NLP Tools · Veritas

Purpose#

The Veritas NLP Tools — a full-screen overlay that takes raw text or a URL, runs a local simulated NLPAnalysis, exposes fixture sample analyses, and renders collapsible sections for sentiment, entities, readability, language complexity, propaganda techniques, logical fallacies, and keywords.

Entry points#

• /domains/veritas index — verify the internal-namespace tool grid links here
• Linked from /domains/veritas/articles — article analysis is a sibling workflow; no direct handoff from this route is verified here
• Direct URL / bookmark — yes (auth required)

Layout regions#

page.tsx is a 'use client' component that renders an sr-only <h1> ("NLP Tools · Veritas · OSHUN") then mounts <VeritasNLPTools onClose={() => router.back()} />. Overlay is fixed full-screen.

• Header: title "NLP Tools", 44px-minimum Close button (aria-label="Close NLP Tools"), responsive wrap behavior, and tablist for input | results | samples.
• Nav tablist: role="tab" buttons expose aria-selected; the results tab is disabled until a result exists.
• Body: focusable scrollable region labelled "NLP tools workspace"; current view is driven by view: NLPView.
  • Input (input) — text/URL mode toggle, textarea or URL field, live word count for text mode, Analyze action, and fixture sample cards
  • Results (results) — populated NLPAnalysis overview and collapsible analysis sections
  • Samples (samples) — dedicated fixture sample library, each keyboard selectable

State drivers: view, inputMode, textInput, urlInput, analyzing, result, and expandedSections.

States#

• Loading — overlay renders synchronously; no async skeleton.
• Input view, text mode — default mode renders textarea, live word count, disabled Analyze action, and sample cards.
• Input view, URL mode — URL field renders, Enter starts analysis, and result carries inputUrl.
• Analyzing — local simulated setTimeout sets analyzing=true, changes the button label to "Analyzing...", then switches to results.
• Results view, populated — custom text/URL results and sample fixture results render from result.
• Results tab disabled when no result — verified via disabled tab state.
• Samples view — dedicated sample library is reachable by tab and cards load fixture results by click/Enter/Space.
• No propaganda / no fallacy branches — factual fixture renders empty "No ... detected" states.
• Propaganda / fallacy populated branches — manipulative fixture renders every technique and fallacy row from SAMPLE_NLP_ANALYSES.
• Offline — URL-mode analysis and Samples navigation remain usable after context.setOffline(true) because no analysis endpoint is fetched.
• Standalone PWA — /?surface=pwa restores to /domains/veritas/nlp and matches standalone display-mode.
• Reduced motion — E2E emulates reduced motion and verifies overview, metric-fill, and sentiment segment transitions/animations are effectively zero.
• Serious accessibility scan — route passes the automatic serious axe gate; sentiment segment text now uses contrast-safe readableInk.

Interactions#

Header#

• Close NLP Tools (button, aria-label="Close NLP Tools", X icon)
  • Function: keyboard Enter calls onClose() → router.back()
• Nav tablist:
  • input — returns to the input workspace
  • results — disabled until result exists; then shows current result
  • samples — opens the dedicated sample library

Input view#

• Input mode toggle — switches inputMode between text and url.
• Text input — textarea accepts pasted text and displays a live token count.
• URL input — text input accepts a URL; pressing Enter runs analysis.
• Analyze — disabled until the active input has content; click/Enter starts local analysis, sets result, and switches to results.
• Input sample card — click/Enter/Space loads the fixture sample into result and switches to results.

Samples view#

• Sample card — click/Enter/Space calls handleLoadSample(sample).
• Enter custom text — returns to the input workspace.

Results view#

• New Analysis — returns to input without clearing the last result.
• Overview — sentiment, word/sentence count, detected language, readability grade, text preview, and overall assessment render.
• Section header — click toggles the section id in expandedSections.
• Sentiment Spectrum — negative/neutral/positive segments and compound marker render; segment contrast and reduced motion are verified.
• Entity Recognition — grouped entities render with category and confidence band labels.
• Readability Metrics — Flesch, Fog, Coleman-Liau, SMOG, ARI, grade level, average sentence length, and syllables-per-word render.
• Language Complexity — vocabulary, lexical diversity, passive voice, sentence complexity, technical terms, jargon density, nominalizations, and hedging metrics render.
• Propaganda Technique Detection — empty and populated branches are verified against fixture data.
• Logical Fallacy Identification — empty and populated branches are verified against fixture data.
• Top Keywords — one chip per fixture keyword renders with frequency and TF-IDF hooks.

Current E2E evidence#

• apps/oshun/web/e2e/veritas-nlp-tools.spec.ts — renders input defaults, runs a custom text analysis, toggles sections, and closes by keyboard covers signed-in navigation from /library, default text input state, disabled results tab, sample count, live word count, Analyze disabled/enabled states, analyzing label, populated custom result, all default-open sections, factual empty propaganda/fallacy branches, entity/keyword counts, section collapse/reopen, New Analysis, and keyboard close/back.
• apps/oshun/web/e2e/veritas-nlp-tools.spec.ts — loads a fixture from Samples by keyboard and verifies manipulative-language results with reduced motion covers the dedicated Samples tab, keyboard sample loading, manipulative fixture result id, propaganda/fallacy counts and row text, reduced-motion overview/metric/sentiment transitions, and the serious axe contrast scan.
• apps/oshun/web/e2e/veritas-nlp-tools.spec.ts — runs URL-mode analysis while offline in standalone mobile without horizontal overflow covers standalone PWA restoration, offline URL-mode analysis via Enter, inputUrl result state, text preview for fetched-content placeholder, offline Samples navigation, mobile no-horizontal-overflow checks, and standalone display-mode.

Data & contracts#

• Reads: SAMPLE_NLP_ANALYSES from @/lib/veritas/veritas-extended-simulation.
• Writes: none externally; all analysis, sample loading, and section expansion is local component state.
• Realtime: none.
• Caching: not applicable.
• Auth/role check: shell middleware (verify).
• Telemetry: no track* imports in this file.
• Page metadata: not set (client component); sr-only <h1> reads "NLP Tools · Veritas · OSHUN".

Cross-references#

• Component: apps/oshun/web/src/components/domains/veritas/VeritasNLPTools.tsx
• Type: NLPView = 'input' | 'results' | 'samples'
• Route E2E: apps/oshun/web/e2e/veritas-nlp-tools.spec.ts
• Sibling tools:
  • domains-veritas-articles.md — full article analyzer
  • domains-veritas-knowledge-graph.md, domains-veritas-research.md
• Customer-facing Veritas hub: veritas.md
• Feature spec: V1/features.md

Open questions / known gaps#

• Analysis pipeline is local/simulated; verify whether V1 wires to a Sophia or Veritas backend NLP service.
• No telemetry on mode switch, analysis submission, sample loading, or section expansion.
• Entry links from /domains/veritas, /domains/veritas/articles, and /veritas are not covered by this route-specific E2E; keep or add coverage in their route specs.
• Live screen-reader and touch-device AT passes remain pending beyond the automated keyboard/Playwright coverage here.