AI Tracing & Monitoring

RavenLens provides powerful tracking capabilities for both direct AI provider usage and AI Agents.

AI Providers

You can track requests to various AI providers. The SDK wraps the provider's client and automatically sends telemetry data to RavenLens.

Supported Providers

• OpenAI
• Google Vertex AI / Gemini
• Anthropic
• Groq
• And more...

Basic Usage (Without Session)

Each trace acts as an individual tracking unit.

import OpenAI from 'openai';
import { OpenAITracker } from '@ravenlens/ai-sdk';

// Initialize OpenAI
const openai = new OpenAI({
    apiKey: process.env.OPENAI_API_KEY
});

// Initialize Tracker with RavenLens configuration
const tracker = new OpenAITracker(openai, {
    apiKey: process.env.RAVENLENS_API_KEY,
    apiVersion: "1.0",
    appID: process.env.RAVENLENS_APP_ID,
    user: { // User object is optional
        id: "your-user-id",
    }
});

const { response, sessionId, traceId } = await tracker.ask({
    messages: [{ role: "user", content: "What is the capital of France?" }],
    model: "gpt-4",
});

const answer = response.choices[0]?.message?.content;
console.log('Answer: ', answer);

Usage with Sessions

Connecting traces to sessions gives you a full conversational history.

Approach 1: Explicit Session Generation

Generate a session ID first and pass it to the tracker.

import OpenAI from 'openai';
import { OpenAITracker, RavenLens } from '@ravenlens/ai-sdk';

const raven = new RavenLens({
    apiKey: process.env.RAVENLENS_API_KEY,
    appId: process.env.RAVENLENS_APP_ID
});

async function main() {
    // 1. Generate new Session ID
    const sessionId = await raven.sessions.generateNewSessionId();

    // 2. Initialize Tracker with existing Session ID
    const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
    const tracker = new OpenAITracker(openai, {
        apiKey: process.env.RAVENLENS_API_KEY,
        apiVersion: "1.0",
        appID: process.env.RAVENLENS_APP_ID,
        sessionId: sessionId // Connects this trace to the session
    });

    const { response } = await tracker.ask({
        messages: [{ role: "user", content: "Hello!" }],
        model: "gpt-4",
    });
}

Approach 2: Chaining Sessions

Use the session ID returned from the first call for subsequent calls.

// ... inside a loop or consecutive calls
const { response, sessionId } = await tracker.ask({ ... });

// Use retrieved `sessionId` for the next tracker initialization
const nextTracker = new OpenAITracker(openai, {
    ...,
    sessionId: sessionId
});

AI Agents

RavenLens supports tracking for Agent frameworks like BeeAI and LangChain.

BeeAI Framework Example

import { ReActAgent } from "beeai-framework/agents/react/agent";
import { OllamaChatModel } from "beeai-framework/adapters/ollama/backend/chat";
import { UnconstrainedMemory } from "beeai-framework/memory/unconstrainedMemory";
import { BeeAiTracker, withRavenLens } from "@ravenlens/ai-sdk"

const tracker = new BeeAiTracker({
    apiKey: process.env.RAVENLENS_API_KEY,
    apiVersion: "1.0",
    appID: process.env.RAVENLENS_APP_ID,
    // sessionId: "...", // Optional: for session grouping
});

// 1. Initialize LLM
const llm = new OllamaChatModel("granite4:micro");

// 2. Wrap LLM with RavenLens
const trackedLLM = withRavenLens(llm, tracker.config);

// 3. Create Agent with tracked LLM
const agent = new ReActAgent({
    llm: trackedLLM,
    tools: [],
    memory: new UnconstrainedMemory(),
});

// 4. Wrap Agent
const trackedAgent = withRavenLens(agent, tracker.config);

// 5. Run and Observe
const { response, sessionId, traceId } = await trackedAgent
    .run({ prompt: "Help me with observability." })
    .observe(tracker.observer());