【智能体】DIVA物流智能体，由Amazon Bedrock提供支持

语言 Chinese, Simplified

SEO Title

The DIVA logistics agent, powered by Amazon Bedrock

解决方案概述

为了解决现有物流代理的局限性，ShellKode使用亚马逊卧室代理、亚马逊卧室知识库和API集成层构建了一个高级代理助理。

当客户与DIVA 2.0交互时，他们会体验到一个无缝的对话界面，可以自然地理解和响应他们的查询。无论是跟踪包裹、检查运费还是询问服务可用性，用户都可以用自己的话提问，而无需遵循僵化的脚本。DIVA 2.0增强的人工智能功能使其能够理解上下文，管理复杂的请求，并提供准确、个性化的响应，显著改善了整体客户体验，减少了人为干预的需要。以下高级架构图说明了AWS服务的应用程序流程和解决方案架构。

The DTDC logistics agent is designed using a modular and scalable architecture to provide seamless integration and high performance. This streamlined workflow demonstrates how a generative AI-powered serverless logistics agent using AWS App Runner, Amazon Bedrock Agents, AWS Lambda, and a vector-based knowledge base handles user queries ranging from tracking requests to serviceability checks and shipping rates intelligently and efficiently.

The logistics agent is hosted as a static website using Amazon CloudFront and Amazon Simple Storage Service (Amazon S3). The logistics agent is integrated with the DTDC website, which provides an intuitive and user-friendly interface for end-user interactions (see the following screenshot).

An end-user accesses the logistics agent through the DTDC website and submits queries like tracking shipments, checking service availability, calculating shipping rates, FAQs, and so on using natural language.The user requests are processed by App Runner, which helps run the web application (including API services, backend web services, and websites) on AWS. App Runner is hosted with multiple API services, such as the Amazon Bedrock Agents API and Dashboard API. App Runner initiates the Amazon Bedrock Agents API based on the user requests.

Amazon Bedrock is a fully managed service that offers a choice of industry leading foundation models (FMs) along with a broad set of capabilities to build generative AI applications, simplifying development with security, privacy, and responsible AI. With Amazon Bedrock, your content is not used to improve the base models and is not shared with any model providers. Amazon Bedrock Guardrails provides configurable safeguards to help safely build generative AI applications at scale. To learn more, see Build safe and responsible generative AI applications with guardrails. AWS Identity and Access Management (IAM) helps administrators securely control who can be authenticated and authorized to use Amazon Bedrock resources.

The Amazon Bedrock agents are configured in Amazon Bedrock. An Amazon Bedrock agent receives the request and interprets the user’s intent using its natural language understanding capabilities. Based on the interpreted intent, the agent triggers an appropriate Lambda function, such as:

Tracking consignments
Pricing information
Location serviceability check
Support ticket creation

The triggered Lambda function calls the following client APIs, retrieves the relevant data, and returns the response to the agent:

Tracking System API – Retrieves real-time status and provides updates on consignment shipment tracking
Delivery Franchise Location API – Checks the service availability to deliver the parcels between the locations
Pricing System API – Calculates the shipping rates based on shipment details provided by the user
Customer Care API – Creates a support ticket for the end-users

The agent passes the response to the large language model (LLM), in this case Anthropic’s Claude 3.0 on Amazon Bedrock, which understands the context of the retrieved data, processes it, and generates a meaningful response for the user.

The knowledge base contains web-scraped content from the DTDC website, internal support documentation, FAQs, and operational data, enabling real-time updates and accurate responses. The knowledge base contents are stored as vector embeddings in Amazon OpenSearch Service, providing quick and relevant responses. For general queries, the logistics agent fetches information from Amazon Bedrock Knowledge Bases, providing accuracy and relevance. Using semantic similarity search, relevant chunks of information are retrieved from the knowledge base based on the user’s query, which Amazon Bedrock then uses to generate a context-aware response. If no relevant data is found in the knowledge base, a fallback response (preconfigured in the Amazon Bedrock prompt) is returned, indicating that the system couldn’t assist with the request.

The logistics agent queries and associated responses are stored in Amazon Relational Database Service (Amazon RDS) for PostgreSQL for enhanced scalability and relational data handling. App Runners initiates the Dashboard API call to update the queries and associated responses in Amazon RDS. We discuss this in more detail the following section.

