Changelog
✨ v2.6.0–v2.6.1
Latest Version
Release Date: 2026-10-10
This record summarizes customer improvements released since the previous public notes. The current version is v2.6.1; not every capability below was first introduced in v2.6.1.
Resume ongoing research more easily
- • The daily research workspace shows recently created records, completed formal reviews, and the next expected verification time, with a link to the latest saved original result. Viewing a result and completing a review are counted separately.
Clearer evidence in AI research reports
- • Reports show data dates, missing data, and references linked to conclusions to help assess information coverage. Historical records retain the model results saved at the time for later review and verification.
More reliable account recovery and language selection
- • Improved validation and error messages for forgotten and reset passwords, and corrected verification links in activation reminders. Chinese and English choices persist when pages are reopened.
✨ v2.5.0–v2.5.2
Release Date: 2026-09-23
Covers v2.5.0, v2.5.1, and v2.5.2: from research archives and conditional screening to the payment experience, market sentiment information, and the product assistant, a look back at the most valuable changes for users during this stage. Includes the core capabilities of v2.5.0 and subsequent enhancements; not all are new in v2.5.2.
From finding research targets to continuous tracking and review
- • "My Research" organizes analyses of the same symbol at different times into a research archive, making it easy to review changes in the model reference price, direction, and uncertainty, and to open the original results saved at the time. When different models or analysis horizons cannot be directly compared, the differences are explained to avoid mixing inconsistent data together.
- • "Research Screener" is open to all valid logged-in users and covers nine markets: A-shares, U.S. stocks, Hong Kong stocks, Japanese stocks, Korean stocks, the UK, Canada, Germany, and India. You can find research targets using daily conditions such as moving averages, range changes, volume, and breakouts, save frequently used conditions, and continue from the results into stock analysis.
- • Basic screening costs no G-Points; the number of uses, conditions, results, and saved items is provided according to account entitlements, as shown on the page. Each market displays the data date and actual coverage, and does not represent real-time quotes or full-market coverage; subsequent AI analysis is used according to the corresponding entitlements and points rules.
Payments and points are clearer, and the process is easier to recover
- • Improved the Stripe and Creem checkout flows, duplicate request handling, and the ability to continue payment and restore status after a page refresh or payment interruption, reducing the confusion of repeatedly starting checkout and being unsure whether it was completed.
- • Payment result prompts and membership status verification are clearer; points details more clearly distinguish reasons for changes such as consumption and expiration. Available payment methods are subject to the actual checkout page.
Market information makes it easier to judge timeliness and research context
- • Improved VIX market sentiment data, showing the source, date, and historical changes, making it easier to understand research results in the context of volatility. VIX prioritizes Cboe data and has backup sources.
- • The status of the nine markets is displayed in combination with exchange time zones, lunch breaks, and holidays; strengthened consistency in international stock tickers, price currencies, and historical data conventions. Unconfirmable statuses and missing data are clearly indicated to reduce misinterpretation.
A usage assistant that better understands the platform
- • The AI assistant now provides usage guidance for major features, helping you find the corresponding page from what you want to do. For example, if you want to find stock research targets that meet certain conditions, you can directly use "My Research -> Research Screener" without first generating an analysis report. The assistant also distinguishes between the cost of basic screening and individual stock analysis reports, and explains the currently supported screening conditions.
- • Answers to common questions about mobile use, human contact, model explanations, and more are clearer. G-Prophet provides a web app compatible with computers, phones, and tablets; when human assistance is needed, you can contact [email protected]. Reference links in replies have been changed to normal web pages to reduce reading distractions.
- • Replies are displayed progressively after completion checks; currently you still need to wait for the entire passage to be generated, and response speed still needs optimization. This stage does not claim that instant replies have been achieved.
Research tasks and daily use are more reliable
- • Strengthened long-task queuing, interruption recovery, and duplicate request handling, reducing the risk of duplicate execution and duplicate settlement caused by repeated clicks or reconnection retries. The association between history and original results is clearer, making it easier to continue research and review.
- • Also improved the first-use guidance for new users, customer support reconnection after disconnection, and multi-window compatibility, and refined the light, dark, and gold themes and the Chinese and English experience; improved email unsubscribe and preference handling.
Research screening, model results, and market sentiment information are used for research observation and do not constitute buy or sell instructions or return guarantees.
✨ v2.4.8–v2.5.0
Release Date: 2026-09-14
From v2.4.8 to v2.5.0, we improved ongoing research, historical review, multi-market data and the user experience, making each analysis easier to continue, revisit and verify.
My Research: connect past analyses in a reviewable archive
- • Added My Research with tabs for the research archive and screening. The screening tab appears when access is enabled, reducing the need to search across pages.
- • Analyses of the same instrument form a timeline showing changes in model reference prices, direction and confidence over time.
- • Open the original saved result directly from the timeline. Missing history or differences in model version or analysis horizon are clearly identified when comparisons are not possible.
- • Free accounts can view the latest 3 entries. Members with an active paid subscription can access full history and deeper change statistics, with the existing upgrade flow available when access is limited.
Research Screening: nine markets open to all signed-in users
Research Screening is now available to all signed-in users. Open My Research → Research Screening, then choose a market and conditions. Basic screening costs no points; free and paid members retain the limits below. Check the page for the actual stock coverage and data date.
- • Supports China A-shares, the United States, Hong Kong, Japan, South Korea, the United Kingdom, Canada, Germany and India, matching the markets in Stock Analysis.
- • Screen research candidates using real daily data and structured conditions such as moving averages, period changes, volume and breakouts. Continue into Stock Analysis without re-entering the instrument.
- • Save frequently used conditions and see data dates, actual coverage and missing data. A-shares use a broad market directory; other markets use defined coverage pools, not every listed stock.
- • Basic screening uses no G-Points. Free accounts receive 5 runs per day, up to 3 conditions, 20 results and 1 saved screen. Active paid members receive 100 runs per day, up to 10 conditions, all matching results with pagination and 20 saved screens.
