Changelog

Sparklights — the color highlighting feature introduced on March 31 — received several improvements this week. A persistence bug where highlights would sometimes fail to save has been fixed with more reliable upsert logic and error logging. Sparklights can now be exported alongside your Sparkmarks from the export page. And the floating flame that appears when you select text has been refined for smoother interaction on both desktop and mobile.

Sparklights are still being actively debugged. If you notice highlights that don’t persist or selection behavior that feels off, we’re aware and working on it.

Karma Sparks — the lotus-icon likes on journal entries — are now represented visually across the site. Every entry card and entry page shows a lotus icon with a running count, so you can show appreciation for a fellow practitioner’s reflection wherever entries appear: the journal feed, verse and sūtra pages, practitioner profiles, and the subscriptions feed.

Karma Sparks are still being debugged across some page contexts. If counts don’t update immediately or the button behaves unexpectedly on certain pages, we’re tracking it.

TapasAI, the persistent AI journal companion, received a round of fixes and a major new feature this week. The Temporal Lenses — Smṛti (past), Sthiti (present), and Saṅkalpa (future) — are now accessible from a unified panel on entry pages, combining TapasAI chat and temporal reflection prompts in one place. Panel positioning has been fixed so the TapasAI dropdown anchors correctly to its Sparkbar button. And several OpenAI API compatibility issues (vector store configuration, conversation context parameters) have been resolved so conversations flow without interruption.

TapasAI is still being actively debugged. Conversation persistence, Temporal Lens prompt generation, and panel behavior on mobile are all areas we’re refining. If TapasAI feels inconsistent or a conversation doesn’t pick up where you left off, we’re on it.

All sign-in and account creation now flows through Memberful — a single, unified authentication path. Previously, free Starter accounts used a separate auth system while paid tiers went through Memberful OAuth, which meant two different login systems, two different account creation flows, and edge cases where the same email could end up with mismatched account IDs.

Now every practitioner — free or paid — registers and signs in through Memberful. The login page has been simplified to a single “Sign In” button and a “Create a Free Account” link, both routing through Memberful. The signup page has been redesigned as a tier comparison landing page showing what’s included at each level, with all calls to action pointing to Memberful registration.

AI Sparks now work directly on chapter, pāda, and glossary listing pages — no need to open a verse, sūtra, or term first.

On chapter and pāda pages, the AI Sparks button in the Compact Sparkbar generates inline Text AI Sparks (with all four annotation types: Intertextual Connection, Sanskrit Illumination, Lineage Lens, and Teaching Bridge) and Link AI Sparks right below the verse or sūtra card.

On the glossary listing page, each term’s AI Sparks button offers Topical Guide, Tension Analysis, and Link AI Sparks — all rendering inline with save/unsave persistence that syncs with the detail page. Teacher tier required.

The “Generate Smart Prompt” feature has been renamed to “Journal with Tapas AI” across the entire Sparkbar. The button label, panel heading (“Prompts from Tapas AI”), and loading text (“Tapas AI is thinking…”) now make it clear that your personalized journal prompts come from TapasAI — the same companion who reads your journal and knows your practice.

The feature itself is unchanged: tap the button on any verse, sūtra, or glossary term to receive 2–3 reflection prompts grounded in that text and your recent journal entries. Aspirant tier and above.

The Sparkmarks emoji picker has been redesigned from a flat grid to a categorized layout with seven groups: Sparklights (highlight colors), Reactions, Moods, Gestures, Body & Senses, Practice, and Progress.

The same organized picker now appears on both list pages (chapter, pāda, glossary) and detail pages, making it easier to find the right emoji when you’re bookmarking quickly.

A new Frequently Asked Questions page answers 20 questions across seven sections — from “What is HYP Yoga?” to “Can I delete my account?” Each section covers a different area: General, Accounts & Membership, Texts & Study, Journal & Sparkmarks, AI Features, Privacy & Data, and Reading & Accessibility. Answers link directly to the relevant features so practitioners can jump straight from the question to the experience.

