\/ul>\n<p><strong>Example :<\/strong><\/p>\n<pre><code>You are assisting the staff of an accounting firm. You will only answer questions based on the documents in the client file. If the information is not there, state this clearly without making assumptions. Always cite the document and the date of the document used. Never calculate an amount that does not explicitly appear in a document. Answer in French, concisely and factually.\n<\/code><\/pre>\n<p>A <strong>default system prompt<\/strong> can be defined at the instance level<br \/>\n(<code>Settings \u2192 Default system prompt<\/code>): it applies to new spaces<br \/>\nwithout crushing those who have their own.<\/p>\n<p>The prompt system accepts <strong>variables<\/strong> \u2014 date, time, username,<br \/>\nthe name of the space, plus those you define.<br \/>\n\u2192 <a href=\"\/en\/wiven-llm\/docs\/utilisation\/commandes\/\">Commands and variables<\/a><\/p>\n<hr \/>\n<h2>Temperature<\/h2>\n<p>Controls the degree of unpredictability of the responses.<\/p>\n<table>\n<thead>\n<tr>\n<th>Value<\/th>\n<th>Behavior<\/th>\n<th>Use<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>0 \u2013 0,3<\/td>\n<td>Deterministic, factual, repeatable<\/td>\n<td>Documentary, legal, accounting<\/td>\n<\/tr>\n<tr>\n<td>0,4 \u2013 0,7<\/td>\n<td>Balance<\/td>\n<td>General use<\/td>\n<\/tr>\n<tr>\n<td>0.8 and above<\/td>\n<td>Varied, creative, less reliable<\/td>\n<td>Writing, ideation<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>For documentary use, a low value is almost always preferable:<br \/>\nAsking the same question twice should give the same answer.<\/p>\n<hr \/>\n<h2>History length<\/h2>\n<p>The number of previous messages returned to the model for each question.<br \/>\n<strong>Default value: 20.<\/strong><\/p>\n<ul>\n<li><strong>Higher<\/strong> \u2014 the assistant follows a long discussion better, but each<br \/>\n  The message costs more context, therefore more time, and, with a provider<br \/>\n  External, more money.<\/li>\n<li><strong>Lower<\/strong> \u2014 faster and cheaper responses, at the cost of follow-up<br \/>\n  Conversational level is lower.<\/li>\n<li><strong>Zero<\/strong> \u2014 Each message is processed individually.<\/li>\n<\/ul>\n<p>Helpful reminder: open a <strong>new thread<\/strong> is often preferable to reduce<br \/>\nThe history. \u2192 <a href=\"\/en\/wiven-llm\/docs\/utilisation\/espaces-de-travail\/\">Workspaces<\/a><\/p>\n<hr \/>\n<h2>Document search parameters<\/h2>\n<p>Space settings, tab <strong>Vector base<\/strong>.<\/p>\n<h3>Number of fragments found<\/h3>\n<p>How many passes are retrieved and provided to the model?. <strong>Default: 4.<\/strong><\/p>\n<ul>\n<li>Increasing coverage improves broad issues, at the cost of context<br \/>\n  consumed and potential noise.<\/li>\n<li>Diminish focuses the response on the most relevant passages.<\/li>\n<\/ul>\n<p>On long documents and cross-cutting questions, 6 to 10 often yield a<br \/>\nbest result. Beyond that, the context becomes saturated.<\/p>\n<h3>Similarity threshold<\/h3>\n<p>The minimum degree of proximity that a fragment must reach in order to be retained.<br \/>\n<strong>Default: 0.25.<\/strong><\/p>\n<table>\n<thead>\n<tr>\n<th>Threshold<\/th>\n<th>Effect<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Low (0 \u2013 0.2)<\/td>\n<td>Almost everything is coming back; noise is intruding on the context<\/td>\n<\/tr>\n<tr>\n<td>Average (0.25 \u2013 0.5)<\/td>\n<td>Usual balance<\/td>\n<\/tr>\n<tr>\n<td>High (0.6 and above)<\/td>\n<td>Only very close passages are being moved up; this risks excluding useful passages.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>Symptom and remedy:<\/strong> \u00ab&quot;He can&#039;t find it even though it&#039;s written there&quot; \u2192 lower it<br \/>\nthreshold. &quot;He quotes unrelated passages&quot; \u2192 mount it.