Flowise integration

Flowise is a visual LangChain builder. Spanlens captures Flowise flows two ways: by routing LLM calls through the Spanlens proxy, which works for any node that calls OpenAI / Anthropic / Gemini, or by attaching the Spanlens LangChain callback handler in a custom node for full span-tree capture.

Option A: Proxy mode (zero code)

In Flowise, open the LLM credential settings for any chat model node and change the base URL to the Spanlens proxy endpoint for that provider.

Use your Spanlens API key (sl_live_*) in place of the provider API key. Spanlens stores your provider key securely and signs the upstream call on your behalf. Every LLM node in the flow now lands in /requests automatically.

Option B: Custom callback node (full span tree)

For full per-node span capture (including the visual flow topology), register a custom Flowise tool that attaches the Spanlens LangChain handler to the chain config.

// flowise-spanlens-tool.ts
import { SpanlensClient } from '@spanlens/sdk'
import { createSpanlensCallbackHandler } from '@spanlens/sdk/langchain'

const client = new SpanlensClient()
const handler = createSpanlensCallbackHandler({ client })

export const onExecuteFlow = async (flowConfig, runtimeContext) => {
  return {
    ...runtimeContext,
    callbacks: [...(runtimeContext.callbacks || []), handler],
  }
}

What you see in Spanlens

Each Flowise flow execution becomes a trace. The flow ID is preserved as a tag. Each LLM node becomes a span with model + cost + latency. Tool nodes become tool spans. Conditional branches and parallel fan-outs are reconstructed from the parent_span_id chain.

Where to go next

• LangChain integration, since Flowise builds on LangChain.
• Proxy reference, full proxy URL list.