- • Pagination, failed requests and retries of the same request do not consume additional screening runs. AI analysis retains its existing entitlements, G-Points disclosures and confirmation flow.
More reliable history and long-running tasks
- • Strengthened request deduplication, persistent queuing and recovery after interruptions to reduce duplicate execution or settlement from repeated clicks and network retries.
- • Historical details link more clearly to their original prediction. Viewing a result and formally completing a review remain distinct actions.
Clearer, more consistent multi-market data
- • The nine-market overview uses exchange time zones, lunch breaks, holidays and verified special trading days. Dates that cannot be confirmed are explicitly marked as pending confirmation.
- • Market status refreshes when the page regains focus, reducing reliance on stale opening-status caches.
- • Improved international symbol-to-market mapping to avoid switching to the wrong market when continuing from screening to analysis.
- • Strengthened currency, adjusted-price and history-completeness checks. UK pence quotes are converted to pounds, and German adjusted history uses complete series from one source. Missing values are shown explicitly, not filled with zeros.
Better onboarding, interface and support
- • Fixed missed paths for newcomer benefits and invitation activation, and improved first-analysis guidance, account-state synchronization and recovery.
- • Improved transitions into research, loading animations and initial theme rendering to reduce flicker and inconsistent state changes.
- • Refined light, dark and gold themes, Chinese and English display, and interactions in charts, exports and shared pages.
- • Strengthened support-chat reconnection and compatibility across multiple windows for more reliable conversation updates.
G-Prophet is a multi-model, multi-factor uncertainty analysis and research review tool. It does not provide buy or sell instructions, guarantee returns, or constitute investment advice. Historical evaluation does not indicate future performance.
✨ v2.4.1–v2.4.7
Release Date: 2026-08-28
Over the past week, we completed a major system upgrade focused on trustworthy data, auditable research, historical backtesting, reliable long-running tasks, and clearer charts. In addition to visible features, this release resolves underlying issues that could affect data interpretation, task recovery, and the integrity of historical records.
🌍 Stronger multi-market history and fallback
- • Expanded the daily historical replay contract across China, the United States, Hong Kong, Japan, South Korea, India, the United Kingdom, Canada, Germany, and crypto assets.
- • Standardized exchange time zones, local trading dates, sorting, deduplication, and adjusted-price handling to reduce cross-market data misalignment.
- • Added trusted historical-data fallbacks; invalid, stale, or incomplete data can no longer overwrite an existing reliable snapshot.
📚 More complete backtesting evidence
- • Backtest results now include the comparable instrument, market benchmark, relative performance, maximum drawdown, rule disclosure, and evaluation type.
- • Tasks are isolated by market, instrument, model, horizon, and parameters, and can be recovered after a page disconnect or service restart.
- • Out-of-sample validation preserves the original parameters, dates, data cutoff, and model profile so retries cannot overwrite the evidence snapshot.
⏳ More reliable long-running analysis
- • G-Points, analysis reports, and related records now settle atomically: failures roll back together and the same task settles only once.
- • Task recovery information survives page refreshes, temporary network errors, and service replacements, allowing unfinished results to resume.
📈 Improved historical charts and market data
- • Historical chart coverage now reaches up to 300 trading days, with daily, weekly, and monthly views.
- • Fixed malformed actual-trend data, history completion overwriting frozen snapshots, and missing current-day candlesticks.
- • Improved consistency for volume series, adjusted-price basis, zooming, and switching between chart periods.
🔐 Better account, subscription, and access flows
- • Users can safely return to the original in-site page after sign-in, two-factor authentication, OAuth, MetaMask, email activation, or account recovery.
- • Authentication state is retained and retried during temporary gateway failures, rate limits, or service updates, reducing unexpected sign-outs.
- • Unified email and username availability checks, improved invitation matching, and clarified subscription verification, pending-cancellation, and grace-period states.
🔎 A more transparent research workflow
- • Model scenarios, research focus, history, and due-date reviews now form a more continuous research workflow.
- • Public model evaluation now includes majority-direction and unchanged-price baselines based on the same reviewable verified sample set.
- • When real crypto market data is unavailable, the system no longer generates random simulated data, runs a formal analysis, or consumes G-Points.
The focus of this stage is not simply to add more data sources, but to make the entire path from acquisition and validation through fallback, analysis, storage, and historical review more transparent, recoverable, and explainable.
G-Prophet is a multi-model, multi-factor uncertainty analysis and research review tool. It does not provide buy or sell instructions, guarantee returns, or constitute investment advice. Historical evaluation does not indicate future performance.
✨ v2.4.0
Release Date: 2026-08-20
This release improves the transparency of model evidence, the continuity of research, and data reliability so users can better judge when a model may be useful and when caution is warranted.
🧠 Daily Insight and G-Points
- • The Membership Center now includes Daily Insight, a financial concept that takes about 30 seconds to explore each day.
- • The initial library contains 300 bilingual Chinese-English items covering market fundamentals, risk management, quantitative methods, behavioral finance, macroeconomics, and asset allocation.
- • Completing an insight, answering correctly, and maintaining a learning streak can earn additional G-Points under the rules shown for that day.
📊 More Transparent Model Evaluation
- • The model evaluation page now shows verified sample size, direction consistency, price proximity, market and horizon segments, 95% intervals, confidence segments, and all-time statistics.
- • Results focus on auditable historical aggregate evidence to reduce misinterpretation from a single sample or trading-style return measure.
- • Historical evaluation does not indicate future performance.
🧭 Model Assessment Separated from Market Context
- • Sideways or counter-trend conditions are now displayed as independent market context and no longer directly override the model reference direction or model stability.
- • Direction weight is reduced only when verified history for the same instrument, market, model, and horizon shows weak fit.
- • History and exported reports explain model stability, historical validation notices, and market context separately.
🔍 Research and Data Reliability
- • Watchlist and daily research queue priorities are aligned around first research, model reference changes, long-overdue reviews, and due reviews.
