@xuda.io/ai_module 1.1.5425 → 1.1.5427

package/index.mjs

@@ -4212,6 +4212,16 @@ const ai_chat_conversation = async function (req, job_id, headers) {
 
     const render_output_suggestions = async function (response_conversation_item_id) {
       console.log('prompt_suggestions', prompt_suggestions);
+      let in_conversation_item_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, response_conversation_item_id);
+      const response_text = (in_conversation_item_doc?.text || in_conversation_item_doc?.prompt || '').toLowerCase().trim();
+
+      if (!response_text || /how can i help you today\??/.test(response_text) || (/^(hi|hello|hey)\b/.test(response_text) && response_text.length < 140)) {
+        in_conversation_item_doc.output_suggestions = [];
+        await db_module.save_app_couch_doc_native(account_profile_info.app_id, in_conversation_item_doc);
+        emitToDashboard('suggestions', JSON.stringify([]));
+        return { code: 44, data: [] };
+      }
+
       const exclude_agents = prompt_suggestions.map((agent) => agent.agent_id);
       const ret = await get_chat_suggestions({ uid, type: 'chat response', ai_agents, conversation_item_id: response_conversation_item_id, exclude_agents });
 
@@ -4220,7 +4230,6 @@ const ai_chat_conversation = async function (req, job_id, headers) {
         emitToDashboard('suggestions', JSON.stringify(ret.data));
         // emitToDashboard('stream_end');
 
-        let in_conversation_item_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, response_conversation_item_id);
         in_conversation_item_doc.output_suggestions = ret.data || [];
         const save_ret = await db_module.save_app_couch_doc_native(account_profile_info.app_id, in_conversation_item_doc);
       }
@@ -11186,14 +11195,44 @@ export const get_chat_suggestions = async function (req) {
 
   try {
     const account_profile_info = await get_active_account_profile_info(uid);
-    const account_doc = await db_module.get_couch_doc_native('xuda_accounts', uid);
     const conversation_item_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, conversation_item_id);
     const conversation_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, conversation_item_doc.conversation_id);
+    const contentToAnalyze = (context || conversation_item_doc?.text || conversation_item_doc?.prompt || conversation_doc?.prompt || '').trim();
 
     if (conversation_item_doc.prompt_suggestions) {
       exclude_agents = [...new Set([...exclude_agents, ...conversation_item_doc.prompt_suggestions.map((agent) => agent.agent_id)])];
     }
 
+    const is_generic_content = function (content) {
+      if (!content) {
+        return true;
+      }
+
+      const normalized = content.toLowerCase().replace(/\s+/g, ' ').trim();
+
+      if (normalized.length < 24 && !/[{}[\]<>`]/.test(normalized)) {
+        return true;
+      }
+
+      if (/^(hi|hello|hey|good morning|good afternoon|good evening)\b/.test(normalized) && normalized.length < 140) {
+        return true;
+      }
+
+      if (/how can i help you today\??/.test(normalized)) {
+        return true;
+      }
+
+      if (/^(thanks|thank you|ok|okay|sure|sounds good|great|nice|welcome)\b/.test(normalized) && normalized.length < 80) {
+        return true;
+      }
+
+      return false;
+    };
+
+    if (is_generic_content(contentToAnalyze)) {
+      return { code: 44, data: [] };
+    }
+
     const get_agents = async function () {
       if (typeof ai_agents === 'undefined') {
         // no agents
@@ -11208,54 +11247,47 @@ export const get_chat_suggestions = async function (req) {
         }
       }
 
-      const list = [];
-
-      for (const agent_id of ai_agents) {
-        try {
-          let ai_agent_doc;
+      const agent_docs = await Promise.all(
+        ai_agents.map(async (agent_id) => {
           try {
-            ai_agent_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, agent_id);
+            if (agent_id === conversation_item_doc.ai_agent_id) return null;
+            if (exclude_agents.includes(agent_id)) return null;
+
+            let ai_agent_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, agent_id);
             if (ai_agent_doc.studio_meta.shared_from_uid) {
               const reference_doc = _.cloneDeep(ai_agent_doc);
-              const app_id = await get_account_default_project_id(ai_agent_doc.studio_meta.shared_from_uid);
+              await get_account_default_project_id(ai_agent_doc.studio_meta.shared_from_uid);
               ai_agent_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, ai_agent_doc.studio_meta.share_item_id);
               ai_agent_doc.reference_doc = reference_doc;
             } else if (ai_agent_doc.studio_meta.installed_from_app_uid) {
               const reference_doc = _.cloneDeep(ai_agent_doc);
-              const app_id = await get_account_default_project_id(ai_agent_doc.studio_meta.installed_from_app_uid);
+              await get_account_default_project_id(ai_agent_doc.studio_meta.installed_from_app_uid);
               ai_agent_doc = await db_module.get_app_couch_doc_native(account_profile_info.app_id, ai_agent_doc.studio_meta.share_item_id);
               ai_agent_doc.reference_doc = reference_doc;
             }
 