Throughout the process, Amazon CloudWatch Logs captures key events such as intent recognition, Lambda invocations, API responses, and fallback triggers for auditing and system monitoring. AWS CloudTrail records and monitors activity in the AWS account, including actions taken by users, roles, or AWS services. It logs these events, which can be used for operational auditing, governance, and compliance.

Amazon GuardDuty is a threat detection service that continuously monitors, analyzes, and processes AWS data sources and logs in your AWS environment. GuardDuty uses threat intelligence feeds, such as lists of malicious IP addresses and domains, file hashes, and ML models to identify suspicious and potentially malicious activity in the AWS environment.

Logistics agent dashboard

The following high-level architecture diagram illustrates the logistics agent dashboard, which captures the end-user interactions and its associated responses.

The logistics agent dashboard is hosted as a static website using CloudFront and Amazon S3. Dashboard access is allowed only for the DTDC admin team.

The dashboard is populated through API calls using Amazon API Gateway with Lambda as a backend, which retrieves the dashboard data from Amazon RDS for PostgreSQL.

The dashboard provides real-time insights into the logistics agent performance, including accuracy, unresolved queries, query categories, session statistics, and user interaction data (see the following screenshot). It provides actionable insights with features such as heat maps, pie charts, and session logs. Real-time data is logged and analyzed on the dashboard, enabling continuous improvement and quick issue resolution.

解决方案的挑战和好处

在实现DIVA 2.0时，DTDC和ShellCode面临着几个重大挑战。整合来自多个遗留系统的实时数据对于提供有关跟踪、费率和可服务性的准确、最新信息至关重要。这可能是通过亚马逊Bedrock代理强大的API集成功能解决的。另一个主要障碍是训练人工智能理解复杂的物流术语和多步骤查询，这是通过使用亚马逊基岩LLM和亚马逊基岩知识库克服的，这些知识库经过了行业特定数据的微调。该团队还必须应对从旧的僵化DIVA系统过渡的微妙过程，同时保持服务连续性和保留历史数据，可能采用并行系统的分阶段方法。最后，将解决方案扩展到处理每月400000多个查询，同时保持性能是一个重大挑战，通过使用Amazon Bedrock Agents的云基础设施来实现最佳的可扩展性和性能来解决。这些挑战突显了在物流等高容量、数据密集型行业升级到人工智能驱动系统的复杂性，并突显了AWS解决方案如何提供必要的工具来克服这些障碍。DTDC通过使用Amazon Bedrock为物流代理提供生成式人工智能，实现了以下好处：

增强与客户支持代理的对话和实时数据访问——该解决方案由亚马逊基岩代理提供支持，提高了自然语言理解能力，使对话更加流畅和引人入胜。通过多步骤推理，它可以更准确地处理更广泛的查询。此外，通过与DTDC的API层无缝集成，物流代理提供对重要信息的实时访问，如跟踪发货、服务可用性和计算运费。高级会话功能和实时数据的结合提供了快速、准确和上下文相关的响应。
智能数据处理和准确的常见问题解答响应——对于复杂的查询，物流代理使用LLM技术处理原始数据并提供结构化、量身定制的响应。这确保用户获得清晰、可操作的见解。对于常见问题，物流代理使用亚马逊基岩知识库提供精确的答案，无需人工支持，减少了等待时间，提高了整体用户体验。
减少对现场客服的依赖和持续改进——尽管物流客服并没有消除对客户支持的需求，但客户支持团队处理的查询数量减少了51.4%。该系统通过集成的实时分析提供了对关键性能指标的宝贵见解，如峰值查询时间、未解决的问题和整体参与度，帮助随着时间的推移完善和提高助手的能力。

结果

生成式人工智能物流代理减轻了客户支持团队的负担，缩短了解决时间，从而带来了更好的客户体验：

DIVA 2.0由Amazon Bedrock提供支持，能够理解自然语言的查询，并支持动态对话，响应准确率为93%
根据过去3个月的仪表板指标数据，他们观察到以下情况：
- 71%的询问与货物有关（256048），而29.5%是一般询问（107132）
- 51.4%的托运查询（131530）没有得到支持票，而48.6%（124518）则创建了新的支持票
- 在导致工单的询问中，40%的人在转向人工智能助理之前从客户支持中心开始，而60%的人在涉及客户支持中心之前从助理开始

DIVA 2.0将客户支持团队处理的查询数量减少了51.4%。DTDC的支持团队现在可以专注于更关键的问题，提高整体效率。