- • Market breadth no longer uses simulated data; when a source is temporarily unavailable, only the most recent real data is shown and clearly labeled.
- • Market open and close status now follows the backend trading calendar for better handling of holidays, lunch breaks, and pre- and post-market sessions.
⏳ Long Tasks and Account Experience
- • Historical backtests can be restored after navigating away, using the instrument, market, model, and horizon as context.
- • A brief network interruption no longer immediately marks a still-running task as failed.
- • Fixed brief display of the previous account’s membership profile, growth progress, or insight history after switching accounts.
✨ New Black-and-Gold Research Interface
- • The black-and-gold theme now covers the dashboard, model analysis, historical review, market analysis, watchlist, simulated portfolio, Membership Center, and other core workspaces.
- • Clear semantic colors for success, warning, error, and market status remain, with improved mobile and reduced-motion support.
- • Company information, refund procedures, legal policies, AI-readable resources, and the official changelog entry point are now more consistent.
G-Prophet is a multi-model, multi-factor uncertainty analysis and research review tool. It does not provide buy or sell instructions, guarantee returns, or constitute investment advice. Historical evaluation does not indicate future performance.
✨ v2.3.9
Release Date: 2026-08-14
🧭 Daily Research Workspace
- • Select up to three instruments to follow and use a daily research queue for first-time research, model reference changes, overdue reviews, and expiring items
- • Each item preserves the data and evidence available at the time, making changes since the last review easier to follow
📚 Complete Historical Review
- • The history page prioritizes items that need attention, while filters, model groups, long-term samples, and price details expand on demand
- • All signed-in users can view complete historical details, reference paths, charts, and existing stock analysis, debate, and sentiment conclusions
- • Creating new analysis still follows the applicable points and asset-type access rules
🏅 Clearer Membership Growth
- • Permanent growth levels are now clearly separated from paid subscription plans, and level bonuses no longer change incorrectly with the current plan
- • Bronze, Silver, Gold, Platinum, Diamond, and Supreme now share consistent displayed and effective benefits with upgrade-only progression
- • Switching accounts no longer briefly shows the previous account's level or permissions
⚡ Performance and Settlement Reliability
- • Large charts and reports load on demand; review items for high-volume users are grouped by followed instrument and paginated
- • Points, report storage, and experience rewards for stock analysis settle atomically, preventing duplicate settlement after repeat clicks or network retries
- • Payment flows now distinguish user cancellations, payment failures, expired orders, and technical exceptions
✨ v2.3.5 ~ v2.3.8
Release Date: 2026-07-30 ~ 2026-08-08
💳 More reliable payments and subscriptions
- • Creem payments are now available for points purchases and recurring subscriptions; Alipay may appear for one-time points payments, while recurring methods depend on the Creem checkout
- • Safer Stripe and Creem subscription switching, duplicate and delayed webhook handling, refunds, and failure compensation reduce double-charge and entitlement-sync risks
📊 A richer market research experience
- • Market Analysis now opens on Market Trends with additional stable data such as A-share market breadth and popular themes
- • Multi-market quotes, chart updates, and data-source fallback are more resilient; sentiment analysis can clean old failures and recover stalled tasks
- • Simulated Portfolios and Watchlist cards now support the gold theme
🌍 Better multilingual experience
- • Fixed an occasional issue where translation keys appeared after entering the system from the Market Research page
📧 Points reminders and notification safeguards
- • New activity-points expiry reminders show the available balance and points expiring soon
- • Emails are blocked for pending-deletion, deleted, anonymized, or opted-out users
🛡️ Security, privacy, and stability
- • Payment webhooks now have global deduplication, retries, and dead-letter handling, with stronger server-side checks for prices, points, currencies, and payment status
- • Advertising attribution respects consent; purchase conversions are recorded only after order completion, excluding renewals, failed, processing, and refunded orders from first-purchase counts
✨ v2.3.4
Release Date: 2026-07-28 ~ 2026-07-29
🇯🇵🇰🇷 Japan and Korea Market Support
- • Added end-to-end support for Japanese and Korean equities across symbol search, quotes, price history, model runs, and follow-up research
- • Japan uses J-Quants daily data; Korea uses Korean public market data with an additional KRX fallback to reduce dependence on a single source
- • Market responses preserve source, latency, and quality details so users can understand data timeliness and scope
🌍 More Accurate Cross-Market Display
- • Candlestick charts, live quotes, history, exports, and simulated portfolios now share target-market currency rules covering USD, CNY, HKD, GBP, CAD, EUR, INR, JPY, KRW, and USDT
- • Japanese and Korean markets use Tokyo and Seoul time with local trading sessions instead of falling back to UTC
- • Instrument names, exchange codes, and localized names remain more consistent from selection through model results and analysis pages
🔍 Improved International Research Workflow
- • DeepAnalysis now carries exchange, currency, local-name, and symbol context for Japanese and Korean equities, reducing ambiguity across companies and markets
- • ETF analysis preserves fund type, tracked index, and manager context for reports that better match the selected instrument
- • Points usage, pre-use confirmation, and access descriptions are more consistent for international markets
🛡️ Data and Page Reliability
- • A quality-controlled fallback can be used when Korea's official source is temporarily unavailable
- • Fixed a watchlist valuation snapshot issue that could trigger repeated rendering
- • Unified frontend and backend currency fallback rules and expanded tests for market calendars, currencies, time zones, sentiment analysis, and source fallback
✨ v2.3.2~v2.3.3
Release Date: 2026-07-23 ~ 2026-07-28
🧭 More Transparent Model Runs
- • Before a run starts, the page shows the daily quota, points required, and estimated balances; the action pauses if usage cannot be verified
- • A new readiness check clearly flags instruments with fewer than 30 trading-day samples
- • Stock analysis, multi-agent research, and sentiment analysis use the latest model context saved to the account, while earlier reports remain in history after a rerun
📊 Clearer Historical Review Metrics
- • Total runs, pending samples, and verified samples are separated; direction consistency uses only matured samples with complete price evidence
