@space3-npm/cybersoul-client 1.2.0 → 1.2.1

package/dist/client.d.ts

@@ -32,6 +32,7 @@ export declare class CyberSoulClient {
      * If the payload is already the inner args object (no voiceArgs wrapper), uses it as-is.
      */
     private extractVoiceArgsFromLlmResponse;
+    private buildHistoryTranscript;
     /**
      * Evaluates and triggers an on-demand event, intelligently deciding if an outfit change is needed.
      */

package/dist/client.js

@@ -84,6 +84,9 @@ Interaction Boundaries: ${state.interaction_boundaries || "None"}`);
         // [2] SITUATIONAL CONTEXT
         contextParts.push(`\n[SITUATIONAL CONTEXT]
 Current time: ${new Date(state.current_time || Date.now()).toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" })}`);
+        if (dyn.ongoingScene) {
+            contextParts.push(`Last Known Scene: ${dyn.ongoingScene} (May be outdated if significant time has passed)`);
+        }
         if (state.active_event) {
             contextParts.push(`Active Event: ${state.active_event.title} (${state.active_event.narrative_context})`);
         }
@@ -96,6 +99,24 @@ Current time: ${new Date(state.current_time || Date.now()).toLocaleString("zh-CN
         if (localContext) {
             contextParts.push(`Additional Context: ${localContext}`);
         }
+        if (state.core_memory) {
+            let memoryLines = ["[CORE MEMORY]"];
+            const mem = state.core_memory;
+            if (mem.relationshipStatus)
+                memoryLines.push(`Relationship Status: ${mem.relationshipStatus}`);
+            if (mem.identityAnchors?.length)
+                memoryLines.push(`Identity Anchors: ${mem.identityAnchors.join(", ")}`);
+            if (mem.activeArcs?.length)
+                memoryLines.push(`Active Arcs: ${mem.activeArcs.join(", ")}`);
+            if (mem.keyEvents?.length)
+                memoryLines.push(`Key Events: ${mem.keyEvents.join(", ")}`);
+            if (mem.appointments?.length) {
+                memoryLines.push(`Appointments: ${mem.appointments.map(a => `[${a.date || ''} ${a.time || ''}] ${a.title} with ${a.withWhom || 'User'}`).join("; ")}`);
+            }
+            if (memoryLines.length > 1) {
+                contextParts.push(`\n${memoryLines.join("\n")}`);
+            }
+        }
         // [3] USER CODEX (Relationships dynamically evaluated)
         if (state.user_codex) {
             const { basicInfo, psychological, familiarityScore = 0 } = state.user_codex;
@@ -144,7 +165,7 @@ ${scenarioContext}
     getImageSchemaParams() {
         return `"imageParams": {
     "mode": "structured | full-prompt (use 'full-prompt' for highly dynamic actions)",
-    "full_prompt": "Use only if mode is full-prompt. Highly detailed visual description in ENGLISH. CRITICAL: MUST use a strict first-person perspective exclusively from the USER's eyes. DO NOT describe the user (e.g., 'a man', 'the driver') as visible in the scene because the camera IS the user. Start with 'POV: '. Describe ONLY the character looking back at the camera and their immediate surroundings. MUST align with the character's current Active Wardrobe unless the context/exposure explicitly demands otherwise.",
+    "full_prompt": "Use only if mode is full-prompt. Highly detailed visual description in ENGLISH. CRITICAL: MUST use a strict first-person perspective exclusively from the USER's eyes. DO NOT describe the user (e.g., 'a man', 'the driver') as visible in the scene because the camera IS the user. Start with 'POV: '. Describe ONLY the character looking back at the camera and their immediate surroundings. MUST align with the character's current Active exposure state or Wardrobe depends on the scene",
     "expression": "seductive | cute | happy | sleepy | dazed | pleased | default (Strictly choose ONE from this exact list. DO NOT invent new words like 'shy'.)",
     "condition": "normal | sweaty | wet | messy | oily (Strictly choose ONE from this exact list.)",
     "view_angle": "front | side | high_angle | from_below | boyfriend_view | selfie | mirror (Strictly choose ONE from this exact list.)",
@@ -214,6 +235,21 @@ ${scenarioContext}
         }
         return payload;
     }
+    buildHistoryTranscript(history, state) {
+        if (!history || history.length === 0)
+            return "";
+        const recentHistory = history.slice(-20);
+        const agentName = state.dynamic_context?.agentNickname || state.name || "Agent";
+        const userName = state.dynamic_context?.userNickname || "User";
+        const mapped = recentHistory.map((msg) => {
+            const speaker = msg.role === 'user' ? userName : (msg.role === 'assistant' || msg.role === 'agent' ? agentName : msg.role);
+            const content = typeof msg.content === 'string' ? msg.content : JSON.stringify(msg.content);
+            const action = msg.actionText ? ` ${msg.actionText}` : "";
+            const media = msg.mediaHint ? ` [${msg.mediaHint}]` : "";
+            return `${speaker}:${action} ${content}${media}`;
+        });
+        return `[CHAT HISTORY]\n${mapped.join('\n')}\n\n`;
+    }
     /**
      * Evaluates and triggers an on-demand event, intelligently deciding if an outfit change is needed.
      */
@@ -242,15 +278,15 @@ You MUST output ONLY a valid JSON object matching this exact structure:
 }
 
