Slickflow is a .NET 8 AI workflow engine with deep integration of LLM, RAG and Agent capabilities. Orchestrate multi-agent collaboration through BPMN 2.0 standards to drive intelligent transformation across vertical industries.
Deeply fusing large models, knowledge retrieval and agent orchestration to build truly enterprise-grade AI workflows.
Built-in LLM nodes support seamless connection with OpenAI GPT, DeepSeek, Alibaba Qwen and other leading models. Call LLM capabilities directly within BPMN process nodes for intelligent decision-making, content generation and structured data extraction.
Built-in RAG service nodes combined with vector databases (Supabase/pgvector, Milvus, Qdrant) enable semantic retrieval of enterprise knowledge bases, injecting domain knowledge into LLMs to significantly improve accuracy and reduce hallucinations.
ReAct loop (ReasonβActβObserve) drives multiple Agent nodes to work collaboratively. Each node independently decides, uses tools, and passes context. BPMN flowcharts visually orchestrate complex multi-step reasoning and cross-system automation.
The robustness of traditional BPMN workflow engines, overlaid with modern AI capabilities, making business automation smarter and more flexible.
Register tool functions automatically with C# attributes. The engine uses reflection to generate JSON Schema, and agents automatically identify and call tools β no manual tool definition templates needed.
Built-in RAGService node type connects to vector databases for semantic retrieval. Supports TF-IDF + LSH intelligent caching to improve retrieval efficiency and response speed.
Users describe business processes in natural language, and the system automatically generates BPMN 2.0-compliant flowcharts. AI pre-fills Agent configurations β no technical background required.
AgentConversationMemory mechanism enables conversation history sharing across nodes. Multiple Agent nodes share the same process instance context, ensuring reasoning continuity.
Define business processes in code with a fluent API. Supports LocalMethod, ServiceTask, RagService and AgentTask node types. Run automated workflows without a database.
Full support for sequence, split/merge, sub-process, multi-instance task workflow patterns. Compatible with SQL Server, MySQL, PostgreSQL, Oracle. Production-grade stability.
AI multi-agent workflows have gone live across multiple verticals β covering the full business chain from pre-sales to back-office automation.
Built on a knowledge base of vehicle specs, subsidy policies and charging-network information, the assistant answers common pre-sales questions to help visitors understand the product. Interested users are invited to leave their contact details so a sales consultant can follow up.
Backed by a professional mental-health knowledge base, the assistant answers common questions about psychological wellbeing and offers general coping suggestions. Users who want further support are guided to leave their contact details for referral to a licensed counselor.
Drawing on a med-spa treatment knowledge base, the assistant answers visitors' common questions about procedures, pricing and what to expect. Users interested in booking a consultation are invited to leave their contact details for a specialist to follow up.
Powered by a lighting-product knowledge base, the customer-facing assistant handles pure knowledge Q&A β spec lookups, model comparisons and application-scenario advice. The procurement-facing module supports AI-assisted auto-RFQ to quickly gather and organise supplier quotes for purchasing decisions.
Three office automation scenarios in one platform: HR uploads rΓ©sumΓ©s and AI scores and ranks candidates automatically; Finance uploads receipt images and OCR extraction generates a structured expense report; a Logistics assistant handles pure knowledge Q&A on shipment policies and common queries using a logistics knowledge base.
Backed by a pet boarding service knowledge base, the assistant answers common questions about services, pricing and care procedures. Pet owners who wish to make a booking are guided to leave their contact details and basic pet information so staff can follow up and arrange check-in.
From LLM calls to business system integration, each layer has clear responsibilities with flexible replacement and custom extension support.
β
β
β
Use C# attribute annotations to automatically register Agent tools. The engine uses reflection to auto-generate JSON Schema, bind parameters and return values β no manual tool definitions, dramatically reducing integration cost.
// Auto-register Agent tools with attribute annotations [AgentToolSet("SupplierSearch")] public class SupplierSearchService { [AgentTool("Search suppliers and get quotes")] public async Task<SupplierQuote> RequestQuote( string productCategory, int quantity, string deliveryRegion) { // Engine auto-handles parameter binding return await _supplierApi .GetQuotesAsync(productCategory, quantity); } } // Code-first AI workflow definition var wf = new Workflow(); wf.Start("Begin") .AgentTask("Needs Analysis", "NeedsAnalysis") .AgentTask("Supplier Search", "SupplierSearch") .AgentTask("Compliance Check", "ComplianceCheck") .AgentTask("Price Evaluation", "PriceEvaluation") .ServiceTask("Create PO", "CreateOrder") .End("Done");
Begin with the open-source Community edition, or choose Enterprise for full AI features and technical support.
MIT Open Source Β· Commercial Use Β· Docker Deployment Β· SQL Server / MySQL / PostgreSQL / Oracle