Phase 4: Template system with auto-select and pre-fill

Backend:
- Add templates.json with 5 template definitions (news, policy, business, fiction, social)
- Add template API (/api/template/list, /api/template/auto-select, /api/template/:id/filter-rules)
- Register template blueprint in Flask app

Frontend:
- Add template API client (frontend/src/api/template.js)
- Add template selector UI in Home.vue (chip buttons + auto-select button)
- Add template state management and auto-select logic

Locale:
- Add template keys for th/en/zh

Entity filter rules in templates.json for context-aware filtering in Step 1.

backend/app/__init__.py

@@ -64,9 +64,11 @@ def create_app(config_class=Config):
     
     # 注册蓝图
     from .api import graph_bp, simulation_bp, report_bp
+    from .api.template import template_bp
     app.register_blueprint(graph_bp, url_prefix='/api/graph')
     app.register_blueprint(simulation_bp, url_prefix='/api/simulation')
     app.register_blueprint(report_bp, url_prefix='/api/report')
+    app.register_blueprint(template_bp, url_prefix='/api/template')
     
     # 健康检查
     @app.route('/health')

backend/app/api/template.py

@@ -0,0 +1,147 @@
+"""
+Template Auto-Selection API
+Analyze seed data and recommend the best simulation template
+"""
+
+import json
+import os
+from flask import Blueprint, request, jsonify
+from ..utils.llm_client import LLMClient
+from ..utils.locale import t, get_locale, get_language_instruction
+from ..utils.logger import get_logger
+
+logger = get_logger('crowdsight.template')
+
+template_bp = Blueprint('template', __name__)
+
+# Load templates
+_templates_path = os.path.join(os.path.dirname(__file__), '..', 'templates.json')
+with open(_templates_path, 'r', encoding='utf-8') as f:
+    _templates_data = json.load(f)
+    _templates = _templates_data['templates']
+
+
+@template_bp.route('/list', methods=['GET'])
+def list_templates():
+    """Return all available templates"""
+    locale = get_locale()
+    lang_key = f'prompt_{locale}' if locale in ('th', 'en', 'zh') else 'prompt_en'
+    
+    result = []
+    for tmpl in _templates:
+        result.append({
+            'id': tmpl['id'],
+            'icon': tmpl['icon'],
+            'category': tmpl['category'],
+            'prompt': tmpl.get(lang_key, tmpl['prompt_en']),
+            'placeholders': tmpl['placeholders'],
+            'name': t(f'templates.{tmpl["id"]}'),
+        })
+    
+    return jsonify({'success': True, 'templates': result})
+
+
+@template_bp.route('/auto-select', methods=['POST'])
+def auto_select_template():
+    """
+    Analyze seed data and recommend the best template + pre-fill prompt.
+    
+    Request JSON:
+        {
+            "text": "extracted text from uploaded documents",
+            "simulation_requirement": "optional user requirement"
+        }
+    
+    Response JSON:
+        {
+            "success": true,
+            "template_id": "news_event",
+            "prompt": "pre-filled prompt with actual values",
+            "confidence": 0.85,
+            "reasoning": "why this template was selected"
+        }
+    """
+    try:
+        data = request.get_json() or {}
+        text = data.get('text', '')[:5000]  # Limit to 5000 chars
+        simulation_requirement = data.get('simulation_requirement', '')
+        
+        if not text:
+            return jsonify({'success': False, 'error': 'No text provided'}), 400
+        
+        locale = get_locale()
+        lang_instruction = get_language_instruction()
+        
+        # Build template descriptions for the LLM
+        template_desc = []
+        lang_key = f'prompt_{locale}' if locale in ('th', 'en', 'zh') else 'prompt_en'
+        for tmpl in _templates:
+            template_desc.append(f"- {tmpl['id']}: {tmpl.get(lang_key, tmpl['prompt_en'])}")
+        
+        templates_list = '\n'.join(template_desc)
+        
+        system_prompt = f"""You are a smart assistant that analyzes document content and recommends the best simulation template.
+
+Available templates:
+{templates_list}
+
+Your task:
+1. Analyze the document content
+2. Select the MOST appropriate template
+3. Fill in the template placeholders with actual values from the document
+4. Return the result in JSON format
+
+{lang_instruction}"""
+
+        user_prompt = f"""Document content:
+{text[:3000]}
+
+{f'User requirement: {simulation_requirement}' if simulation_requirement else ''}
+
+Return JSON:
+{{
+    "template_id": "best_template_id",
+    "prompt": "the template with placeholders filled in with actual values from the document",
+    "confidence": 0.0-1.0,
+    "reasoning": "brief explanation of why this template was chosen"
+}}"""
+
+        llm = LLMClient()
+        result = llm.chat_json(
+            messages=[
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_prompt}
+            ],
+            temperature=0.3
+        )
+        
+        # Validate template_id
+        valid_ids = [t['id'] for t in _templates]
+        if result.get('template_id') not in valid_ids:
+            result['template_id'] = 'news_event'  # Default fallback
+        
+        return jsonify({
+            'success': True,
+            'template_id': result.get('template_id', 'news_event'),
+            'prompt': result.get('prompt', ''),
+            'confidence': result.get('confidence', 0.5),
+            'reasoning': result.get('reasoning', ''),
+        })
+    
+    except Exception as e:
+        logger.error(f"Template auto-select failed: {e}")
+        return jsonify({'success': False, 'error': str(e)}), 500
+
+
+@template_bp.route('/<template_id>/filter-rules', methods=['GET'])
+def get_filter_rules(template_id):
+    """Return entity filter rules for a template (used by Step 1)"""
+    for tmpl in _templates:
+        if tmpl['id'] == template_id:
+            return jsonify({
+                'success': True,
+                'template_id': template_id,
+                'filter_rules': tmpl.get('entity_filter', {})
+            })
+    
+    return jsonify({'success': False, 'error': 'Template not found'}), 404

