Return_to_Blueprints
Agentic Workflow/Phase_06/25 min build

Agentic Support with Multi-Tool Routing

An advanced, 9-node agentic workflow that combines semantic caching with an orchestrator LLM to route complex customer queries to specific API tools before final synthesis.

Execution_Steps

  1. 01

    Ingestion & Cache Layer

    Start by connecting an `input` node to a `cache` node. Link the cache to a `router`. Configure the router to direct Cache Hits to a direct `output` node, ensuring instantaneous replies for frequent queries.

  2. 02

    Deploy the Orchestrator

    Connect the Miss path of the router to an `llm_call` node using "claude-3-5-sonnet". In its system prompt, instruct the model to analyze the user intent and select the appropriate tool.

  3. 03

    Attach Tools

    Drag two `tool_call` nodes onto the canvas. Connect both to the Orchestrator LLM. Configure one for Order Lookups and the other for Knowledge Base searches.

  4. 04

    Synthesis & Final Delivery

    Connect the outputs of both Tool Call nodes to a final `llm_call` node configured with "gpt-4o-mini". This node synthesizes the raw JSON tool outputs into a friendly response, feeding into the final `output` terminal.

Expected_Metrics

P50_LATENCY:< 4500ms
COST_SAVING:68.0%
SLA_LIMIT:6000ms

Ready to verify?

Open the canvas and simulate these parameters in real-time.

Node_Architecture

inputCustomer ChatChat Mode
cacheQuery CacheSemantic / TTL 7200s
routerCache GateHit / Miss Strategy
outputFast Cache DeliveryInstant Response
llm_callIntent Orchestratorclaude-3-5-sonnet
tool_callOrder Lookup APIREST Fetch
tool_callRefund Policy SearchVector Query
llm_callFinal Synthesizergpt-4o-mini
outputResolution DeliveryStandard

Recursive_Read_Next

Architecture

The AI Essay Detector

Data Pipeline

Enterprise Support RAG Pipeline