ready
Search Is Finding. Discovery Is Understanding.Position TheoB Discovery Engine as an intelligence refinery, not a generic search clone.
Do not confuse ranked links with trusted understanding.THEOB DISCOVERY ENGINE FOUNDATION LAYER
Search is finding. Discovery is understanding.
TheoB Discovery Engine pursues truth, provides context and action, and turns scattered web results into scored, cited, deduplicated, agent-ready intelligence inside the Vault.
Universal Time Scroll
Every TheoB pathway can move through Past, Present, and Future without losing context.
Present
Read current signals, conditions, and live context.
🎙
Universal Voice Orb
Voice ready
SYSTEM STATE
StableFounder ControlledHuman Reviewed
THEOB DISCOVERY ENGINE
The internet, refined into intelligence.
TheoB Discovery Engine pursues truth, provides context and action, and turns scattered web results into scored, cited, deduplicated, agent-ready intelligence inside the Vault.
From infinite information to living intelligence capsules.
trueFoundation Ready
falseLive Search Enabled
falseVault Ingestion
falseCapsule Compression
Search is finding. Discovery is understanding.
TheoB Discovery Engine pursues truth, provides context and action, and turns scattered web results into scored, cited, deduplicated, agent-ready intelligence inside the Vault.
From infinite information to living intelligence capsules.
Discovery Engine foundation is ready as a non-destructive architecture layer. Live provider retrieval, Vault ingestion, multimodal interpretation, capsule compression, and agent evidence handoff remain disabled until provider access, scoring policy, redaction, persistence, file handling, and source licensing rules are approved.
ready
Pursue TruthPrioritize evidence quality, source transparency, contradiction detection, and context.
Truth pursuit must show uncertainty, not pretend certainty.ready
Provide Context And ActionDiscovery output should help agents and humans understand what matters, why it matters, and what action is appropriate.
No context-free data dumping.ready
Deduplicate Before TrustRepeated copies of the same claim should be clustered before credibility is scored.
Do not mistake repetition for verification.ready
Citations Before ConclusionsEvery useful discovery result should point back to visible source trails.
No black-box truth claims.ready
Vault-Ready IntelligenceDiscovery should produce clean reference objects that can be stored, retrieved, scored, compressed, and used by agents.
Only structured, redacted, source-safe records should enter the Vault.review-required
Provider-Aware RetrievalSearch and data providers must be used through legal, rate-aware, terms-aware access patterns.
No brittle scraping or unlimited-data fantasy wiring.review-required
Agent-Ready EvidenceAgents should receive scored reference cards, extracted claims, contradiction signals, and confidence levels.
Agents should not reason from raw noisy search results.review-required
Multimodal DiscoveryDiscovery must eventually support text, images, diagrams, CAD, engineering schematics, maps, signals, and structured files.
Interpretation must add structure, not hallucination.review-required
Capsule-Ready OutputDiscovery should prepare future records for TheoB Intelligence Capsule Engine compression and reactivation.
Compression must reduce size, not truth.ready
Human-Readable TransparencyHumans should see why a source was included, downgraded, clustered, or rejected.
Trust requires visible reasoning surfaces.ready
Non-Destructive FoundationThe foundation layer defines the architecture without executing live searches or mutating the Vault.
Design first. Wire live retrieval later.planned
general-web-searchRetrieve broad web results from provider-aware search APIs.
Respect provider terms, quotas, caching rules, and attribution requirements.Google Custom SearchBing Web SearchYahoo-compatible search sourcesBrave Search
planned
encyclopedic-knowledgeProvide baseline entity context, definitions, and structured relationships.
Treat encyclopedic sources as context, not final authority.WikipediaWikidataDBpedia
planned
academic-researchRetrieve papers, citations, authors, abstracts, and research trails.
Separate peer-reviewed research from preprints, commentary, and SEO summaries.CrossrefSemantic ScholarOpenAlexPubMed where relevant
planned
government-public-dataUse primary public datasets where available.
Prefer primary data for factual and statistical claims.data.govNOAANASAWorld BankFAOUN datasets
planned
news-current-eventsTrack current developments and compare reporting across outlets.
Avoid treating speed as accuracy.approved news APIspublisher RSSlicensed feeds
planned
domain-specific-vault-feedsFeed TheoB domain intelligence with specialized sources.
Domain sources must be scored, dated, and source-linked.cacao research feedsclimate datasetsagriculture sourcessupply chain references
planned
multimodal-file-sourcesPrepare future discovery for non-text intelligence sources.