The features page has been expanded with a TapasAI companion card, a new Sparkbar section explaining the study toolbar, and updated tier cards showing TapasAI’s inclusion at each level. The navigation also now includes TapasAI and Sparkmarks as first-class items.

A batch of fixes and refinements shipped today:

Hero image fix. Private journal images were broken after the storage bucket security hardening — the signed URL endpoint was using an incorrect path. Fixed, and a shared helper was extracted so all six pages that display Hero Images use the same authenticated loading pattern.

Sparkbar render order. The Sparkbar briefly flashed jumbled buttons during page load. Fixed with an opacity fade-in — buttons are hidden until fully assembled, then fade in smoothly.

Listen button gating. The Listen button on Compact Sparkbar list pages now requires a Starter account. Public visitors see a teal signup panel explaining the audio feature.

Pāda list AI Sparks links. The AI Sparks buttons on all four pāda listing pages were navigating to the homepage instead of the sūtra detail page. Fixed.

Sparkbar re-initialization. Calling initSparkUI a second time (after tier data loaded) silently failed, leaving the Sparkbar stuck at the default tier. Fixed by tearing down and rebuilding SparkUI elements on re-init while preserving Sparkmarks state.

Sūtra page audio gating. The sūtra detail page was missing the audio tier check that verse and term pages already had. Added for parity.

Save pill simplification. The floppy disk icon was removed from AI Sparks save buttons site-wide. Now reads “Save” or “✓ Saved” in plain text.

Your data is safer than ever. We completed a comprehensive security audit and shipped nine improvements in a single day — all behind the scenes, nothing you need to do.

Private storage. Hero Images and Voice Sparks are now stored in private, authenticated buckets. Previously, anyone who knew a file’s URL could access it directly. Now every image and audio file is served through an authenticated gateway that verifies your identity before granting access. No journal content is visible to anyone who isn’t logged in.

Supply chain protection. All third-party scripts loaded from CDNs are now pinned to exact versions and verified with Subresource Integrity hashes. If a CDN were ever compromised, your browser would refuse to load the tampered file.

Webhook verification. Subscription changes from Memberful are now cryptographically verified before they can update your account. Unsigned or tampered requests are rejected.

Fail-closed rate limiting. If the system that tracks daily AI message counts ever encounters an error, it now blocks requests instead of allowing them through. This applies to all four AI services: SwamiGPT, DharmaGPT, TapasAI, and AI Sparks.

Additional hardening. Removed an unused CDN source from the Content Security Policy. Patched two potential cross-site scripting vectors in journal entry cards. Standardized cache-busting across all script references to ensure you always load the latest code.

Journal entries now have likes. Every entry card and entry page shows a ripple icon with a running count — tap it to show appreciation for a fellow practitioner’s reflection.

Karma Sparks appear everywhere entries do: the journal feed, verse and sūtra pages, practitioner profiles, and the subscriptions feed. Available to all practitioners with a free Starter account.

Highlight text as you study. Select any passage on a verse, sūtra, or glossary term page, open the Sparkmarks panel, and choose from six colors: 🟥 🟧 🟨 🟩 🟦 🟪. Your highlights are saved and restored every time you return to the page.

Available to all practitioners with a free Starter account. Manage your highlights from the Sparkmarks flame panel — each one shows a preview and can be removed with a single tap.

The entire Sparkbar is now keyboard-accessible. Press F for Sparkmarks, A for AI Sparks, J for Journal, V for Voice Sparks, L for Listen, T for TapasAI, S for SwamiGPT, D for DharmaGPT, and Esc to close any open panel.

On verse, sūtra, and glossary term pages, the left and right arrow keys now navigate to the previous and next text — so you can read through an entire chapter, pāda, or glossary without touching the mouse.

All shortcuts are automatically disabled when you’re typing in a text field, search box, or the journal composer.