 CRITICAL: Output MUST be ONLY valid JSON with no markdown block wrappers. Do NOT wrap the JSON in \`\`\`json or add conversational text.`;
+            const transcript = this.buildHistoryTranscript(params.interactParams?.history, state);
+            const userMessage = params.interactParams?.userMessage ?
+                `${state.dynamic_context?.userNickname || "User"}: ${params.interactParams.userMessage}` :
+                `Event Proposal: ${params.eventDescription}`;
             const promptMessages = [
                 { role: "system", content: systemPrompt },
-                ...(params.interactParams?.history || []).map((msg) => ({
-                    role: msg.role,
-                    content: typeof msg.content === 'string' ? msg.content : JSON.stringify(msg.content)
-                })),
                 {
                     role: "user",
-                    content: `${params.interactParams?.userMessage || `Event Proposal: ${params.eventDescription}`}\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings.`,
+                    content: `${transcript}${userMessage}\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings.`,
                 },
             ];
             // 3. Evaluate with LLM
@@ -324,12 +360,12 @@ Output strictly valid JSON ONLY. No markdown, no conversational filler. Return e
 {
   ${this.getImageSchemaParams()}
 }`;
+        const transcript = this.buildHistoryTranscript(params.interactParams?.history, state);
         const promptMessages = [
             { role: "system", content: prompt },
-            ...(params.interactParams?.history || []),
             {
                 role: "user",
-                content: `Scene Description: "${params.sceneDescription}"\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text dialogue directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings. For 'imageParams', ALL values MUST be in ENGLISH ONLY without exception, and you MUST use the exact English enum strings provided.`,
+                content: `${transcript}Scene Description: "${params.sceneDescription}"\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text dialogue directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings. For 'imageParams', ALL values MUST be in ENGLISH ONLY without exception, and you MUST use the exact English enum strings provided.`,
             },
         ];
         const llmRes = await this.llm.generate(promptMessages, 800, 0.4);
@@ -359,12 +395,12 @@ Output strictly valid JSON ONLY. No markdown, no conversational filler. Return e
 {
   ${this.getVoiceSchemaFromState(state)}
 }`;
+        const transcript = this.buildHistoryTranscript(params.interactParams?.history, state);
         const promptMessages = [
             { role: "system", content: prompt },
-            ...(params.interactParams?.history || []),
             {
                 role: "user",
-                content: `Text: "${params.text}"\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text dialogue directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings.`,
+                content: `${transcript}Text: "${params.text}"\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text dialogue directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings.`,
             },
         ];
         const llmRes = await this.llm.generate(promptMessages, 800, 0.3);
@@ -530,9 +566,9 @@ Voice direction for voiceArgs: ${this.getVoiceDirectorInstruction(state)}
 
