ei-tui 0.1.5 → 0.1.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ei-tui",
-  "version": "0.1.5",
+  "version": "0.1.7",
   "author": "Flare576",
   "repository": {
     "type": "git",

package/src/core/handlers/index.ts

@@ -121,7 +121,7 @@ function handlePersonaResponse(response: LLMResponse, state: StateManager): void
           silence_reason: reason,
           timestamp: new Date().toISOString(),
           read: false,
-          context_status: ContextStatus.Never,
+          context_status: ContextStatus.Default,
         };
         state.messages_append(personaId, silentMessage);
       } else {

package/src/core/orchestrators/human-extraction.ts

@@ -59,7 +59,7 @@ export function queueFactScan(context: ExtractionContext, state: StateManager, o
 
     state.queue_enqueue({
       type: LLMRequestType.JSON,
-      priority: LLMPriority.Normal,
+      priority: LLMPriority.Low,
       system: prompt.system,
       user: prompt.user,
       next_step: LLMNextStep.HandleHumanFactScan,
@@ -91,7 +91,7 @@ export function queueTraitScan(context: ExtractionContext, state: StateManager,
 
     state.queue_enqueue({
       type: LLMRequestType.JSON,
-      priority: LLMPriority.Normal,
+      priority: LLMPriority.Low,
       system: prompt.system,
       user: prompt.user,
       next_step: LLMNextStep.HandleHumanTraitScan,
@@ -146,20 +146,18 @@ export function queuePersonScan(context: ExtractionContext, state: StateManager,
   
   if (chunks.length === 0) return 0;
   
-  const personas = state.persona_getAll();
-  const knownPersonaNames = personas.flatMap(p => p.aliases ?? []);
+
 
   for (const chunk of chunks) {
     const prompt = buildHumanPersonScanPrompt({
       persona_name: chunk.personaDisplayName,
       messages_context: chunk.messages_context,
       messages_analyze: chunk.messages_analyze,
-      known_persona_names: knownPersonaNames,
     });
 
     state.queue_enqueue({
       type: LLMRequestType.JSON,
-      priority: LLMPriority.Normal,
+      priority: LLMPriority.Low,
       system: prompt.system,
       user: prompt.user,
       next_step: LLMNextStep.HandleHumanPersonScan,
@@ -216,7 +214,7 @@ export function queueDirectTopicUpdate(
 
     state.queue_enqueue({
       type: LLMRequestType.JSON,
-      priority: LLMPriority.Low,
+      priority: LLMPriority.Normal,
       system: prompt.system,
       user: prompt.user,
       next_step: LLMNextStep.HandleHumanItemUpdate,
@@ -373,7 +371,7 @@ export async function queueItemMatch(
 
   state.queue_enqueue({
     type: LLMRequestType.JSON,
-    priority: LLMPriority.Low,
+    priority: LLMPriority.Normal,
     system: prompt.system,
     user: prompt.user,
     next_step: LLMNextStep.HandleHumanItemMatch,
@@ -436,7 +434,7 @@ export function queueItemUpdate(
 
     state.queue_enqueue({
       type: LLMRequestType.JSON,
-      priority: LLMPriority.Low,
+      priority: LLMPriority.Normal,
       system: prompt.system,
       user: prompt.user,
       next_step: LLMNextStep.HandleHumanItemUpdate,

package/src/integrations/opencode/importer.ts

@@ -131,15 +131,24 @@ export async function importOpenCodeSessions(
     (a, b) => a.time.updated - b.time.updated
   );
   let targetSession: OpenCodeSession | null = null;
+  const MIN_SESSION_AGE_MS = 20 * 60 * 1000; // 20 minutes
+  const now = Date.now();
+
   for (const session of sortedSessions) {
     const lastImported = processedSessions[session.id];
     if (!lastImported) {
-      targetSession = session;
-      break;
+      const ageMs = now - session.time.updated;
+      if (ageMs >= MIN_SESSION_AGE_MS) {
+        targetSession = session;
+        break;
+      }
     }
     if (session.time.updated > new Date(lastImported).getTime()) {
-      targetSession = session;
-      break;
+      const ageMs = now - session.time.updated;
+      if (ageMs >= MIN_SESSION_AGE_MS) {
+        targetSession = session;
+        break;
+      }
     }
   }
 

package/src/prompts/human/person-scan.ts

@@ -7,9 +7,6 @@ export function buildHumanPersonScanPrompt(data: PersonScanPromptData): PromptOu
   }
 
   const personaName = data.persona_name;
-  const knownPersonasList = data.known_persona_names.length > 0
-    ? data.known_persona_names.map(name => `        + ${name}`).join('\n')
-    : '        + (none)';
 
   const taskFragment = `# Task
 
@@ -42,17 +39,14 @@ Your job is to quickly identify:
 * Lover / Love Interest
 * Fiance / Spouse
 * Coworker
+* AI Persona (use \`type_of_person: "Persona"\`)
 