-            if (agent_id === conversation_item_doc.ai_agent_id) continue; //exclude the current agent from list
-            if (exclude_agents.includes(agent_id)) continue;
-
-            let skip = false;
-            for (const [key, val] of Object.entries(ai_agent_doc?.agentConfig?.agent_tools || [])) {
+            for (const [, val] of Object.entries(ai_agent_doc?.agentConfig?.agent_tools || [])) {
               if (['web_search', 'table', 'file_search'].includes(val.type)) {
-                skip = true;
-                break;
+                return null;
               }
             }
-            if (skip) continue;
 
-            list.push({
+            return {
               agent_id,
               agent_name: ai_agent_doc.properties.menuName,
               agent_category: ai_agent_doc.studio_meta.agent_category,
               agent_industry: ai_agent_doc.studio_meta.agent_industry,
-              agent_instructions: ai_agent_doc?.agentConfig?.agent_instructions,
-              agent_user_guide: ai_agent_doc?.agentConfig?.agent_user_guide,
-            });
+              agent_instructions: ai_agent_doc?.agentConfig?.agent_instructions?.slice(0, 300),
+              agent_user_guide: ai_agent_doc?.agentConfig?.agent_user_guide?.slice(0, 200),
+            };
           } catch (err) {
-            throw new Error(err.message);
+            console.error(err.message);
+            return null;
           }
-        } catch (err) {
-          console.error(err.message);
-        }
-      }
+        }),
+      );
 
-      return list;
+      return agent_docs.filter(Boolean);
     };
 
     const get_suggestions = async function () {
@@ -11281,22 +11313,24 @@ export const get_chat_suggestions = async function (req) {
         1. Return fewer than ${no_of_suggestions} agents when only a few are relevant.
         2. Return an empty array when none of the agents are a strong match.
         3. Do not fill the list just to reach ${no_of_suggestions}.
-        4. Only include an agent when its relevance_score is 70 or higher.
+        4. Only include an agent when its relevance_score is 85 or higher.
         5. Never invent agent ids. Use only ids from the ## Agent Source.
         6. Never return any agent whose id appears in ## Excluded Agent IDs.
+        7. Base every suggestion on explicit evidence in ## Content To Analyze, not on assumptions about what might be useful later.
+        8. If ## Content To Analyze is only a greeting, acknowledgement, chit-chat, or other generic text with no concrete task, return [].
+        9. Do not use the user profile by itself as a reason to recommend technical, coding, HTML, or app-building agents.
 
         Selection Strategy:
-        1. Format Compatibility: Analyze the ## ${type} to Process. If it contains structured JSON, component definitions, or raw HTML, prioritize agents capable of transforming these formats (e.g., converting HTML/JSON to functional apps).
-        2. Workflow Continuity: Focus on the "Lifecycle" of the project. If a UI or content has been drafted, the logical next step is conversion to a production-ready format, styling, or deployment. 
-        3. Contextual Relevance: Match the user's business profile and prompt category only when they materially improve the fit.
+        1. Evidence First: Recommend an agent only when the content clearly mentions a task, artifact, or format that the agent directly handles.
+        2. Format Compatibility: If the content explicitly contains structured JSON, component definitions, code, markup, or a direct request for those formats, prioritize agents specialized for that exact format.
+        3. Workflow Continuity: Use next-step reasoning only when the current content already demonstrates a concrete project step.
         4. Reject generic or weakly related agents. If the match is ambiguous, leave the agent out.
 
 
-        ## ${type}: ${context ? context : 'Use the previous response output as the content to analyze for next agents.'}
+        ## Content To Analyze (${type}): ${contentToAnalyze || 'No usable content.'}
  
         Input Context:
 
-        User Profile: ${JSON.stringify(account_doc.account_info)}
         Prompt Category: ${conversation_doc?.category_info?.category} 
         Excluded Agent IDs: ${JSON.stringify(exclude_agents)}
         
@@ -11343,7 +11377,7 @@ export const get_chat_suggestions = async function (req) {
             if (seen.has(agent.agent_id)) {
               return false;
             }
-            if (typeof agent.relevance_score !== 'number' || agent.relevance_score < 70) {
+            if (typeof agent.relevance_score !== 'number' || agent.relevance_score < 85) {
               return false;
             }
             const short_action = agent.short_action?.trim();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xuda.io/ai_module",
-  "version": "1.1.5425",
+  "version": "1.1.5427",
   "description": "Xuda AI Module",
   "main": "index.mjs",
   "type": "module",