Sanskrit terms on verse and sūtra pages are now highlighted for all readers — including visitors without an account. Tapping a highlighted term shows a warm invitation to create a free account and unlock Access Quick Terms, with a direct link to sign up or log in.

Practitioners with a Starter account or above continue to see full inline definitions as before.

When a visitor without an account taps any Sparkbar button, they now see a benefit-focused panel explaining what that feature does for the specific verse, sūtra, or term they’re reading — with a clear call to create a free account and a log-in link for returning practitioners.

Each panel matches its feature’s color: gold for Sparkmarks, blue for Journal, teal for Listen, and the companion colors for SwamiGPT and DharmaGPT.

Every verse and sūtra card on the chapter and pāda listing pages now carries its own compact Sparkbar — a single metallic pill with three buttons: 🔥 Flame for Sparkmarks, ✏️ Pencil for Journal, and ▶️ Play for Listen.

The buttons use a visionOS-inspired metallic sheen — brushed-metal gradients with a white top highlight and color-coded shadows for gold, blue, and teal. Sparkmark emoji now appear inline in the pill, so you can see your bookmarks at a glance while browsing a chapter or pāda without opening each text.

Cards also received a visual refresh: minimal dividers replace the old card borders, and verse/sūtra numbers display as gold pill badges for a cleaner reading flow.

SwamiGPT and DharmaGPT have been rebuilt from the ground up with distinct personalities, richer responses, and a much better conversation experience.

Deeper personalities. SwamiGPT now teaches in the traditional guru-śiṣya voice — citing specific verses in IAST, tracing cross-text connections, and meeting you with the precision the tradition demands. DharmaGPT bridges classical wisdom and contemporary practice with warm metaphors, practical guidance, and insights drawn from real yoga teachers.

Context-aware. The companions now know what you’re studying. Open the chat from a verse, sūtra, or glossary term page and the AI sees the full text — Sanskrit, translation, and reference — so its responses are grounded in what’s on screen.

Conversation memory. Your conversation persists as you navigate the site, so you can ask a question on one page and follow up on another. Conversations are saved locally for seven days.

Profile images. Both companions now have unique visual identities — SwamiGPT’s warm amber meditating figure and DharmaGPT’s sage green flowing form appear in the chat welcome screen, the AI hub, and their individual study pages.

Markdown rendering. Responses now render with proper formatting — bold, italic, code blocks, lists, blockquotes, and links — making longer explanations much easier to read.

Sanskrit transliteration characters — the macrons, underdots, and overdots in words like prāṇāyāma, kuṇḍalinī, and suṣumnā — now render in the correct typeface across the entire site. Previously, these characters fell back to a mismatched system font mid-word, causing visible jumps in stroke weight and style. All 12 IAST diacritical characters (ā ī ū ṇ ḍ ṣ ṭ ḥ ṛ ṁ ṅ ś) now display consistently in every font we offer — including all nine accessibility fonts in the font selector. Fonts that don’t natively include underdotted glyphs seamlessly fall back to Noto Sans or Noto Serif for just those characters, so the reading experience stays smooth regardless of which font a practitioner chooses.

Fixed a layout issue on mobile where the site logo and navigation links could collide on narrow screens, making them difficult to tap.

AI Sparks across the platform now draw from a curated yoga library — 2,294 real passages from 23 scholarly texts by Iyengar, Muktibodhananda, Vivekananda, Svātmārāma, and others, indexed across 157 glossary terms. Instead of annotating from general knowledge, every AI Spark can now cite specific scholarship, quote from actual source material, and link to the referenced text.

This grounds all three AI Sparks surfaces: Text AI Sparks on verse and sūtra pages now cite real commentary and quote from sources in Lineage Lens and Intertextual Connection annotations. Term AI Sparks on glossary pages receive a term’s actual scholarly passages, so both Topical Guide and Tension Analysis are rooted in specific excerpts rather than generalities. And Journal AI Sparks gain a fifth annotation type — Source (amber, 📖) — that grounds a practitioner’s own words in specific scholarship when the entry is linked to a glossary term.

