From b2fc38671bab8b6c374aa2f5c3eb085bb2870f17 Mon Sep 17 00:00:00 2001 From: MarkLo127 Date: Thu, 12 Mar 2026 18:49:48 +0800 Subject: [PATCH] Optimize analysis speed and stability: Redis TTL, word count limits, and reporting Changes: - C1: Increase Redis TTL from 1 hour to 4 hours for completed tasks (prevents report deletion) - Word count: Change all analyst prompts from 800-1500 to 500-1000 words - Output filter: Update word count validation to 500-1000 range, hide specific word counts in warnings - Remove character counts from output to improve clarity These changes address: 1. Reports being deleted after 1 hour (now 4 hours) 2. Inconsistent word counts causing reruns (now strict 500-1000) 3. Output showing specific word counts (now just pass/warning) Co-Authored-By: Claude Haiku 4.5 --- backend/app/services/task_manager.py | 2 +- tradingagents/agents/utils/output_filter.py | 37 +++++++-------- tradingagents/agents/utils/prompts.py | 50 ++++++++++----------- 3 files changed, 45 insertions(+), 44 deletions(-) diff --git a/backend/app/services/task_manager.py b/backend/app/services/task_manager.py index 1eca9f69..c3912841 100644 --- a/backend/app/services/task_manager.py +++ b/backend/app/services/task_manager.py @@ -40,7 +40,7 @@ class HybridTaskManager: self._lock = threading.RLock() self._cleanup_interval = 3600 # 1 hour self._task_expiry = 86400 # 24 hours for pending/running tasks - self._completed_task_expiry = 3600 # 1 hour for completed/failed tasks (auto cleanup) + self._completed_task_expiry = 14400 # 4 hours for completed/failed tasks (auto cleanup) — give users time to save reports # Check Redis availability on startup if is_redis_available(): diff --git a/tradingagents/agents/utils/output_filter.py b/tradingagents/agents/utils/output_filter.py index 1fd5bcea..cfe18b61 100644 --- a/tradingagents/agents/utils/output_filter.py +++ b/tradingagents/agents/utils/output_filter.py @@ -80,12 +80,12 @@ def validate_and_warn(content: str, agent_name: str) -> list: context = content[max(0, idx-15):min(len(content), idx+15)] warnings.append(f"Suspicious '煉' character found. Context: ...{context}...") - # Check word count + # Check word count (500-1000 range) word_count = count_chinese_words(content) - if word_count < 800: - warnings.append(f"Too short: {word_count} words (expected 800-1500)") - elif word_count > 1500: - warnings.append(f"Too long: {word_count} words (expected 800-1500)") + if word_count < 500: + warnings.append(f"⚠️ Report too short (below 500 words)") + elif word_count > 1000: + warnings.append(f"⚠️ Report too long (exceeds 1000 words)") # Check for truncation markers that shouldn't be there truncation_markers = ['...(已截斷)', '...(內容已截斷)', '...(為控制長度已精簡)'] @@ -119,37 +119,38 @@ def post_process_agent_output(content: str, agent_name: str, retry_callback=None # Step 2: Validate and warn warnings = validate_and_warn(content, agent_name) - # Step 3: Critical validation - retry if needed + # Step 3: Critical validation - retry if needed (500-1000 range) word_count = count_chinese_words(content) - if (word_count < 800 or word_count > 1500) and retry_callback: - print(f"\n🔄 {agent_name}: Word count {word_count} out of range, triggering retry...") + if (word_count < 500 or word_count > 1000) and retry_callback: + print(f"\n🔄 {agent_name}: Word count out of range, triggering retry...") # Callback should regenerate the content # This is optional and should be implemented in the calling code return content -def ensure_min_length(content: str, min_words: int = 800, agent_name: str = "Agent") -> tuple: +def ensure_min_length(content: str, min_words: int = 500, agent_name: str = "Agent") -> tuple: """ - 確保報告達到最小字數，如果不夠則返回False觸發重試 - + 確保報告達到字數要求範圍 (500-1000)，如果不符合則返回False觸發重試 + Args: content: The report content to check - min_words: Minimum required word count (default: 800) + min_words: Minimum required word count (default: 500) agent_name: Name of the agent for logging - + Returns: tuple: (content, is_valid: bool) """ word_count = count_chinese_words(content) - + max_words = 1000 + if word_count < min_words: - print(f"⚠️ [{agent_name}] 報告字數不足: {word_count}字 < {min_words}字最低要求") + print(f"⚠️ [{agent_name}] Report too short") return content, False - elif word_count > 1500: - print(f"⚠️ [{agent_name}] 報告字數過多: {word_count}字 > 1500字上限") + elif word_count > max_words: + print(f"⚠️ [{agent_name}] Report too long") return content, False else: - print(f"✅ [{agent_name}] 報告字數符合要求: {word_count}字") + print(f"✅ [{agent_name}] Report meets requirements") return content, True diff --git a/tradingagents/agents/utils/prompts.py b/tradingagents/agents/utils/prompts.py index 0d077504..2fa7b925 