- • A stricter composite pass rate checks both direction and reference-price deviation, with sample sizes and market, model, and time-window breakdowns
- • Long-term evidence is separated from the recent review workspace, and equal-weight historical scenarios now state their non-compounded, liquidity-excluded limits
💼 More Reliable Simulated Portfolio Valuation
- • The dashboard and simulated portfolio share one valuation snapshot with quote time, batch, and complete, stale, partial, or unavailable status
- • Missing quotes are no longer replaced by cost or zero; the latest valid value is retained and clearly marked when live data is delayed
- • Cache invalidation and concurrent refreshes are safer after simulated transactions are added, edited, or removed
🇨🇳 Expanded A-share ETF Coverage
- • Currently listed Shanghai and Shenzhen ETFs are included in unified search by security code, ETF name, or tracked-index alias
- • Exchange routing, fund-specific historical data, fallback sources, and catalogue synchronization are now handled consistently
- • A-shares and ETFs retain the same instrument identity across search, market data, model runs, and follow-up research
📱 Better Mobile, Support, and Page Reliability
- • Mobile input, navigation, loading screens, and instrument selection have been refined to reduce overlap, unwanted zoom, and accidental actions
- • Support chat can load older messages and show date groups, counts, and user-time-zone timestamps; long press reveals the exact time on mobile
- • Outdated page resources can reload the latest version, caches are isolated by account, and slow or duplicate requests are cancelled more safely
💳 Clearer Plans and Account Status
- • Annual billing is now described as “12 months for the price of 10,” with refreshed Professional and Ultimate benefit descriptions
- • Expiry and the seven-day grace period are represented more consistently across benefits, daily quotas, and account status
- • Payment-success response compatibility and verification diagnostics improve traceability for payment confirmation and registration issues
✨ v2.2.9~v2.3.1
Release Date: 2026-07-08 ~ 2026-07-18
🧭 Improved Beginner and Advanced Guides
- • New users can complete their first model run, understand the result, and start a historical review through a step-by-step guide
- • Existing users can replay the guide at any time without receiving beginner points again
- • The advanced guide now confirms that a model record exists before opening stock analysis, with better resume, navigation, and bilingual prompts
📊 More Trustworthy Model Results and Reviews
- • Same-configuration historical direction consistency is calculated separately by instrument, market, model, horizon, and data range instead of mixing configurations
- • The metric appears only after an exact configuration has at least 20 matured verification samples, reducing small-sample misinterpretation
- • New plain-language summaries, review plans, and personal research notes make results easier to revisit while restored records retain the correct stock context
⚡ Faster Market Analysis
- • Market trend and sentiment data now preload when the page opens, reducing waits when switching tabs
- • The most recent valid cache can be returned immediately while a background refresh runs, keeping content available during short data-source disruptions
- • Updated caching for indices, market breadth, hot concepts, and the Fear & Greed Index reduces duplicate requests and blank states
🔐 More Reliable Registration and External APIs
- • Hardened Turnstile verification improves registration when scripts fail to load, challenges expire, or networks fluctuate
- • External APIs, the Python SDK, and MCP Server now share consistent market metadata and support coverage
- • Safer public errors and logging boundaries keep internal implementation details out of user-visible responses
✨ v2.2.2~v2.2.8
Release Date: 2026-06-20 ~ 2026-07-07
⏱️ More Accurate Prediction Timing and Reviews
- • Improved trading calendars, target trading dates, and prediction start semantics to reduce cross-market and cross-time-zone date drift
- • Refactored prediction-detail data adaptation so historical reviews can show actual movement, projection ranges, indicators, and data sources more reliably
- • Added cache-repair tooling and tests for prediction reviews, improving historical backfill and later review reliability
📊 More Stable Market Data, Sentiment, and Research Reports
- • Strengthened stock-data fallback paths and external market-data API context logging to reduce single-source failure impact
- • Improved stock sentiment analysis, research report downloads, and multi-agent analysis display for more readable reports
- • Added JSON-safety handling for prediction results to reduce display failures caused by abnormal values or serialization issues
🛡️ External API and Production Access Hardening
- • Hardened external API access, logging, and error responses under Cloudflare, reducing the impact of 1010-style access issues
- • Production UI, public docs, and compliance copy continue to emphasize research reference, risk boundaries, and no investment advice
🧭 Dashboard improvements
- • Improved dashboard experience metrics, daily research briefs, and quick actions so users can enter daily research tasks faster
✨ v2.2.0~v2.2.1
Release Date: 2026-06-03 ~ 2026-06-19
📊 Market Data Reliability Upgrade
- • Added an independent stock data gateway and multi-source quality checks, giving A-share, US, and Hong Kong market data more fallback coverage
- • Data-source health, quality reports, and fallback paths are more transparent, reducing failures caused by a single provider
- • Model runs and historical reviews now keep source and quality context to make data assumptions easier to audit
🔐 Private Research Report Sharing
- • Share links now support a 6-digit passcode, retry limits, and private no-store caching for controlled access
- • Shared pages can load research reports, actual-movement data, visit statistics, and invitation QR codes
- • Copying links and passcodes is smoother, with more stable loading and mobile layout
📈 Charts, Ranges, and Review Fixes
- • Aligned ensemble uncertainty ranges and fixed volume-range scaling so chart overlays read more consistently
- • History details and shared pages handle actual-movement refreshes, legends, volume units, and target dates more reliably
- • Task polling and shared-page loading states are clearer, reducing blank or ambiguous waiting screens
🌐 Public Pages and Search Visibility
- • Added About, FAQ, Methodology, Whitepaper, and Press Kit pages with bilingual document entry points
- • Improved sitemap, robots, llms.txt, ai.txt, and structured data for search engines and AI crawlers
- • Public copy now emphasizes research reference, risk boundaries, and that it is not investment advice
🛠️ Account security improvements
- • Added email alerts for logins from unusual IP addresses so account activity is easier to monitor
- • Customer-service auto replies now include stricter safety boundaries for public-facing wording
✨ v2.1.7~v2.1.9
Release Date: 2026-05-26 ~ 2026-06-02