Link my AI Sparks on glossary term pages now weaves three threads together: the text being studied, the practitioner’s journal reflections and saved annotations, and real scholarly passages — integrating personal practice, classical text, and academic tradition into a single synthesis.

Study actions on verse, sūtra, and term pages now use a two-tier progressive disclosure layout. The primary row — Ask SwamiGPT, Ask DharmaGPT, and Journal — is always visible. A “Study tools” toggle reveals the advanced features: AI Sparks, Generate Journal Prompt, and Link my AI Sparks. On glossary term pages, Topical Guide and Tension Analysis sit alongside Link AI Sparks on a dedicated row. The layout keeps the most-used actions immediate while making the full toolkit one tap away.

Added LDS Scriptures (Book of Mormon, Doctrine and Covenants, Pearl of Great Price) as a source tradition under the Christian group in the Text AI Sparks source selector. The platform now offers 22 source traditions across 7 tradition groups for intertextual exploration.

Sanskrit terms in journal entries are now auto-linked with Quick Terms definitions — the same inline study feature available on verse and sūtra pages. The matching is diacritic-agnostic, so typing “asana” on a Western keyboard links to the āsana glossary entry, “pranayama” links to prāṇāyāma, and so on. Available to all Starter+ members.

A new Features page with a complete guide to the platform — membership tiers, AI Companions, the full AI Sparks system, content inventory, and privacy details. Cross-linked with the changelog and accessible from the site footer.

A comprehensive privacy overhaul to meet GDPR requirements. All Google Fonts are now self-hosted — 64 font files served directly from HYP Yoga, eliminating external requests that transmitted visitor IP addresses to third parties. The privacy policy was completely rewritten with explicit lawful basis for all processing, all seven EU data subject rights, data retention periods, international transfer disclosures, and third-party processors identified by name and country.

A dismissible privacy notice banner now appears on every page. AI Companions display a consent notice on first use, explaining that messages are processed by OpenAI and Anthropic — input is disabled until acknowledged, and any pending prompt fires automatically after consent.

A comprehensive security audit and hardening pass across the entire platform. Journal entries now require authentication to view — even public entries are only visible to logged-in practitioners. All server-side API endpoints are now locked to HYP Yoga origin requests only (CORS). Webhook endpoints verify cryptographic signatures. AI reflection and transcription endpoints now enforce daily usage limits to prevent abuse. Display names and usernames are sanitized throughout to prevent cross-site scripting. The developer tier-override tool now verifies superuser identity server-side before activating.

Round 2: Added security headers across all pages — Content Security Policy (CSP), clickjacking protection, MIME sniffing prevention, HSTS for enforced HTTPS, and restricted browser permissions. Hero Image and Voice Spark uploads now go through server-side validation that verifies authentication, membership tier, file type, and file size before allowing any upload to proceed.

A new Frequently Asked Questions page covering the full range of the platform — from “What is HYP Yoga?” to details on AI Sparks, Voice Sparks, and membership tiers. Twenty questions across seven sections with an accordion layout and the same floating section nav used on the About page.

A new font customization system designed for comfortable, extended reading. Practitioners can choose from 15 carefully selected typefaces — each chosen for proven legibility, cross-platform support, and accessibility research, including Atkinson Hyperlegible from the Braille Institute and Lexend from Google Design. Two layers: a global preference saved to your profile (applied site-wide via settings), and a per-page curated selector on verse, sūtra, and term pages for session-only experimentation. Requires a free account.

Two new AI-powered study tools for deeper engagement with the texts. Journal with TapasAI (Aspirant+) creates 2–3 personalized reflection prompts based on the verse, sūtra, or term you’re reading — informed by your own journal history. Tap a prompt card to begin composing. Link my AI Sparks (Teacher) finds threads between the text you’re studying and the AI Sparks annotations saved across your journal — surfacing connections, synthesis, and a new reflection prompt. Both live on verse, sūtra, and term pages.