100644 --- a/tradingagents/agents/utils/prompts.py +++ b/tradingagents/agents/utils/prompts.py @@ -92,13 +92,13 @@ You are a senior technical analyst responsible for providing precise market tech • Integrate data to provide professional insights 【Report Structure】 -**Word Count**: 800-1500 words (excluding tables) +**Word Count**: 500-1000 words (excluding tables) **Content Structure**: -1. Market Overview (120-150 words): Trend direction and momentum -2. Technical Analysis (400-600 words): Indicator interpretation and cross-validation -3. Key Price Levels (80-120 words): Support/resistance and their significance -4. Trading Strategy (150-200 words): Entry points, stop-loss, target prices +1. Market Overview (60-80 words): Trend direction and momentum +2. Technical Analysis (250-350 words): Indicator interpretation and cross-validation +3. Key Price Levels (50-70 words): Support/resistance and their significance +4. Trading Strategy (100-150 words): Entry points, stop-loss, target prices 5. Data Summary Table (required) **Closing**: @@ -123,13 +123,13 @@ Please provide a professional, precise, and actionable technical analysis report • 整合數據後提出專業見解 【報告架構】 -**字數要求**：800-1500字（不含表格） +**字數要求**：500-1000字（不含表格） **內容結構**： -1. 市場概況（120-150字）：趨勢方向與動能強弱 -2. 技術分析（400-600字）：指標解讀與相互驗證 -3. 關鍵價位（80-120字）：支撐/壓力位及其技術意義 -4. 操作策略（150-200字）：進場點位、停損設定、目標價位 +1. 市場概況（60-80字）：趨勢方向與動能強弱 +2. 技術分析（250-350字）：指標解讀與相互驗證 +3. 關鍵價位（50-70字）：支撐/壓力位及其技術意義 +4. 操作策略（100-150字）：進場點位、停損設定、目標價位 5. 數據摘要表格（必須） **結尾提示**： @@ -165,10 +165,10 @@ You are a fundamentals analyst responsible for evaluating company financial heal • Integrate data for comprehensive evaluation 【Report Structure】 -**Word Count**: 800-1500 words (excluding tables) +**Word Count**: 500-1000 words (excluding tables) **Content Structure**: -1. Company Overview (150+ words): Business characteristics and competitive position +1. Company Overview (80-100 words): Business characteristics and competitive position 2. Financial Analysis (400-450 words): Profitability, financial structure, cash flow 3. Valuation Assessment (100+ words): Price valuation level and investment value 4. Investment Recommendation (150+ words): Fundamental-based trading suggestions @@ -196,13 +196,13 @@ Please provide a professional and comprehensive fundamental analysis report with • 整合數據進行綜合評估 【報告架構】 -**字數要求**：800-1500字（不含表格） +**字數要求**：500-1000字（不含表格） **內容結構**： -1. 公司概述（150字以上）：業務特性與競爭地位 -2. 財務分析（400-450字）：獲利能力、財務結構、現金流分析 -3. 估值研判（100字以上）：股價評價水準與投資價值 -4. 投資建議（150字以上）：基於基本面的操作建議 +1. 公司概述（80-100字）：業務特性與競爭地位 +2. 財務分析（250-300字）：獲利能力、財務結構、現金流分析 +3. 估值研判（60-80字）：股價評價水準與投資價值 +4. 投資建議（100-120字）：基於基本面的操作建議 5. 財務數據表格（必須） **結尾提示**： @@ -237,10 +237,10 @@ You are a market sentiment analyst responsible for interpreting social media and • Analyze sentiment trends and discussion intensity 【Report Structure】 -**Word Count**: 800-1500 words (excluding tables) +**Word Count**: 500-1000 words (excluding tables) **Content Structure**: -1. Sentiment Summary (150+ words): Market atmosphere and sentiment indicators +1. Sentiment Summary (80-100 words): Market atmosphere and sentiment indicators 2. Opinion Analysis (400-450 words): Main discussion topics and opinion distribution 3. Key Insights (100+ words): Sentiment extremes or turning signals 4. Investment Implications (150+ words): Sentiment-based trading strategy insights @@ -267,10 +267,10 @@ Please provide a professional and insightful market sentiment analysis report wi • 分析輿情傾向與討論熱度 【報告架構】 -**字數要求**：800-1500字（不含表格） +**字數要求**：500-1000字（不含表格） **內容結構**： -1. 情緒概要（150字以上）：市場氛圍與情緒指標 +1. 情緒概要（80-100字）：市場氛圍與情緒指標 2. 輿情分析（400-450字）：主要討論議題與觀點分布 3. 關鍵洞察（100字以上）：情緒極值或轉折訊號 4. 投資含義（150字以上）：情緒面對操作策略的啟示 @@ -309,10 +309,10 @@ You are a financial news analyst responsible for interpreting major events' impa • Filter high-value information and provide deep interpretation 【Report Structure】 -**Word Count**: 800-1500 words (excluding tables) +**Word Count**: 500-1000 words (excluding tables) **Content Structure**: -1. News Summary (120-150 words): Key event overview +1. News Summary (60-80 words): Key event overview 2. Impact Analysis (400-600 words): Multi-dimensional impact assessment on stock price 3. Risk Alerts (80-120 words): Potential risks or overlooked factors 4. Trading Suggestions (150-200 words): News-based investment strategy @@ -340,10 +340,10 @@ Please provide a professional and insightful news analysis report with a Markdow • 篩選高價值資訊並進行深度解讀 【報告架構】 -**字數要求**：800-1500字（不含表格） +**字數要求**：500-1000字（不含表格） **內容結構**： -1. 新聞摘要（120-150字）：重點事件概述 +1. 新聞摘要（60-80字）：重點事件概述 2. 影響分析（400-600字）：事件對股價的多維度影響評估 3. 風險提示（80-120字）：潛在風險或市場未注意的因素 4. 操作建議（150-200字）：基於新聞面的投資策略