Deep dives into AI agents, workflow engines, RAG retrieval, and LLM integration — mirrored from cnblogs.com/slickflow
Integrates LLM Agent nodes into BPMN process flows, giving each agent reasoning, tool-calling, and collaboration capabilities. Built around Harness engineering conventions following a declare → register → execute pattern for agent tool management. The ReAct loop (Reason → Act → Observe) enables multi-step reasoning tasks. A 5-node procurement flow demo (Requirements Analysis → Supplier Search → Compliance Check → Price Evaluation → Order Execution) fully automates the journey from user request to purchase order.
Read Full Article →[AgentToolSet] attribute registrationA multi-turn Q&A customer service system built on Slickflow.NET's workflow engine and large language models. React frontend + ASP.NET Core backend, combined with a vector database for knowledge retrieval augmentation, enabling precise domain-specific answers.
v3.5.0 introduces three core AI capabilities: natural language "text-to-flow" diagram generation (Qianwen reasoning model), multimodal image classification (GPT-4 Vision), and RAG-enhanced knowledge base Q&A. Supports flexible switching between DeepSeek, OpenAI, Qianwen, and more.