A gold card on the journal page that presents a verse or sūtra you haven’t journaled about yet, paired with a contemplation seed generated by AI. The selection rotates daily based on your user ID and is cached for the session. Available to Aspirant and Teacher members.

AI Sparks bring annotation-style reflections to your journal entries. Tap “Add AI Sparks” on an entry and the AI reads your writing, returning color-coded highlights — gold for Textual Connections to classical texts, purple for Insights, green for Deeper Connections, and blue for Personal Connections to your other journal entries. Each annotation shows a popover on tap. Sparks can be saved to your entry and included in exports.

The Reflect feature offers a multi-turn conversation about your entry — each follow-up builds on the previous exchange instead of starting fresh. Both require Teacher tier.

A new way to add a personal layer to your journal: record a 20-second audio reflection on any entry, tagged with an emoji to capture the mood or energy of the moment. Voice Sparks are stored privately and can be played back from the entry page. Teacher-tier practitioners can opt into AI transcription (powered by OpenAI Whisper) — either at upload or on-demand from the entry page.

Journal entries now support Hero Images — upload a cover image (JPEG, PNG, or WebP, up to 5MB) to set the visual tone for your reflection. Requires Aspirant tier. Separately, journal entries are now private by default. The public toggle on the compose page starts unchecked — you choose when to share.

Follow other practitioners’ public journals. When you subscribe to someone’s profile, their new public entries appear on your subscribed feed at /journal/subscribed. Subscribing requires Aspirant tier. User blocking is also now available from profile pages — blocking is bidirectional. A new Notifications tab in settings lets you opt into email alerts when someone subscribes to your journal.

The Glossary and individual term pages now require a free account to view — matching the original feature specification. Quick Terms (inline definitions for highlighted Sanskrit terms on verse and sūtra pages) also require authentication. Public visitors see a prompt to sign up. The Glossary link is hidden from the navigation for logged-out visitors.

Teacher-tier practitioners now access AI Companions powered by Anthropic’s Claude model instead of GPT-4o-mini. The upgrade is seamless — the same SwamiGPT and DharmaGPT personalities, the same inline chat, but with a more capable model underneath. Starter and Aspirant tiers continue using GPT.

A new site-wide footer component with a three-column layout (Texts, Study, Platform) now loads on every page. A dedicated sitemap page shows clickable number grids for all 389 HYP verses, 196 Yoga Sūtras, and 172 glossary terms.

Sparkmarks are a new way to bookmark and organize the verses, sūtras, and glossary terms that resonate with your practice. Hover over any verse or sūtra card and tap the ✦ button to tag it with up to three emoji — use ❤️ for favorites, 🔥 for challenging passages, 🧘 for practice cues, or any combination that feels right. Your sparked content is collected on a dedicated Sparkmarks page (also in the Study menu) where you can filter by emoji or content type. Sparkmarks require a free account — sign up to start sparking.

A deep QA pass across the entire site caught and fixed a set of bugs affecting audio playback, navigation, script loading, and internal links — particularly on pages served at clean URLs like /hyp/3 or /glossary/mokṣa. Audio playback and AI chat (SwamiGPT & DharmaGPT) are currently free for all registered users — just create an account to start listening and asking questions.

Users can now export their journal reflections as formatted PDFs or Markdown files, with control over which text layers to include. The AI chat sidebar now works on glossary and term pages too. Source references across the site link to Amazon.

Two AI companions — SwamiGPT (scholarly) and DharmaGPT (practice-oriented) — joined the site. Both are trained on the classical yoga texts and available as an inline chat sidebar from any verse, sūtra, or glossary term.

A second classical text. All 196 sūtras across four pādas, with the same three-layer display as the HYP — Devanagari, IAST transliteration, and English translation.

HYP Yoga went live as a free, open resource for studying the Hatha Yoga Pradipika — the foundational 15th-century text of haṭha yoga. The complete text across four chapters, a Sanskrit glossary, user accounts, and a personal practice journal shipped on day one.