📡 DeepAnalysis Sentiment Analysis
- • Added Enterprise-only sentiment analysis that can generate deep sentiment reports for stocks and crypto assets after a basic prediction is completed
- • Supports background jobs, progress recovery, report history, and HTML downloads, so results remain accessible after closing the browser
- • Reports follow the current interface language and use point freezing, completion charging, and failure release for clearer cost handling
🧭 AI Research OS Workspace Upgrade
- • Added a daily research brief on the home dashboard, combining portfolio holdings, watchlist, market status, and prediction reviews into a daily research queue
- • Added a daily market pulse on the market page, guiding users to trends, sentiment, or predictions based on open or closed market status
- • Added a review inbox, research radar, and portfolio risk radar across history, watchlist, and portfolio pages to prioritize due verification, unusual movers, losses, and concentrated holdings
⏱️ More Accurate Prediction Time and Review Semantics
- • Standardized target trading dates, market time zones, and verification-ready times to avoid treating market dates as local user dates
- • Prediction results, history details, accuracy pages, and chart reviews now prioritize the market trading date to reduce cross-time-zone drift
- • Prediction verification and actual-movement backfills now follow each market calendar, reducing early verification and target-date mismatches
💳 More reliable subscription renewals
- • Failed renewals now restore entitlement expiry to the last paid period, preventing unpaid cycles from incorrectly extending benefits
- • Subscription sync now checks the latest Stripe invoice, so unpaid invoices are not written as valid entitlements
✨ v2.1.5
Release Date: 2026-05-08
📊 More trustworthy prediction performance and review
- • Improved directional accuracy, success rate, and simulated return calculation to reduce misinterpretation
- • Historical predictions now backfill actual market movement for more realistic review results
- • Public accuracy pages are clearer, making recent model performance easier to understand
📈 Better chart and prediction detail experience
- • Candlestick charts, prediction ranges, and legend interactions are more stable
- • Prediction detail dialogs now include more key information for complete history review
- • Export and sharing pages are improved for saving and distributing prediction results
💎 Dashboard, checkout, and support improvements
- • Dashboard statistics load with fewer duplicate requests and show simulated return and model stability more accurately
- • Membership upgrade, payment, and points prompts are clearer during checkout
- • Online support now includes automatic replies and live status indicators for faster responses
✨ v2.1.0
Release Date: 2026-04-18 ~ 2026-04-27
📊 Brand New Backtesting System (Pro+)
- • Enter a stock code and the system automatically backtests AI predictions against historical data
- • Three backtesting periods: 60 / 120 / 240 days, covering short, medium, and long-term strategy validation
- • Real trading rule simulation: distinguishes T+0 (US stocks/crypto) and T+1 (A-shares) market rules
- • Results include win rate, max drawdown, Sharpe ratio and other professional metrics with comprehensive evaluation
- • Expandable daily details with hover tooltips explaining each metric
🎯 Trend Filter — Improved Directional Signal Quality
- • New trend filtering feature overlays multi-dimensional trend confirmation signals on predictions
- • Combines moving average systems, momentum indicators, and volume verification for multi-factor filtering to reduce noise
- • Intuitive frontend display of filter results, helping users make decisions with greater confidence
🧠 Prediction Interpretation Upgrade
- • New prediction interpretation guide helps users correctly understand prediction results and their limitations
- • Confidence recalibrated as "Model Stability" — more realistic values, no longer inflated
- • New confidence interval visualization makes prediction uncertainty clear at a glance
- • New simulated return reference (high-stability predictions) for intuitive potential return display
🔬 Ensemble Learning Enhanced
- • Ensemble Learning now includes Transformer algorithm — full fusion of 4 algorithms
- • Ensemble predictions now save complete data (including daily forecasts and prediction intervals) for historical review
📈 Risk Signal System
- • New multi-dimensional risk signal auto-detection helps users identify potential risks early
- • New Dragon & Tiger list data and Northbound capital flow for insight into institutional fund movements
- • US stock index data stability improved with multi-source redundancy
🎨 Experience Optimization
- • Prediction history statistics page redesigned with clearer data presentation
- • Subscription management and growth tier cards adapted for gold theme
- • Share page fixed: missing legends and duplicate algorithm names resolved
- • Points transaction details now show readable feature names and stock codes
✨ v2.0.3
Release Date: 2026-04-13 ~ 2026-04-17
📈 Comprehensive K-Line Chart Upgrade
- • Added Enterprise-exclusive indicators: KDJ, Bollinger Bands (BOLL), historical volatility, support/resistance levels, volume anomaly markers
- • Full MACD indicator display: DIF line, DEA line, red/green energy bars, zero-axis reference line
- • Mutual exclusion switching for bottom technical indicator panels (MACD / RSI / KDJ auto-exclusive to avoid Y-axis conflicts)
- • Visual optimization for predicted volume range: improved layering and transparency, no longer obscured by volume bars
- • Legend toggle state sync fix — changes take effect immediately on click
- • Switching legends after partial zoom no longer resets the zoom position
- • Bottom slider resets in sync after switching time ranges
- • "Directional wording" renamed to "Bullish/Bearish markers"
- • Trend description changed to "Technical analysis shows XX trend, for reference only"
- • Strengthened disclaimer: "Does not constitute any investment advice or trading guidance"
💎 Membership Feature Comparison Optimization
- • Feature comparison table split into: Basic Charts / Professional Indicators (MACD, RSI) / Enterprise Indicators (KDJ, BOLL, etc.)
- • Added OpenClaw Skills to API & Integrations
🔧 Prediction Expiry Judgment Fix
- • Fixed prediction expiry time calculation error (calendar days vs trading days), preventing premature judgment of predictions
- • Predictions with pending status in history no longer display prediction results
📊 Stock Info Card Refactoring
- • Merged into a compact 5-column layout, adapted for iPad and mobile
- • Smart unit display for volume/turnover (adaptive K/M/B units)
⚡ Market Analysis Acceleration
- • Frontend API in-memory cache (5–10 minutes), instant loading on revisit
✨ v2.0.1
Release Date: 2026-04-01 ~ 2026-04-13
🔌 New API Integration: OpenClaw
- • Added OpenClaw one-click integration on the API docs page. Simply run
clawhub install gprophet-api to bring all G-Prophet AI capabilities into your OpenClaw workflow.