Never flatten visual, spatial, or schematic meaning into weak text-only summaries.imagesPDFsCADengineering schematicsarchitectural plansmapsdatasets
Classify the user or agent request into research, comparison, validation, monitoring, visual interpretation, schematic interpretation, or action-support intent.
Output: intent profileChoose which search, data, academic, government, multimodal, or Vault sources should be queried.
Output: provider planCollect results from approved providers through legal and rate-aware interfaces.
Output: raw provider result setCluster duplicate URLs, mirrored articles, repeated claims, syndicated pages, and near-identical summaries.
Output: duplicate clustersScore source authority, primary-source status, author clarity, citation trail, date freshness, and commercial contamination.
Output: source scoreExtract key claims, dates, entities, numbers, and relationships from useful results.
Output: claim cardsPrepare future interpretation for images, diagrams, schematics, CAD, and architectural files.
Output: multimodal observation cardsDetect when sources disagree and label the disagreement without hiding it.
Output: conflict mapCreate structured reference cards that humans can inspect and agents can use.
Output: reference cardsPrepare records for Vault ingestion after redaction, licensing, retention, and persistence rules are approved.
Output: vault-ready recordPrepare reference records for future TheoB Intelligence Capsule Engine compression and reactivation.
Output: capsule-ready intelligence objectHand scored, cited, deduplicated evidence to agents instead of raw search noise.
Output: agent evidence bundlesource authorityprimary-source statuspublication dateretrieval dateauthor identitycitation trailduplicate cluster countclaim consistencyconflict leveldomain reputationcommercial or SEO contaminationhistorical reliabilityvisual interpretabilityschematic interpretabilityagent usabilityVault readinesscapsule readiness
Allowed NowRender Discovery Engine foundation.Define product positioning.Define provider categories.Define discovery pipeline stages.Define scoring dimensions.Define future reference card shape.Define future capsule handoff shape.Keep live search disabled.Keep Vault ingestion disabled.Keep capsule compression disabled.
Not Allowed YetRun live search provider queries.Scrape search engines or websites.Store external content in the Vault.Compress source material into capsules.Process CAD, schematics, images, or datasets automatically.Claim unlimited free provider data.Bypass provider terms, quotas, or attribution rules.Hand raw noisy search results directly to agents.Execute agent actions from discovery results.
queryId: stable discovery query idquery: user or agent query textintent: research/comparison/validation/monitoring/visual-interpretation/schematic-interpretation/action-supportproviderPlan: selected provider categoriestimeSensitivity: static/current/historical/live-monitoringdeduplicationRequired: true/falseconflictDetectionRequired: true/falsevaultIngestionRequested: true/falsecapsuleReadinessRequested: true/falsecreatedAt: ISO timestampproductionMutation: false
referenceId: stable reference idqueryId: linked discovery query idtitle: source titlesourceUrl: canonical source URLprovider: source providersourceType: web/news/academic/government/encyclopedic/vault/domain-feed/image/diagram/schematic/CAD/datasetauthor: safe author label when availablepublishedAt: ISO date or unknownretrievedAt: ISO timestampsummary: short source-grounded summarykeyClaims: array of extracted claim objectsentities: array of safe entity labelsvisualObservations: array for images, diagrams, maps, schematics, or CAD when availableconfidenceScore: 0-100conflictScore: 0-100sourceQualityScore: 0-100duplicateClusterId: linked duplicate cluster idcitationTrail: array of source links or source idsvaultReadiness: not-ready/review-ready/readycapsuleReadiness: not-ready/review-ready/readyagentUsability: low/medium/highredactionStatus: redacted-safe
capsuleCandidateId: stable capsule candidate idcapsuleType: text/image/diagram/schematic/signal/entity/compositesourceTrail: linked reference cards and source IDssummary: compressed meaning without losing source traceabilityentities: safe extracted entitiestimeRange: known time rangeconfidenceScore: 0-100conflictScore: 0-100retrievalTags: array of semantic tagsreactivationModes: expand/compare/diagramify/timelineify/simulate/route-to-agentpreservationRule: A capsule must preserve enough truth to be reawakened faithfully.
Discovery Provider RegistryDiscovery Query Intent ClassifierDiscovery Source Scoring ReadinessDiscovery Deduplication ReadinessDiscovery Vault Ingestion ReadinessTheoB Intelligence Capsule Engine FoundationCapsule Type RegistryImage Capsule Interpretation LayerDiagram And Schematic Capsule ReadinessContextual Design Translation ReadinessVisual Semantics Color Intelligence RegistryQuantum Intelligence Evolution LayerUniversal Intelligence Hub Foundation