Research Without Leaving the Book.
A full browser lives inside Ishvana. You read, bookmark, analyze, and extract entities, and the things you save flow straight into your Legendry as real references. No more forty Chrome tabs you're afraid to close.
Built-in Browser
Research without context-switching. A full Electron-powered browser lives inside Ishvana so you can read articles, explore references, and bookmark sources without leaving your writing environment.
Embedded Browser
Full web browser inside Ishvana. Navigate, read, bookmark without context-switching. Back/forward/reload, URL bar with Ctrl+L, and a dashboard home page with quick-launch cards.
Smart Content Capture
Auto-extracts main content on every page load. Priority cascade: article → main → largest text-heavy element → full body. Up to 50,000 characters captured per page.
Ad Blocker
Built-in ad and tracking pixel blocker. Toggle on/off per session. Browse research sources without visual clutter or tracking.
Dashboard Start Page
Landing page with bookmark stats and quick-launch cards: Search the Web, Resume Last Trail, Review Saved Evidence, Ask Lagan, and Transcribe Video.
Keyboard Shortcuts
Ctrl+L focus URL bar, Ctrl+D bookmark current page, Ctrl+F focus search in active tab, Escape return to dashboard.
Bookmarks That Actually Remember the Page
Every bookmark is a structured research artifact. On save, Ishvana captures the content, generates tags, summarizes the page, pulls out named entities, and scores it for relevance to your active project. Six months later you can still find the article, because the whole thing is actually searchable.
Auto-generated tags, project filters, search, and inline actions — every bookmark is a structured research artifact.
- One-Click Bookmarking
- Bookmark any page with one click (Ctrl+D). Toggle analysis on save to auto-generate tags, a structured summary, and smart suggestions linking content to your project.
- Entity Extraction
- Named entities — people, places, concepts, works — get pulled from bookmarked content and presented as tags. Link entities directly to your Legendry entries.
- Project & Tag Organization
- Filter bookmarks by project, tag, or search text. Sort by date, project, or alphabetically. Every bookmark tracks URL, title, content excerpt, category, project, user tags, auto-generated tags, and timestamps.
- Project Relevance Scoring
- Every bookmark gets scored in the context of your active project. A page about medieval trade routes gets flagged as relevant if your world has similar economic systems.
- Bookmark Detail
- Full detail view with embedded analysis, edit capabilities, send-to-Legendry action, and right-click context menu: view, open externally, copy URL, re-analyze, or delete.
Perception Engine
A four-tier analysis system that goes well past summarization. Extract entities, map relationships, read sentiment, and trace second-order implications — all calibrated to what your project actually cares about, so the analysis lands on things your book needs instead of generic article-summary fluff.
- One-Click Page Analysis
- Hit "Analyze this page" and get summary, key points, entities, project relevance, reading time, content type classification, and word count. Works on any loaded page.
- 4 Perception Depths
- Choose your analysis level: Surface (fast entities + sentiment), Analytical (entities, relationships, tone, key points), Strategic (patterns, implications, source positioning), Prophetic (second-order effects, trajectory analysis, deepest synthesis).
- Entity Extraction with Confidence
- Every entity gets a confidence score (0–100%) with visual confidence bars. People, places, organizations, and concepts identified, scored, and contextualized.
- Relationship Mapping
- Entity-to-entity connections extracted with relationship types and confidence scores. See how concepts in an article connect to each other and to your world.
- Sentiment & Tone Analysis
- Overall sentiment (positive/neutral/negative), confidence score, tone descriptors (professional, casual, academic), and emotional indicators with per-emotion scores.
- Raw Text Analysis
- Don't need a webpage? Paste any text into the raw text input and run the same perception engine on it. Analyze excerpts, field notes, or early drafts.
Summary, key points, named entities, and project relevance — extracted in one click. Perception depth controls how deep the analysis goes.
Research Pipeline
A seven-stage research orchestration engine. You enter a question, and Lagan searches your knowledge base, consults your lore, gathers specialist insights, and synthesizes everything into a scored, sourced brief you can read in two minutes instead of spending a day building yourself.
Freeform research prompts, lore/specialist toggles, and web monitors all in one panel.
7-Stage Pipeline
Enter a freeform research prompt and Lagan runs: term extraction → RAG query → lore search → specialist consultation → external sources → summary synthesis → relevance scoring. Toggle project lore and specialist agents on/off.
RAG Knowledge Search
Semantic search across your entire ChromaDB knowledge base. Results ranked by relevance with source metadata. Direct queries available outside the full pipeline too.
Deep Intelligence Query
Entity-aware research queries that return structured results. Generate strategic insights at 4 depth levels. Multi-source synthesis identifies agreements, disagreements, and unique insights across sources.