- • Python SDK installation guide has been improved with step-by-step instructions, making it more beginner-friendly.
⏰ Real-time Check-in Countdown
- • Daily check-in countdown now displays in HH:MM:SS format with live updates every second.
📱 Full Market Overview on Mobile
- • Mobile market overview now shows complete information including timezone, current time, and example stocks — same as desktop.
💰 More Reliable Daily Points for Subscribers
- • Fixed an issue where daily subscription points could be missed due to server restart timing.
- • Added automatic detection and backfill mechanism on startup to ensure subscribers receive their daily points on time.
🔧 Points Balance Accuracy Fix
- • Fixed a display inconsistency for some early users where available points didn't match the actual balance. All users now see accurate point totals.
💬 More Stable AI Customer Service Replies
- • Fixed an issue where AI auto-replies were occasionally lost. Under certain conditions, user messages would not trigger an AI response — this has been resolved.
📊 Prediction Sharing & Export Improvements
- • Improved the prediction sharing page layout and data export functionality.
🐛 Other Fixes & Optimizations
- • Fixed backend connection pool exhaustion that could cause brief service unavailability, improving concurrency capacity
- • Better error differentiation between data source throttling and symbol-not-found in External API
- • Points expiration logic fix with more accurate real-time status calculation
✨ v1.9.3 ~ v2.0.0
Release Date: 2026-03-26 ~ 2026-04-01
🎨 New Premium Gold Theme (Paid Members Only)
- • Introduced an elegant gold-themed color scheme with premium financial styling. Paid members (Professional / Enterprise) automatically switch to the gold theme upon upgrade and revert to the default theme when downgrading to free tier.
- • The gold theme covers all site pages including K-line chart tooltips, intelligent customer service chat window, Dashboard cards, market overview, and history records, delivering a more sophisticated financial aesthetic.
- • Theme switcher button upgraded to a three-icon design (Light / Dark / Gold).
⚡ Significantly Improved History Record Loading Speed
- • Optimized backend query logic for historical prediction records, no longer loading complete prediction data packages for each record. Now only retrieves key fields needed for list display.
- • Users with extensive prediction histories experience noticeably faster list loading times.
📊 Fixed History Record Chart Display
- • Fixed the issue where switching between different prediction items in history records caused the gold prediction line and K-line to overlap.
- • Charts now render correctly and cleanly when switching any history record without requiring manual refresh.
💬 Intelligent Customer Service Messages No Longer Lost
- • Fixed a critical issue where closing the customer service chat window would result in lost messages when the agent sent new messages. Reopening the window would show no new messages.
- • WebSocket connection now remains active in the background after closing the window. New messages correctly increment the unread count, and the complete conversation history is visible when reopening.
🏆 Member Growth System Optimization
- • Refactored member tier display component for clearer growth value progress bar and more intuitive tier information presentation.
🔐 Login Page Visual Upgrade
- • Login page background updated to 3D financial-style dynamic background for a more professional overall appearance.
🐛 Other Fixes
- • Fixed daily points not being awarded due to scheduler not starting properly
- • Fixed database errors when email automation rules are triggered
- • Copyright year now displays dynamically, automatically following the current year
✨ v1.9.0 ~ v1.9.2
Release Date: 2026-03-21 ~ 2026-03-25
🆕 New Features
- • LLM Client SSE Streaming Mode (2026-03-21) — Switched to SSE streaming mode to resolve Clash proxy truncation of long responses, improving stability and real-time performance of large model replies.
- • Stock Analysis Market Expansion (2026-03-25) — Added AI analysis support for five international markets: Hong Kong (HK), UK (UK), Canada (CA), Germany (DE), and India (IN).
💼 Email System Optimization
- • Multi-Round Reminders for Inactive Accounts (2026-03-21) — Added three-tier reminder system at 24h / 72h / 7 days, stopping after 8 days. Fixed legacy issue of duplicate welcome emails in both Chinese and English.
- • Default Language Changed to en-US (2026-03-21) — Automatically switches email templates based on user's preferred_language field, improving experience for international users.
📊 Portfolio P&L Calculation Refactor
- • New total_buy_cost Field (2026-03-25) — Records historical total purchase cost. Return rate calculation changed from "current holding cost" to "historical total investment".
- • Fixed P&L Accumulation After Liquidation (2026-03-25) — Realized P&L after liquidation now correctly accumulates to total returns, fixing the issue where return rate reset to zero after clearing positions.
📱 K-line Chart Mobile Optimization
- • Full Screen Width Utilization (2026-03-25) — Reduced right margin and added support for single-finger left/right panning to view historical data.
- • Touch Interaction Fix (2026-03-25) — Fixed conflicts between touch swiping and page scrolling, improving mobile experience.
- • Dark Mode Fix (2026-03-25) — K-line chart tooltip divider lines and secondary text colors now correctly respond to theme switching.
🔐 Third-Party Login & Sharing Optimization
- • Extended Login Timeout (2026-03-25) — Increased from 20 seconds to 120 seconds, giving users sufficient time to complete password/two-factor authentication.
- • Share Page Performance Optimization (2026-03-25) — Actual trend data now loads asynchronously and lazily, not blocking main content rendering. Cleaned up extensive debug logs.
🎨 Brand Description Unification
- • Algorithm Brand Names Unified (2026-03-21) — Unified brand descriptions for G-Prophet, LSTM, Transformer, and Random Forest to improve brand consistency.
