Taste-intent understanding
Recognizes compound intent like “for commuting, goes with a trench coat, no obvious logos” — aesthetics and scene included.
CONSUMER DECISION AGENT
Consumer Decision Agent, with you through a high-quality decision
Start from one natural-language request: it understands taste intent, compares candidates, explains its reasons, and closes the loop from discovery to decision in real consumption contexts.
Taste search entryExplainable decisionsCross-cultural fit
Not just search — it carries the decision for you
Search engines return endless links; recommendation feeds return “you might like this”. Neither takes responsibility for the decision.
The Decision Agent enters through the taste search engine and reasons on the Taste Knowledge Graph, aiming to finish the final mile: a compared, explainable, clear recommendation under real constraints.
- Understand: recover scene, aesthetic leanings, budget and hidden constraints from conversation
- Compare: organize candidates along graph relations, ranked by taste fit rather than sales
- Explain: reasons, trade-offs and alternatives — users keep the final judgment
- Act: connect lists, price comparison, reminders and purchase paths to close the loop
Anatomy
How the Agent makes good decisions
Structured comparison
Candidates across price bands, brands and styles are compared within one taste coordinate system.
Cultural-context fit
Understands the tacit requirements of gifting, festivals, workplaces and regional culture.
Explainable recommendation
Every suggestion carries reasons and evidence — no black-box “you might like this”.
Decision-path planning
Complex decisions break into traceable steps: clarify, filter, compare, decide.
Long-term taste memory
With user consent, taste preferences accumulate — it gets sharper over time instead of starting from zero.
Decision flow
How a high-quality decision happens
Express and clarify
State the need in natural language; the Agent proactively clarifies key constraints and taste leanings.
Taste reasoning
It expands candidates over the knowledge graph, reads cultural context and compares across dimensions.
Decide and act
A clear recommendation with reasons and alternatives; next steps follow, and feedback refines the result.
Taste Search: search by taste, not keywords
When a need like “French-vintage style, fits a small apartment, budget ¥3,000” cannot be understood by keyword search, taste search offers a new entry point.
It maps the aesthetic and scene language in natural language onto the Taste Knowledge Graph, returning not web links but understood products, brands and option sets ready for the Agent to reason on.
- Scene-level expression: colloquial, compound and cross-language requests
- Semantic results: ranked by taste fit rather than keyword matching
- Agent-native: results are structured objects the Agent can reason with directly
FAQ
About the Consumer Decision Agent
How is this different from a general chatbot?
General chat relies on parametric language knowledge and easily errs on concrete products. This Agent is grounded in the Taste Knowledge Graph: candidates and relations are traceable, and its goal is completing a decision, not generating an answer.
Will the Agent place orders automatically?
No. It offers recommendations, reasons and alternatives; the final decision belongs to the user. Any transactional action is executed only after explicit user confirmation.
How are my preferences and privacy protected?
Taste memory is enabled only with consent and can be viewed, exported or cleared at any time. Personal data is strictly isolated from brand-side insights, which only receive aggregated, de-identified trend information.
Which categories and markets are supported?
Priority categories are apparel, beauty, home, food and beverage, electronics and lifestyle. Localized decisions support Chinese Mainland, HK/Macao/Taiwan, Japan, Korea and Southeast Asia.
Entrust your next decision to taste
Start with one request and get a compared, explainable, high-quality choice.