zappybee 0.1.0

ZappyBee Python SDK - Observability for AI agents

zappybee

ZappyBee observability SDK for AI agents (Python). Track traces, steps, token usage, and costs.

Install

pip install zappybee

Quick Start

import os
from zappybee import ZappyBee

ZappyBee.init(
    api_key=os.getenv("ZAPPYBEE_API_KEY", "tc_live_..."),
    # Optional. Defaults to http://localhost:3001
    base_url=os.getenv("ZAPPYBEE_BASE_URL"),
)

Manual Tracing

from zappybee import ZappyBee

trace = await ZappyBee.start_trace(
    "my-agent",
    user_id="user-123",
    session_id="session-abc",
    input={"prompt": "Hello"},
)

step = await trace.start_step(
    "LLM Call",
    "llm_call",
    model="gpt-4o",
    input={"messages": [{"role": "user", "content": "Hello!"}]},
)

# ... your LLM call here ...

await step.end(
    status="success",
    model="gpt-4o",
    output={"role": "assistant", "content": "Hello!"},
    prompt_tokens=10,
    completion_tokens=20,
)

await trace.end(status="success")

Auto-Instrumentation (OpenAI / Anthropic)

from zappybee import ZappyBee

client = ZappyBee.wrap(openai_client)      # wraps chat.completions.create()
client = ZappyBee.wrap(anthropic_client)   # wraps messages.create()

The wrapper supports both non-stream and stream calls (when the client exposes an async iterator).

Backwards Compatibility

If you previously used tokencat, it is still available:

from tokencat import TokenCat

TokenCat.init(api_key="tc_live_...")

License

MIT