backend/app/templates.json

@@ -0,0 +1,70 @@
+{
+  "templates": [
+    {
+      "id": "news_event",
+      "icon": "📰",
+      "category": "news",
+      "prompt_th": "จำลองปฏิกิริยาของประชาชนบนโซเชียลมีเดียหลังจาก [เหตุการณ์] ในช่วง [ระยะเวลา] ชั่วโมง",
+      "prompt_en": "Simulate public reactions on social media after [event] within [timeframe] hours",
+      "prompt_zh": "模拟[事件]发生后公众在社交媒体上的反应，时间范围[时间]小时",
+      "placeholders": ["event", "timeframe"],
+      "entity_filter": {
+        "exclude_types": [],
+        "focus": "affected_parties, public, media, officials"
+      }
+    },
+    {
+      "id": "policy_regulation",
+      "icon": "📋",
+      "category": "policy",
+      "prompt_th": "จำลองผลกระทบของนโยบาย [ชื่อนโยบาย] ต่อกลุ่มต่างๆ ในสังคม รวมถึงการโต้ตอบบนโซเชียลมีเดีย",
+      "prompt_en": "Simulate the impact of [policy name] on different social groups, including social media interactions",
+      "prompt_zh": "模拟[政策名称]对不同社会群体的影响，包括社交媒体互动",
+      "placeholders": ["policy_name"],
+      "entity_filter": {
+        "exclude_types": [],
+        "focus": "citizens, experts, media, government, affected_groups"
+      }
+    },
+    {
+      "id": "business_ad",
+      "icon": "📢",
+      "category": "business",
+      "prompt_th": "จำลองการแพร่กระจายและผลตอบรับของ [แคมเปญ/สินค้า] บนโซเชียลมีเดีย รวมถึงปฏิกิริยาของกลุ่มเป้าหมาย",
+      "prompt_en": "Simulate the spread and reception of [campaign/product] on social media, including target audience reactions",
+      "prompt_zh": "模拟[活动/产品]在社交媒体上的传播和反馈，包括目标受众的反应",
+      "placeholders": ["campaign_product"],
+      "entity_filter": {
+        "exclude_self": true,
+        "exclude_types": ["Advertiser", "BrandOwner"],
+        "focus": "target_audience, competitors, influencers, media"
+      }
+    },
+    {
+      "id": "fiction_story",
+      "icon": "📖",
+      "category": "fiction",
+      "prompt_th": "จำลองเรื่องราวต่อจาก [เนื้อเรื่อง] โดยให้ตัวละครแต่ละตัวมีปฏิสัมพันธ์กันบนโซเชียลมีเดีย และทำนายตอนจบ",
+      "prompt_en": "Simulate the story continuation from [plot] with characters interacting on social media, and predict the ending",
+      "prompt_zh": "模拟从[情节]开始的故事延续，角色在社交媒体上互动，并预测结局",
+      "placeholders": ["plot"],
+      "entity_filter": {
+        "exclude_types": ["Author", "Narrator"],
+        "focus": "characters, story_entities"
+      }
+    },
+    {
+      "id": "social_culture",
+      "icon": "🌍",
+      "category": "social",
+      "prompt_th": "จำลองผลกระทบของ [กระแสสังคม] ต่อกลุ่มคนต่างๆ บนโซเชียลมีเดีย รวมถึงการแพร่กระจายของข้อมูล",
+      "prompt_en": "Simulate the impact of [social trend] on different groups on social media, including information spread",
+      "prompt_zh": "模拟[社会潮流]对不同群体在社交媒体上的影响，包括信息传播",
+      "placeholders": ["social_trend"],
+      "entity_filter": {
+        "exclude_types": [],
+        "focus": "public_figures, communities, media, influencers, general_public"
+      }
+    }
+  ]
+}