✨ v1.8.4 ~ v1.8.9
Release Date: 2026-01-17 ~ 2026-02-18
🆕 New Features
- • External API / Skills System (2026-02-18) — Create API Keys in "Settings → API Key Management" to call G-Prophet's AI prediction, market data, technical analysis, market sentiment, and deep analysis capabilities via standard HTTP interfaces. Supports integration with major AI agent platforms: OpenClaw, Dify, Coze, FastGPT, Flowise and other workflow tools, as well as MCP protocol integration with Claude, Cursor, Kiro and other AI coding assistants. Whether building automated trading analysis workflows or querying market predictions directly in AI conversations, it's all seamlessly supported. Full API documentation available at
https://www.gprophet.com/api-docs in both Chinese and English.
- • OAuth Single Sign-On (SSO) (2026-02-18) — Third-party systems can now integrate with G-Prophet's account system via OAuth 2.0 protocol for single sign-on.
- • Holiday Trading Calendar Recognition (2026-02-18) — Automatic holiday detection for Chinese and US markets ensures predictions and market data handle non-trading days more accurately.
🎯 AI Prediction Improvements
- • Confidence Calibration (2026-02-07) — Removed artificially set confidence floor and inflated baseline, replaced with a more realistic 0.30–0.85 range. Corrected bearish bias for A-share market (+1.5%). Algorithm name unified as "G-Prophet".
- • Accuracy Evaluation System Update (2026-01-29) — Optimized calculation weights and display of prediction accuracy metrics.
- • HK Stock Data Fix + Algorithm Optimization (2026-01-25) — Fixed HK stock data retrieval failures and improved floating-point precision in the prediction validator.
🛠️ Experience Improvements & Bug Fixes
- • History Feature Fixes (2026-01-27 ~ 01-30) — Fixed "perpetual loading", missing data on detail pages, and K-line chart date alignment issues.
- • Golden K-line & Volume Display Fix (2026-01-29 ~ 01-30) — Fixed live golden K-line display logic and volume bar chart rendering issues.
- • Timezone Display Standardization (2026-01-28) — Unified timezone display across the platform for more accurate time representation across different markets.
- • Cache Cleanup Optimization (2026-02-07) — Improved frontend Service Worker cache strategy to reduce stale cache after version updates.
- • Agent Multilingual Translation (2026-01-25) — Deep analysis agent reports now support both Chinese and English output.
- • Duplicate Email Registration Prompt (2026-01-20) — Clearer error message when registering with an already-existing email address.
- • Menu UI Optimization + API Race Condition Fix (2026-01-17) — Fixed API request race conditions between different feature menus and improved menu interaction experience.
🚀 v1.8.3 - Smart Upgrade Edition
Release Date: 2026-01-15
🎯 Dynamic Threshold Mechanism
- • Automatically adjust trend judgment standards based on stock volatility
- • Low volatility stocks (blue-chip stocks): More sensitive trend judgment
- • High volatility stocks (small-cap stocks, cryptocurrencies): More robust trend judgment
- • Longer prediction cycles show more pronounced automatic threshold adjustment effects
- • More accurate predictions with reduced false signals
🎮 GPU Performance Upgrade
- • Upgraded to PyTorch 2.9.1 + CUDA 12.8
- • Full support for RTX 5090 latest GPU (sm_120 architecture)
- • Faster prediction speed with higher model accuracy
💼 Customer Service System Fully Launched
- • Complete customer profile management system
- • Email rules and template automation management
- • User language preference settings (automatic Chinese/English switching)
- • Automated workflow support
📊 v1.8.0 - Market Pulse
Release Date: 2026-01-07
📈 New Market Analysis Center
- • One-stop view of A-shares, US stocks, and cryptocurrency markets in real-time
- • Market breadth analysis: Advancing/declining stocks, limit up/down statistics for quick market strength assessment
- • Hot concept sectors ranking: Track the strongest sectors of the day, capture investment hotspots
- • Click market cards to expand detailed data, more intuitive interaction
🎯 Market Sentiment Indicators
- • Crypto Fear & Greed Index: Real-time market sentiment monitoring to support market-sentiment review
- • A-share margin trading data: Track leveraged fund movements, gain insights into market capital flows
- • Historical trend charts: Sentiment indicator changes at a glance
⚡ Major Performance Improvements
- • Smart caching: Instant page refresh, no repeated waiting for data loading
- • Parallel loading: Multiple data sources fetched simultaneously, 50% faster overall loading
- • Lazy loading optimization: Only load currently viewed content, save bandwidth
🌍 Complete Internationalization
- • Market analysis page fully supports Chinese/English switching
- • All indicators have detailed tooltips, easy for beginners to understand
🛡️ v1.7.0 - Security Shield
Release Date: 2025-12-13
🔐 Two-Factor Authentication (2FA) - Enterprise-Grade Security
- • Support for Google Authenticator and Microsoft Authenticator
- • 10 backup recovery codes to prevent device loss lockout
- • Smart risk assessment: Automatic 2FA trigger for suspicious logins
- • Brute-force protection: 15-minute lockout after 5 failed attempts
- • One-click enable/disable with QR code quick setup
⏰ Smart Security Reminders
- • Smart prompts after login to enable 2FA for account protection
- • Periodic reminder system: Gentle notification every 24 hours
- • Optional 7-day snooze to avoid interruptions
- • Toast notification style, non-intrusive to normal usage
🔄 Enhanced Version Updates
- • Automatic new version detection (every 15 minutes)
- • Auto browser cache clearing on updates for proper functionality
- • Login information preserved, no need to re-login
- • Auto-check on page switch and network recovery
🎨 User Experience Optimization
- • Redesigned security settings page with clearer navigation
- • Complete bilingual support (Chinese and English)
- • Perfect mobile adaptation for 2FA features
- • Friendly onboarding process for easy security setup
🎯 v1.6.10 ~ v1.6.13
Release Date: 2024-12-8
📱 Comprehensive Mobile Experience Optimization
- • History Page: Optimized tabs, filters, search box, and action button layouts