<\/p>\n<h3>Search mode<\/h3>\n<p>Some vector databases allow you to combine semantic search with<br \/>\nA keyword search. Semantic search retrieves the wording.<br \/>\ndifferent; keyword search finds the exact references<br \/>\n(item numbers, product codes). Combining the two is useful on the<br \/>\ntechnical corpus.<\/p>\n<hr \/>\n<h2>What exactly the model sees<\/h2>\n<p>For each question, the assembled context contains, in this order:<\/p>\n<ol>\n<li>THE <strong>prompt system<\/strong> of space, resolved variables; ;<\/li>\n<li>THE <strong>pinned documents<\/strong>, in its entirety; ;<\/li>\n<li>your <strong>memoirs<\/strong> relevant and, where applicable, the elements of<br \/>\n   <strong>Second Brain<\/strong> ;<\/li>\n<li>THE <strong>fragments found<\/strong> through research; ;<\/li>\n<li>THE <strong>latest messages<\/strong> thread; ;<\/li>\n<li>your <strong>question<\/strong>.<\/li>\n<\/ol>\n<p>All of this must fit within the model&#039;s context window. When it is<br \/>\nWhen the system is saturated, the oldest elements are discarded \u2014 hence the importance of not<br \/>\ndon&#039;t pin everything down and don&#039;t let a thread drag on forever.<\/p>\n<hr \/>\n<h2>Sources and verification<\/h2>\n<p>Each answer built on your documents displays the extracts used. They<br \/>\nallow to:<\/p>\n<ul>\n<li><strong>check<\/strong> that the answer depends on the correct passage; ;<\/li>\n<li><strong>diagnose<\/strong> a search problem \u2014 off-topic excerpts<br \/>\n  indicate a threshold or a division that needs to be reviewed; ;<\/li>\n<li><strong>go back to the source<\/strong> for complete control.<\/li>\n<\/ul>\n<p>An answer without a displayed source was not built on your documents: either<br \/>\nThe search yielded nothing, or the model responded based on its knowledge.<br \/>\ngeneral (Chat mode).<\/p>\n<hr \/>\n<h2>Adjusting a space: three profiles<\/h2>\n<table>\n<thead>\n<tr>\n<th>Profile<\/th>\n<th>Fashion<\/th>\n<th>Temperature<\/th>\n<th>Fragments<\/th>\n<th>Threshold<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>Strict documentary<\/strong> \u2014 files, regulatory<\/td>\n<td>Request<\/td>\n<td>0 \u2013 0,2<\/td>\n<td>6 \u2013 10<\/td>\n<td>0,25 \u2013 0,4<\/td>\n<\/tr>\n<tr>\n<td><strong>General Assistance<\/strong><\/td>\n<td>Cat<\/td>\n<td>0,5 \u2013 0,7<\/td>\n<td>4<\/td>\n<td>0,25<\/td>\n<\/tr>\n<tr>\n<td><strong>Exploratory research<\/strong> on large corpus<\/td>\n<td>Cat<\/td>\n<td>0,3<\/td>\n<td>8 \u2013 12<\/td>\n<td>0,15 \u2013 0,25<\/td>\n<\/tr>\n<\/tbody>\n<\/table>","protected":false},"parent":7450,"menu_order":303,"template":"","meta":[],"class_list":["post-7464","wvn_doc","type-wvn_doc","status-publish","hentry"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.6 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Conversations \u2014 Documentation WivenLLM<\/title>\n<meta name=\"description\" content=\"Modes Chat et Requ\u00eate, prompt syst\u00e8me, temp\u00e9rature, historique et param\u00e8tres de recherche \u2014 ce que chaque r\u00e9glage change vraiment.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wiven.ai\/en\/wiven-llm\/docs\/utilisation\/conversations\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Conversations \u2014 Documentation WivenLLM\" \/>\n<meta property=\"og:description\" content=\"Modes Chat et Requ\u00eate, prompt syst\u00e8me, temp\u00e9rature, historique et param\u00e8tres de recherche \u2014 ce que chaque r\u00e9glage change vraiment.