visibe 0.1.2

AI Agent Observability Platform - Track CrewAI, LangChain, LangGraph, and more

Visibe SDK for Python

Observability for AI agents. Track costs, performance, and errors across your entire AI stack — whether you're using CrewAI, LangChain, LangGraph, AutoGen, or direct OpenAI calls.

---

📦 Getting Started

Installation

pip install visibe

Basic Configuration

Set your API key in a .env file:

VISIBE_API_KEY=sk_live_your_api_key_here

Then initialize the SDK — one line instruments everything:

import visibe

visibe.init()

That's it. Every OpenAI, LangChain, LangGraph, CrewAI, and AutoGen client created after this call is automatically traced.

Quick Usage Example

import visibe
from openai import OpenAI

visibe.init()

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}]
)
# This call is automatically traced — cost, tokens, duration, and content are captured.

---

🧩 Integrations

Visibe integrates with the most popular AI/agent frameworks in Python. Every integration supports three levels of control:

Framework	visibe.init()	obs.instrument()	obs.track() / manual
OpenAI	✅	✅	✅
LangChain	✅	✅	✅
LangGraph	✅	✅	✅
CrewAI	✅	✅	✅
AutoGen	✅	✅	✅
AWS Bedrock	✅	✅	✅

Also works with OpenAI-compatible providers: Azure OpenAI, Groq, Together.ai, DeepSeek, and others.

OpenAI

from visibe import Visibe
from openai import OpenAI

obs = Visibe()
client = OpenAI()

obs.instrument(client)

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}]
)

Group multiple calls into one trace:

with obs.track(client, name="my-conversation"):
    r1 = client.chat.completions.create(model="gpt-4o-mini", messages=[...])
    r2 = client.chat.completions.create(model="gpt-4o-mini", messages=[...])
# ^ Both calls sent as one grouped trace

Works with chat completions and Responses API, streaming, tool calls, sync and async clients.

LangChain / LangGraph

from visibe import Visibe
from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

obs = Visibe()
llm = ChatOpenAI(model="gpt-4o-mini")
graph = create_react_agent(llm, tools)

obs.instrument(graph, name="my-agent")

result = graph.invoke({"messages": [("user", "Your prompt here")]})

Dynamic pipe chains (prompt | llm | parser) are also automatically instrumented when using visibe.init(). Nested sub-graphs are tracked with hierarchical agent names.

CrewAI

from visibe import Visibe
from crewai import Agent, Task, Crew

obs = Visibe()

architect = Agent(role="Plot Architect", goal="Design mystery plots", backstory="...")
designer = Agent(role="Character Designer", goal="Create characters", backstory="...")
narrator = Agent(role="Narrator", goal="Write the story", backstory="...")

task1 = Task(description="Create a plot outline", agent=architect, expected_output="...")
task2 = Task(description="Design characters", agent=designer, expected_output="...", context=[task1])
task3 = Task(description="Write the story", agent=narrator, expected_output="...", context=[task1, task2])

crew = Crew(agents=[architect, designer, narrator], tasks=[task1, task2, task3])

obs.instrument(crew, name="mystery-writer")
result = crew.kickoff()
# ^ Single trace with all agents, LLM calls, and per-task cost breakdown

With visibe.init(), trace names are auto-derived from agent roles (e.g. "Plot Architect, Character Designer, Narrator"). Training and testing runs (crew.train(), crew.test()) are traced too.

AutoGen

from visibe import Visibe
from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_agentchat.agents import AssistantAgent

obs = Visibe()
model_client = OpenAIChatCompletionClient(model="gpt-4o-mini")

obs.instrument(model_client, name="my-conversation")

assistant = AssistantAgent("assistant", model_client=model_client)
result = await assistant.run(task="Help me with this task")

AWS Bedrock

from visibe import Visibe
import boto3

obs = Visibe()
bedrock = boto3.client("bedrock-runtime", region_name="us-east-1")

obs.instrument(bedrock)

response = bedrock.converse(
    modelId="anthropic.claude-3-haiku-20240307-v1:0",
    messages=[{"role": "user", "content": [{"text": "Hello!"}]}]
)

Group multiple calls into one trace:

with obs.track(bedrock, name="my-workflow"):
    r1 = bedrock.converse(modelId="anthropic.claude-3-haiku-20240307-v1:0", messages=[...])
    r2 = bedrock.converse(modelId="amazon.nova-lite-v1:0", messages=[...])
# ^ Both calls sent as one grouped trace

Supports all Bedrock API methods: converse, converse_stream, invoke_model, and invoke_model_with_response_stream. Works with all models available via Bedrock — Claude, Nova, Llama, Mistral, and more.

---

⚙️ Configuration

from visibe import Visibe

# API key from environment (recommended)
obs = Visibe()

# Or pass directly
obs = Visibe(api_key="sk_live_abc123")

# Group traces by session
obs = Visibe(session_id="user-session-123")

Environment Variables

Variable	Description	Default
VISIBE_API_KEY	Your API key (required)	—
VISIBE_API_URL	Override API endpoint	https://api.visibe.ai
VISIBE_AUTO_INSTRUMENT	Comma-separated frameworks to auto-instrument	All detected
VISIBE_CONTENT_LIMIT	Max chars for LLM/tool content in spans	1000
VISIBE_DEBUG	Enable debug logging (1 to enable)	0

---

📊 What Gets Tracked

Metric	Description
Cost	Total spend + per-agent and per-task cost breakdown
Tokens	Input/output tokens per LLM call
Duration	Total time + time per step
Tools	Which tools were used, duration, success/failure
Errors	When and where things failed
Spans	Full execution timeline with LLM calls, tool calls, and agent events

---

📚 Documentation

For advanced usage, detailed integration guides, and API reference, check out the full documentation:

• OpenAI integration
• LangChain integration
• CrewAI integration
• AutoGen integration
• AWS Bedrock integration

---

🔗 Resources

• PyPI Package — Install the latest version
• Visibe Dashboard — View your traces and analytics

---

📃 License

MIT — see LICENSE for details.