insights-js/packages/ai

Hanzo Insights AI

TypeScript SDK for LLM observability with Hanzo Insights.

SEE FULL DOCS

Installation

npm install @hanzo/insights-ai

Direct Provider Usage

import { OpenAI } from '@hanzo/insights-ai'
import { Insights } from '@hanzo/insights-node'

const phClient = new Insights('<YOUR_PROJECT_API_KEY>', { host: 'https://us.i.insights.hanzo.ai' })

const client = new OpenAI({
  apiKey: '<YOUR_OPENAI_API_KEY>',
  insights: phClient,
})

const completion = await client.chat.completions.create({
  model: 'gpt-5-mini',
  messages: [{ role: 'user', content: 'Tell me a fun fact about hedgehogs' }],
  insightsDistinctId: 'user_123', // optional
  insightsTraceId: 'trace_123', // optional
  insightsProperties: { conversation_id: 'abc123', paid: true }, //optional
  insightsGroups: { company: 'company_id_in_your_db' }, // optional
  insightsPrivacyMode: false, // optional
})

console.log(completion.choices[0].message.content)

// YOU HAVE TO HAVE THIS OR THE CLIENT MAY NOT SEND EVENTS
await phClient.shutdown()

Custom and unsupported providers

Use this when working with Vercel AI SDK telemetry. @hanzo/ai exposes an OTEL SpanProcessor that maps spans to Insights AI events and sends them through @hanzo/insights-node.

import { NodeSDK } from '@opentelemetry/sdk-node'
import { Insights } from '@hanzo/insights-node'
import { generateText } from 'ai'
import { openai } from '@ai-sdk/openai'
import { InsightsSpanProcessor } from '@hanzo/insights-ai/otel'

const phClient = new Insights('<YOUR_PROJECT_API_KEY>', { host: 'https://us.i.insights.hanzo.ai' })

const sdk = new NodeSDK({
  spanProcessors: [
    new InsightsSpanProcessor(phClient),
  ],
})

await captureAiGeneration(phClient, {
  distinctId: 'user_123',
  traceId: 'trace_abc',
  provider: 'cloudflare-workers-ai',
  model: '@cf/zai-org/glm-4.7-flash',
  input: messages,
  output: result.response,
  modelParameters: { reasoning_effort: 'high' },
  usage: { inputTokens: result.usage?.prompt_tokens, outputTokens: result.usage?.completion_tokens },
  latency: (Date.now() - start) / 1000,
  properties: { feature: 'transcript-toc' },
})

await phClient.shutdown()

captureAiGeneration is the same primitive that every other @insights/ai wrapper funnels through, so the resulting events are indistinguishable from those produced by withTracing, OpenAI, Anthropic, etc.

OpenTelemetry

• aiSdkSpanMapper is the default mapper.
• You can pass custom mappers in InsightsSpanProcessor options to support additional span schemas.

npm install @insights/ai @opentelemetry/sdk-node @opentelemetry/sdk-trace-base @opentelemetry/exporter-trace-otlp-http

For dynamic properties, pass values in experimental_telemetry.metadata on each AI SDK call. These are captured from ai.telemetry.metadata.* and forwarded as Insights event properties. Use processor options (insightsProperties) only for global defaults.

A self-contained SpanProcessor that handles batching and export internally. Use this when your setup accepts a span processor.

• The OTEL route currently maps supported spans into Insights AI events (manual capture path).
• Existing wrapper-based tracing (for example withTracing) still works and is unchanged.

LLM Observability docs

Please see the main Insights docs.

Questions?

Check out our community page.