Pipeline Results
Every run produces: extracted research terms, RAG results with scores, matched lore entries, specialist knowledge, external sources, a synthesized context summary, and an overall relevance score.
Semantic Search
Search by meaning. ChromaDB-powered vector search runs across all your collections — documents, bookmarks, lore entries — with saved queries and confidence scoring, so you can find what you remember even when you've forgotten the exact words you used.
- Natural Language Queries
- Ask "articles about resistance movements" and find semantically relevant results even if those exact words don't appear anywhere in the content. The retrieval works on meaning, so synonyms and paraphrases still land.
- Collection Filtering
- Multi-select which ChromaDB collections to search. Search your documents, web bookmarks, lore entries, or everything at once. Narrow or broaden scope per query.
- Saved Searches
- Name and save frequently-used queries for one-click re-execution. Each saved search tracks creation timestamp and last-used date.
- Query History
- Recent queries tracked automatically. One-click re-run any past search. Configurable history limit so your most-used queries are always at hand.
- Results with Confidence
- Every result shows a similarity score (displayed as percentage), content snippet preview, source collection, and metadata. Ranked by relevance, highest first.
Web Monitoring
Track pages over time. Add any URL to a monitor, set a check interval, and get notified when content changes. Automatic snapshots, diff-based detection, and configurable sensitivity.
URL Watch
Add any URL to monitor. Set check interval in minutes and change sensitivity threshold (0.1–1.0). Toggle notifications on/off per monitor.
Automatic Snapshots
Each check stores a baseline snapshot. Changes detected via content diffing with percentage-change calculation against the previous snapshot.
Change Detection
Visual badges show when monitored pages have changed. See total changes detected today. Open changed pages directly in the browser for review.
Monitor Management
Active monitor list with status (active/initializing/error), last-checked timestamp, and controls to re-check now, view changes, or stop monitoring.
Lagan, the Research Copilot
Multi-turn conversation with the research agent. Lagan already has context on your active project, your bookmarks, your search results, and whatever page you've just been analyzing. Ask for source triage, synthesis, strategy, or a sanity check on a claim that seemed too convenient.
Multi-turn research chat with project-aware synthesis, source triage, and structured recommendations.
- Multi-Turn Conversation
- Full chat interface with Lagan, your research copilot. Ask for source triage, synthesis, strategy recommendations, or fact-checking — Lagan has context of your active project, bookmarks, search results, and analysis.
- Tool-Equipped Agent
- Lagan can call specialized tools during conversation — searching your lore, querying knowledge bases, analyzing content. See which tools were used in each response via the tools-used strip.
- Per-Turn Actions
- Copy any response to clipboard. Send assistant responses to Legendry as reference entries. Regenerate responses. Edit-and-resend (truncates history from edit point). Delete individual turns.
- Streaming Responses
- Messages render in real-time as Lagan generates them. Auto-scroll keeps you at the latest response. Full markdown rendering with tables, code blocks, and structured layouts.
YouTube Transcription
Transcribe and analyze video content as research material. Preview metadata, choose transcription sources, track async jobs, and get structured analysis with entity extraction, topics, and project relevance.
Video Preview
Enter a YouTube URL and preview metadata: title, channel, duration, thumbnail, caption availability, and language options. No download required.
Multi-Source Transcription
Three caption sources with automatic fallback: manual captions (creator-made) → auto captions (YouTube-generated) → Whisper transcription (OpenAI model). Best source selected automatically.
Job Tracking
Async processing with real-time status: queued → downloading → transcribing → analyzing → complete. Progress bar, status badges, and error messages. Cancel or delete any time.
Timestamped Transcript
Full transcript with clickable timestamped segments. Search within transcript text. Word count and duration metadata displayed.
Video Analysis
Summary, entity extraction with confidence, topic identification, key quotes, project relevance scoring, and chunk storage in vector DB for later semantic search.
Integration & Export
Connect research to external sources and flow findings directly into your world. Content enrichment, external API fetching, and one-click export to Legendry from anywhere in the module.
- Content Enrichment
- Blend page content with Wikipedia, academic databases, and custom sources. Get enriched entities with confidence scores and source attribution. Configure max items per source.
- External API Fetcher
- Fetch data from any HTTP endpoint (GET or POST). Preview responses, parse JSON, and integrate external data into your research workflow.
- Send to Legendry
- One-click export from the Extract tab, Lagan chat, or bookmark analysis. Creates a reference lore entry with title, content, and source URL automatically populated. Research flows directly into your world.
- Content Extraction
- Review captured page content in the Extract tab. See word count, truncation status (50k character limit), copy to clipboard, or send directly to your lore database. Raw content always accessible.