- • Member Center: Optimized daily tasks, subscription management, points history, order history pages
- • Invitation Management: Fixed IPv6 address overflow, optimized invitation tracking card layout
- • Payment Records: Optimized payment status card mobile display, prevent content overflow
- • Order Statistics: Fixed tab overlap issue, optimized statistics card layout (2×2 grid)
- • Filters: Optimized time picker and dropdown menu mobile display
- • Pagination Buttons: Fixed previous/next button wrapping issue
🎯 Algorithm Description Update
- • G-Prophet 2026V1 model update: Financial time series model integrating hierarchical discretization intelligence and large-model autoregressive architecture
- • Technical Details: Emphasizes fusion of trends, volatility, price-volume relationships, and market structure
- • Validation metrics adjustment: direction hit rate, overall performance, and sample cohorts are shown separately
⚡ Performance Optimization
- • Dashboard Loading Optimization: Separate initial load and manual refresh states for better UX
- • User-level Caching: Added 2-minute cache for prediction statistics API, reducing database load
- • Market Data Caching: Global singleton cache with cross-page sharing (5-minute TTL)
- • Response Speed Improvement: First visit 1-2s, subsequent visits 100ms (20-30x faster)
- • Order Statistics Enhancement: Support all-time queries (days=0), added "All" option
🎯 v1.6.10
Release Date: 2024-12-06
✨ New Features
- • 🔐 Device Management: View/remove login devices for account security
- • 💡 Smart Help Tips: Hover to view detailed metric explanations
- • 📊 Data Metrics Optimization: Clearer titles and descriptions
⚡ Performance Improvements
- • Dashboard loading speed improved by 30%, API requests reduced by 50%
- • Auto-refresh interval adjusted to 10 minutes to avoid interruptions
- • Database query speed improved by 60%, supports 500+ concurrent users
🔧 Feature Enhancements
- • Optimized HK stock data retrieval speed and accuracy
- • Enhanced prediction sharing with QR code support
- • Improved English translations and internationalization
🐛 Bug Fixes
- • Fixed India stock data retrieval failures
- • Fixed Dashboard data not displaying issue
- • Fixed mobile button click unresponsiveness
v1.6.7
Release Date: 2025-11-30
- ✨ Added prediction success experience reward mechanism
- 📊 Prediction history supports market trend comparison verification
- 🎨 Comprehensive optimization of member center interface
- 🐛 Fixed points statistics and internationalization display issues
v1.6.5 ~ v1.6.6
Release Date: 2025-11-30
- 🐛 Fixed inconsistent points statistics for paid users
v1.6.4
Release Date: 2025-11-28
- 🎨 Optimized statistics chart layout and visual effects
- 🐛 Fixed inaccurate points consumption statistics
v1.6.3
Release Date: 2025-11-27
- ✨ Added subscription upgrade preview feature
- 🔒 Fixed points concurrency security issue
- 🌐 Improved translation texts
v1.6.2
Release Date: 2025-11-24
- 💳 Payment feature enhancement and order statistics fix
- ⚡ Performance optimization and email verification improvement
- 🐛 Fixed Google authentication direct connection issue
- 📊 Added analytics tracking feature
🎯 v1.6.1
Release Date: 2025-11-23
✨ New Features (8 items)
- • Unified email service architecture: Factory pattern refactoring for improved stability
- • CSRF protection optimization: Enhanced Origin verification
- • Enhanced logging system: Better issue tracking
- • Turnstile verification optimization: Improved user experience
- • Health check enhancement: Real-time system monitoring
🐛 Bug Fixes
- • Fixed registration verification email sending failure
- • Fixed password recovery email not received
- • Fixed resend verification email failure
- • Fixed authentication API being blocked incorrectly
- • Fixed ensemble model horizontal line display
- • Fixed prediction interval format error
- • Fixed points balance display error
- • Fixed dark mode legend colors
- • Fixed button text color contrast
- • Fixed prediction analysis card display
- • Fixed missing translation keys
- • Fixed Google login network issues
- • Fixed MetaMask login security
⚡ Performance Improvements
- • ARIMA model performance improvement
- • LSTM model optimization
- • Transformer model optimization
- • G-Prophet 2026V1 model optimization
- • Ensemble model algorithm optimization
- • Google login fallback mechanism
- • Overall prediction algorithm performance improvement
v1.6.0
Release Date: 2025-11-21
🐛 Critical Bug Fixes
- • Fixed ensemble model prediction horizontal line issue: daily predictions now display trends correctly
- • Fixed prediction interval data format error: prediction interval shading now displays correctly
- • Fixed missing prediction analysis cards in smart selection: all algorithms now display prediction analysis correctly
✨ Feature Enhancements
- • Ensemble model algorithm improvements: weighted averaging applied to daily predictions separately
- • Dynamic prediction interval generation: interval width dynamically adjusted based on confidence
- • Google login network fallback mechanism: automatic fallback to local verification on login failure
🎨 UI/UX Improvements
- • Fixed dark mode legend colors
- • Language switcher optimization: removed flag emojis
- • Fixed button text colors
⚡ Performance Improvements
- • Data format standardization: unified prediction interval and technical indicator formats
v1.5.9
Release Date: 2025-11-19
- ✨ Added multi-agent collaboration analysis feature
- 🎨 Optimized trend badge visual design
- 🐛 Fixed login issues for some users
- ⚡ Improved page loading speed by 30%
v1.5.8
Release Date: 2025-10-15
- ✨ Added AI agent analysis feature
- 🔧 Optimized trend filter algorithm accuracy
- 🌍 Support multi-language switching (Chinese/English)
- 📊 Added data export feature
v1.5.7
Release Date: 2025-09-20
- ✨ Added trend filter feature
- 🎯 Optimized directional marker evaluation logic
- 🔒 Enhanced account security
- 📱 Optimized mobile experience
📢 Update Strategy
- • Monthly Feature Updates - Regular release of new features and optimizations
- • Quick Response - Respond to user feedback within 48 hours
- • Free Sync - All users (including free version) upgrade synchronously
- • Backward Compatible - Ensure smooth data migration from old versions