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wiven.ai\/en\/wiven-llm\/docs\/utilisation\/conversations\/\" \/>\n<meta property=\"og:site_name\" content=\"Wiven AI\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/conversations\\\/\",\"url\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/conversations\\\/\",\"name\":\"Conversations \u2014 Documentation WivenLLM\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.wiven.ai\\\/#website\"},\"datePublished\":\"2026-08-19T19:10:06+00:00\",\"description\":\"Modes Chat et Requ\u00eate, prompt syst\u00e8me, temp\u00e9rature, historique et param\u00e8tres de recherche \u2014 ce que chaque r\u00e9glage change vraiment.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/conversations\\\/#breadcrumb\"},\"inLanguage\":\"en\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/conversations\\\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/conversations\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.wiven.ai\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Documentation\",\"item\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Utilisation\",\"item\":\"https:\\\/\\\/www.wiven.ai\\\/wiven-llm\\\/docs\\\/utilisation\\\/\"},{\"@type\":\"ListItem\",\"position\":4,\"name\":\"Conversations\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.wiven.ai\\\/#website\",\"url\":\"https:\\\/\\\/www.wiven.ai\\\/\",\"name\":\"Wiven AI\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.wiven.ai\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Conversations \u2014 WivenLLM Documentation","description":"Chat and Query modes, system prompt, temperature, history and search settings \u2014 what each setting really changes.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wiven.ai\/en\/wiven-llm\/docs\/utilisation\/conversations\/","og_locale":"en_US","og_type":"article","og_title":"Conversations \u2014 Documentation WivenLLM","og_description":"Modes Chat et Requ\u00eate, prompt syst\u00e8me, temp\u00e9rature, historique et param\u00e8tres de recherche \u2014 ce que chaque r\u00e9glage change vraiment.","og_url":"https:\/\/www.wiven.ai\/en\/wiven-llm\/docs\/utilisation\/conversations\/","og_site_name":"Wiven AI","twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/","url":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/","name":"Conversations \u2014 WivenLLM Documentation","isPartOf":{"@id":"https:\/\/www.wiven.ai\/#website"},"datePublished":"2026-08-19T19:10:06+00:00","description":"Chat and Query modes, system prompt, temperature, history and search settings \u2014 what each setting really changes.","breadcrumb":{"@id":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/#breadcrumb"},"inLanguage":"en","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/conversations\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.wiven.ai\/"},{"@type":"ListItem","position":2,"name":"Documentation","item":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/"},{"@type":"ListItem","position":3,"name":"Utilisation","item":"https:\/\/www.wiven.ai\/wiven-llm\/docs\/utilisation\/"},{"@type":"ListItem","position":4,"name":"Conversations"}]},{"@type":"WebSite","@id":"https:\/\/www.wiven.ai\/#website","url":"https:\/\/www.wiven.ai\/","name":"Wiven AI","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wiven.ai\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en"}]}},"_links":{"self":[{"href":"https:\/\/www.wiven.ai\/en\/wp-json\/wp\/v2\/wvn_doc\/7464","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wiven.ai\/en\/wp-json\/wp\/v2\/wvn_doc"}],"about":[{"href":"https:\/\/www.wiven.ai\/en\/wp-json\/wp\/v2\/types\/wvn_doc"}],"up":[{"embeddable":true,"href":"https:\/\/www.wiven.ai\/en\/wp-json\/wp\/v2\/wvn_doc\/7450"}],"wp:attachment":[{"href":"https:\/\/www.wiven.ai\/en\/wp-json\/wp\/v2\/media?parent=7464"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}