Anti-Noise Funnel
Regex firewall + semantic filtering block 90% of noise, focusing only on high-value signals.
INBOUNDSIGHT
InboundSight
AI-Driven Strategic Intelligence Radar
Full-cycle intelligence system for inbound tourism. Monitors 10+ social media sources to mine product demands and travel pain points. Transforms fragmented noise into actionable business insights through multi-layer signal filtering.
90%+
Noise Filtered
10+
Social Sources
99.9%
Accuracy
SIGNAL INTELLIGENCE
From fragmented noise to executable intelligence
InboundSight filters, refines, and translates unstructured complaints, demands, and pain points from social and external sources into structured intelligence for product, ops, and strategy teams.
Role-Playing Recon
Dynamic persona switching (Spy/Tech Forensic) for deep vertical analysis of rivals & bugs.
Demand Alchemy
Turning unstructured user complaints directly into structured Product Requirement Documents (PRDs).
Neuro-Symbolic Fix
AI for flexibility, Code for stability. Hard-coded logic prevents classification hallucinations.
AI TECHNIQUES
Four-layer intelligence pipeline
From noise-canceling, business analysis, and Feishu sync to Text-to-SQL QA, each layer is a hybrid AI-plus-code mechanism tuned for inbound tourism intelligence.
01
Layer 1: Noise-Canceling
Three-Stage Filtering Pipeline
Regex blocking -> Hash deduplication -> AI Judge. 90% of noise is physically blocked.
02
Layer 2: The Analyst
Business Logic Analyst
Reasoning, not just summarizing. Auto-switches to 'Competitor Mode' for 'Agoda' or 'Risk Mode' for 'Scam'.
03
Layer 3: The Action
Feishu Sync Engine
Smart Sync: Creates tickets for new demands, updates heat for recurring issues. Includes source & tags.
04
Layer 4: The Oracle
Text-to-SQL QA Engine
No vector fuzzy matching. Converts 'Top 3 complaints' into complex SQL aggregation queries directly.
DEEP DIVE FLOW
Market signal processing system
Noise-canceling -> Analyst -> Feishu sync -> Oracle QA. This is the real production loop behind InboundSight.
InboundSight Intelligence Grid
Market Signal Processing System
Noise-Canceling -> Analyst -> Feishu Sync -> Oracle QA.
Layer 1
Noise-Canceling Pipeline
Regex Gatekeeper + Silent Updater + AI Judge. Blocks 90% noise.
Regex Gatekeeperquick_relevance_check
Silent Updatercheck_content_exists
Layer 2
The Analyst Agent
Context-Aware + Self-Correction. AI that knows business.
Context-Aware Analysisswitch_system_prompt
Self-Correctionvalidate_and_fix
Layer 3
The Action Trigger
Smart Sync Strategy. New tickets for demands, hot updates for issues.
Smart Sync Strategynew_demand / hot_update
Rich Context Injectionsource_url / tags
Layer 4
The Oracle QA System
Text-to-SQL Dynamic Query. No fuzzy vector matching.
Intent Recognitiontrend / compare / list
Dynamic SQL GenGROUP BY / ORDER BY
90%
Noise Blocked
Action
Feishu Sync
SQL
Native Query