The Verdant Intelligence Platform senses every node of the chain, decides what to grow and where it sells, and orchestrates the flow from plot to plate — in real time.青源智能平台是一套面向农业供应链的 AI 操作系统——六大模块感知每一个节点、决策种什么，并实时编排从田间到餐桌的全程流转。
Demand matched需求匹配率
Active plots活跃地块
Forecast accuracy预测准确率
Traceable batches可溯源批次
Demand vs. planting需求 vs 种植
Planting allocation种植配额
to demand cap · MILP对齐需求上限 · MILP
Quality & traceability质量与溯源
CV grading · batch provenanceCV 分级 · 批次溯源
Product interface shown is illustrative of the platform’s capabilities.所示产品界面用于示意平台能力。
The platform turns a fragmented chain into a single feedback loop: it senses the plot, forecasts the market, decides what to grow, coordinates every tier, traces every batch, and closes the loop at the point of sale.平台把碎片化的链条转化为单一反馈闭环：感知地块、预测市场、决策种什么、协调各级主体、追溯每一批次，并在销售端完成闭环。
Sense
Forecast
Decide
Coordinate
Trace
Sell
Each module is a discipline of its own — perception, matching, forecasting, traceability, orchestration and demand. Together they form the Verdant Intelligence Platform.每个模块自成一门学科——感知、撮合、预测、溯源、编排与需求。它们共同构成青源智能平台。
VerdantSense
Unifies climate, soil, agronomy, price history, live demand and remote-sensing data, with computer-vision perception of crop, disease, yield and grading from satellite and UAV imagery.
VerdantMatch
Retrieves the proven crop mix of the most similar villages (Case-Based Reasoning) and allocates planting to the demand cap with a mixed-integer program — plant to demand, never overproduce.
VerdantForecast
Machine-learning models forecast demand and price on live order data, so planting decisions track where the market is heading, not where it has been.
VerdantTrace
Computer-vision grading and batch-level traceability give every product a verifiable provenance — from the plot it grew on to the buyer who receives it.
VerdantFlow
A multi-agent system coordinates farmer, aggregator, buyer and quality agents across the chain — each node senses and sends data, the platform queries the engine, and the decision is routed back.
VerdantMarket
A B2B2C network linking verified, traceable supply to paying demand — national and urban buyers, e-commerce and live-commerce channels, and corporate off-take.
The platform is geospatial at its core. Satellite and UAV imagery, soil, climate and demand are fused on a live field map — so suitability, crop assignment and traceability are resolved plot by plot, not village by village.平台以地理空间为核心。卫星与无人机影像、土壤、气候与需求在实时田块地图上融合——适宜性、作物分配与溯源因此精确到每一地块，而非以村为单位。
Layers图层
Plot A-12
AIt senses every node of the chain, queries a decision engine, and routes the answer back — in real time.
Every node senses and streams its state to the platform; the platform queries the decision engine and routes the answer back. The links run both ways, in real time — supply and demand stay in lockstep.每个节点都在感知并将自身状态实时回传平台；平台调用决策引擎，再把结果路由回去。链路双向贯通、实时运转——供给与需求始终同步。
BROKER
central engine中央引擎
Farmer
农户
Demand
需求
Aggregator
集货
Quality
质量
Perception, retrieval, optimization, forecasting and coordination — chained into a single decision the platform can act on.感知、检索、优化、预测与协调——串联成平台可直接执行的单一决策。
Perception
CVCrop, disease, yield & grading from imagery.
Retrieval
CBR · kNNReuse the proven crop mix of the most similar villages.
Optimization
MILPAllocate planting to the demand cap — never overproduce.
Forecasting
MLForecast demand on live order data.
Coordination
MASOrchestrate every tier of the supply chain.
Demand-driven allocation
MILPmax Σ (E[πⱼ]ρₚⱼ − κₚⱼ) xₚⱼ s.t. Σ xₚⱼ ≤ Aₚ, Σ ρₚⱼxₚⱼ ≤ DⱼPlant to the demand cap — never overproduce. π price · ρ yield · κ cost · A area · D demand.
Plant to the demand cap —按需求上限种植—— never overproduce.绝不过量生产。
solvers:求解器： exact MILP (small) · GA · SA · PSO · Tabu Search (large) · learned policy
Case-Based Reasoning is the moat. Each village we onboard enriches a shared case base — its proven crop mix and outcomes become a warm start for the next similar village. The engine doesn’t copy the way software does; it learns from real ground truth, so accuracy and speed grow with scale.案例推理（CBR）正是护城河。每接入一个村庄，都会丰富共享案例库——其经过验证的作物组合与产出成果，成为下一个相似村庄的热启动起点。引擎并非像软件那样简单复制，而是从真实地面实况中学习，因此精度与速度随规模一同增长。
Same engine, any village
Capital-light — no new model per village.
Case base compounds
Each deployment is faster and cheaper than the last.
10+ provinces → national
Proven replication beyond Yunnan.
A global problem
~200M smallholders in China; ~500M+ worldwide.
The demand network closes the loop: graded, traceable produce flows to e-commerce and live-commerce channels, and to corporate and institutional buyers under contract.需求网络完成闭环：分级、可溯源的农产品流向电商与直播带货渠道，并按合约供给企业及机构买家。
Consumer channels消费渠道
Branded, premium produce reaches national and urban shoppers directly — provenance verified, batch by batch.品牌化的优质农产品直达全国及城市消费者——产地来源逐批验证。
Off-take demand包销需求
National & urban buyers
Beijing · Shanghai · Kunming
E-commerce shoppers
Pinduoduo · Tmall · Douyin
Health-food & traceable-food communities (B2B2C)
Corporate & institutional buyers
Architecture grounded in Case-Based Reasoning & Multi-Agent Systems research for agri-food supply chains (Procedia Computer Science 232, 2024).
拾千农 · 得万村 · Onboard a thousand farmers, reach ten thousand villages.
VerdantSource turns coordination into compounding returns — connecting fragmented supply to real demand, village by village.VerdantSource 青源把协调变成可复利的回报——把分散的供给与真实需求逐村连接起来。