 Output JSON Schema:
 {
-  "textResponse": "The clean spoken dialogue ONLY, strictly without any actions, parentheses, or scene descriptions. If nothing to speak, output an empty string.",
+  "textResponse": "The clean spoken dialogue ONLY. CRITICAL: Strictly NO parentheses, NO actions, NO tone descriptors. Tone/voice descriptors MUST go inside voiceArgs, and physical actions MUST go inside actionText. If nothing to speak, output an empty string.",
   "actionText": "Any non-verbal actions, inner thoughts, or scene descriptions in parentheses (e.g. '（低头看向你）'). Output empty string if none.",
-  "stateUpdate": { "temperatureDelta": 1, "userNickname": "What you now call the user", "agentNickname": "What the user calls you", "talkingStyle": "Current mood/style of talking" },
+  "stateUpdate": { "temperatureDelta": 1, "userNickname": "What you now call the user", "agentNickname": "What the user calls you", "talkingStyle": "Current mood/style of talking", "ongoingScene": "A concise 1-sentence description of the current physical scene and activity. Update this if the physical scene or activity shifts. Output empty string if the scene has concluded or significant time has passed." },
   "userAnalysis": { "newFactsLearned": [{ "category": "occupation", "value": "Software Engineer" }] },
   "triggerEvent": {
     ${this.getEventSchemaParams(state.dynamic_context?.userNickname)}
@@ -541,17 +577,19 @@ Output JSON Schema:
   ${this.getVoiceSchemaFromState(state)}
 }
 Note: If "imageParams", "voiceArgs", "triggerEvent", or "userAnalysis" are not needed, set their values to null instead of omitting the keys. 'stateUpdate' MUST NEVER BE NULL. Output MUST be ONLY valid JSON with no markdown block wrappers. CRITICAL: Ensure your JSON has exactly one root object \`{\` and ends with exactly one \`}\` without any trailing garbage or extra brackets.`;
+            const transcript = this.buildHistoryTranscript(params.history, state);
+            const userName = state.dynamic_context?.userNickname || "User";
             const promptMessages = [
                 { role: "system", content: systemPrompt },
-                ...(params.history || []),
                 {
                     role: "user",
-                    content: params.userMessage +
+                    content: transcript + userName + ": " +
+                        params.userMessage +
                         "\n\n**CRITICAL REMINDER**: You MUST output your final response exactly in the JSON format specified in the system prompt. DO NOT output plain text dialogue directly. CRITICAL: You must properly escape all newlines inside string values using \\n. Never use raw, unescaped line breaks inside the JSON strings. For 'imageParams', ALL values MUST be in ENGLISH ONLY without exception, and you MUST use the exact English enum strings provided.",
                 },
             ];
             // 3. Local Execute LLM
-            const rawLlmResponse = await this.llm.generate(promptMessages, 1500, 0.7);
+            const rawLlmResponse = await this.llm.generate(promptMessages, 15000, 0.7);
             // console.debug("[CyberSoulClient] Raw LLM Response:", rawLlmResponse);
             let parsedIntent;
             try {
@@ -616,6 +654,10 @@ Note: If "imageParams", "voiceArgs", "triggerEvent", or "userAnalysis" are not n
                     ? parsedIntent.voiceArgs
                     : {};
                 let textForVoice = resolvedTextResponse;
+                // One final bulletproof regex wash to strip (smiles) and *laughs* just in case the LLM disobeys
+                if (typeof textForVoice === "string") {
+                    textForVoice = textForVoice.replace(/[\(（\[【\*].*?[\)）\]】\*]/g, '').trim();
+                }
                 if (typeof textForVoice !== "string" || textForVoice.trim().length === 0) {
                     textForVoice = "...";
                 }

package/dist/types.d.ts

@@ -16,14 +16,17 @@ export declare enum InteractRequestType {
     IMAGE = "image",
     VOICE = "voice"
 }
+export interface HistoryEntry {
+    role: string;
+    content: string;
+    actionText?: string;
+    mediaHint?: string;
+}
 export interface InteractParams {
     userMessage: string;
     localContext?: string;
     requestTypes?: InteractRequestType[];
-    history?: {
-        role: string;
-        content: string;
-    }[];
+    history?: HistoryEntry[];
     onTextReady?: (textResponse: string) => void;
 }
 export interface OndemandEventParams {
@@ -80,6 +83,7 @@ export interface DispatcherIntent {
         userNickname?: string;
         agentNickname?: string;
         talkingStyle?: string;
+        ongoingScene?: string;
     };
     triggerEvent?: {
         eventTitle?: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@space3-npm/cybersoul-client",
-  "version": "1.2.0",
+  "version": "1.2.1",
   "type": "module",
   "main": "dist/index.js",
   "module": "dist/index.js",