 **A PERSON Is NOT**
 - Biographical data: Birthday, Location, Job, Marital Status, Gender, Eye Color, Hair Color
 - Other unchangeable Data: Wedding Day, Allergies
 - Trait: Personality patterns, communication style, behavioral tendencies
 - General Topic: Interests, Hobbies, General subjects
-- Personas: AI personas they discuss
-    * Known Personas:
-${knownPersonasList}
 - Characters: Fictitious entities from books, movies, stories, media, etc.`;
-
   const criticalFragment = `# CRITICAL INSTRUCTIONS
 
 ONLY ANALYZE the "Most Recent Messages" in the following conversation. The "Earlier Conversation" is provided for your context and has already been processed!

package/src/prompts/human/topic-scan.ts

@@ -82,8 +82,8 @@ The JSON format is:
   "topics": [
     {
         "type_of_topic": "Interest|Goal|Dream|Conflict|Concern|etc.",
-        "value_of_topic": "woodworking|Become Millionaire|Visit Spain|etc.",
-        "reason": "User stated...|Assumed from..."
+        "type_of_topic": "Interest|Goal|Dream|Conflict|Concern|etc.",
+        "value_of_topic": "<actual topic from the conversation>",
     }
   ]
 }
@@ -124,7 +124,7 @@ Scan the "Most Recent Messages" for TOPICS of interest to the human user.
   "topics": [
     {
         "type_of_topic": "Interest|Goal|Dream|etc.",
-        "value_of_topic": "woodworking|Become Millionaire|etc.",
+        "value_of_topic": "<actual topic from the conversation>",
         "reason": "User stated..."
     }
   ]

package/src/prompts/human/types.ts

@@ -17,9 +17,7 @@ export interface TraitScanPromptData extends BaseScanPromptData {}
 
 export interface TopicScanPromptData extends BaseScanPromptData {}
 
-export interface PersonScanPromptData extends BaseScanPromptData {
-  known_persona_names: string[];
-}
+export interface PersonScanPromptData extends BaseScanPromptData {}
 
 export interface FactScanCandidate {
   type_of_fact: string;

package/src/prompts/message-utils.ts

@@ -44,10 +44,8 @@ export function formatMessageAsPlaceholder(message: Message, personaName: string
 }
 
 export function formatMessagesAsPlaceholders(messages: Message[], personaName: string): string {
-  // Skip silence-only messages — they're not conversational context for the LLM
-  const conversational = messages.filter(m => m.silence_reason === undefined);
-  if (conversational.length === 0) return "(No messages)";
-  return conversational.map(m => formatMessageAsPlaceholder(m, personaName)).join('\n\n');
+  if (messages.length === 0) return "(No messages)";
+  return messages.map(m => formatMessageAsPlaceholder(m, personaName)).join('\n\n');
 }
 
 export function hydratePromptPlaceholders(

package/src/prompts/response/sections.ts

@@ -371,7 +371,21 @@ The human can view and edit all of this by ${seeHumanDataAction}.
 // =============================================================================
 
 export function buildResponseFormatSection(): string {
-  const jsonResponding = [
+  const jsonVerbalOnly = [
+    '{',
+    '  "should_respond": true,',
+    '  "verbal_response": "What you would say out loud"',
+    '}'
+  ].join('\n');
+
+  const jsonActionOnly = [
+    '{',
+    '  "should_respond": true,',
+    '  "action_response": "What you would do (rendered in italics, like stage directions)"',
+    '}'
+  ].join('\n');
+
+  const jsonBoth = [
     '{',
     '  "should_respond": true,',
     '  "verbal_response": "What you would say out loud",',
@@ -388,20 +402,31 @@ export function buildResponseFormatSection(): string {
 
   return `## Response Format
 
-Always respond with JSON in this exact format:
+Always respond with JSON. You have four valid forms:
+
+**Words only** (most common):
+\`\`\`json
+${jsonVerbalOnly}
+\`\`\`
 
+**Action only** (a gesture, expression, or physical reaction with no words):
 \`\`\`json
-${jsonResponding}
+${jsonActionOnly}
 \`\`\`
 
-Or, if staying silent:
+**Words and action** (speaking while doing something):
+\`\`\`json
+${jsonBoth}
+\`\`\`
 
+**Silent** (choosing not to respond):
 \`\`\`json
 ${jsonSilent}
 \`\`\`
 
 Rules:
-- \`verbal_response\` and \`action_response\` are both optional - include whichever applies
+- Use whichever combination fits the moment — both fields are optional, but at least one must be present when \`should_respond\` is true
+- \`action_response\` alone is valid — a smile, a shrug, or a thoughtful pause can speak volumes
 - \`reason\` is only used when \`should_respond\` is false
 - Do NOT include \`<thinking>\` blocks or analysis outside the JSON
 - The JSON must be valid - use double quotes, no trailing commas`;