BeeAI

IBM Research developed BeeAI Framework as a full-featured toolkit for crafting autonomous agents and coordinated multi-agent systems. The framework supports both Python and TypeScript, offering components for reasoning, action execution, and inter-agent collaboration on complex problem-solving tasks.

This guide explains how to route telemetry from BeeAI applications into InteractiveAI using OpenTelemetry-based instrumentation.

Prerequisites

• InteractiveAI account with API credentials

• LLM provider credentials (OpenAI, Ollama, or other supported provider)

---

Installation

pip install beeai-framework interactiveai openinference-instrumentation-beeai beeai-framework[wikipedia]

---

Configuration

Set your API credentials as environment variables:

import os

# InteractiveAI credentials
# Obtain keys from Settings > API Keys in the dashboard
os.environ["INTERACTIVEAI_PUBLIC_KEY"] = "pk-ia-..."
os.environ["INTERACTIVEAI_SECRET_KEY"] = "sk-ia-..."
os.environ["INTERACTIVEAI_HOST"] = "https://app.interactiveai.com"

# LLM provider credentials
# Adjust based on your chosen provider
os.environ["OPENAI_API_KEY"] = "sk-proj-..."
# Ollama runs locally and requires no API key
# Other providers use their respective environment variables

Initialize the client and confirm connectivity:

import os
from dotenv import load_dotenv

load_dotenv()

from interactiveai import Interactive

client = Interactive(
    public_key=os.getenv("INTERACTIVEAI_PUBLIC_KEY"),
    secret_key=os.getenv("INTERACTIVEAI_SECRET_KEY"),
    host=os.getenv("HOST"),
)

if client.auth_check():
    print("Connection established")
else:
    print("Authentication failed - verify credentials")

---

Enabling Trace Capture

BeeAI Framework includes a dedicated OpenTelemetry instrumentor. Activate it to capture all framework operations:

from openinference.instrumentation.beeai import BeeAIInstrumentor

BeeAIInstrumentor().instrument()

Once enabled, every agent execution, tool invocation, and model call generates spans that flow to InteractiveAI.

---

Enriching Traces with Context

Layer the InteractiveAI SDK on top of BeeAI instrumentation to attach user identifiers, session data, and custom metadata:

from interactiveai import Interactive
from openinference.instrumentation.beeai import BeeAIInstrumentor
from beeai_framework.agents.react import ReActAgent
from beeai_framework.agents.types import AgentExecutionConfig
from beeai_framework.backend.chat import ChatModel
from beeai_framework.backend.types import ChatModelParameters
from beeai_framework.memory import TokenMemory
from beeai_framework.tools.search.wikipedia import WikipediaTool

client = Interactive()
BeeAIInstrumentor().instrument()

with client.start_as_current_span(name="beeai-research-task") as span:
    
    # Attach context to the trace
    client.update_current_trace(
        user_id="user_789",
        session_id="session_def",
        tags=["beeai", "react-agent", "research"],
        metadata={"use_case": "travel-planning", "environment": "staging"}
    )
    
    llm = ChatModel.from_name(
        "openai:gpt-4o-mini",
        ChatModelParameters(temperature=0.7),
    )
    
    agent = ReActAgent(
        llm=llm,
        tools=[WikipediaTool()],
        memory=TokenMemory(llm)
    )
    
    query = "Summarize the main tourist attractions in Tokyo"
    result = await agent.run(
        prompt=query,
        execution=AgentExecutionConfig(max_iterations=5),
    )
    
    # Capture input and output
    client.update_current_trace(
        input=query,
        output=result.result.text
    )

client.flush()

---

Trace Visibility

The InteractiveAI dashboard displays comprehensive trace data including:

• Agent reasoning steps and iteration counts

• Tool calls with arguments and responses

• Model requests with token usage and latency

• Memory operations and context management

• Complete input/output chains for debugging