AgentBasis Python SDK
Management & Observability SDK for AI Agents in production
The AgentBasis Python SDK provides a simple, lightweight way to track the performance, traces, sessions, and behavior of AI agents. It sends data using the OpenTelemetry (OTel) standard, making it compatible with AgentBasis and other observability backends.
This is the foundation SDK that enables deep observability for coded agents built with:
- Pure Python
- LLM Providers:
- OpenAI
- Anthropic
- Gemini
- Frameworks
- LangChain
- Pydantic AI
Installation
pip install agentbasis
Quick Start
1. Initialize the SDK
Start by initializing the SDK with your API key and Agent ID. This usually goes at the top of your main application file.
import agentbasis
# Initialize with your API Key and Agent ID
agentbasis.init(
api_key="your-api-key-here",
agent_id="your-agent-id-here"
)
2. Manual Tracking (Decorators)
Use the @trace decorator to automatically track any function.
from agentbasis import trace
@trace
def chat_with_user(query):
# Your agent logic here
return "Response to: " + query
# When you call this, data is automatically sent to AgentBasis
chat_with_user("Hello world")
3. OpenAI Integration
Automatically track all your OpenAI calls (models, tokens, prompts) with one line of code.
from agentbasis.llms.openai import instrument
# Enable OpenAI instrumentation
instrument()
# Now just use the OpenAI client as normal
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4",
messages=[{"role": "user", "content": "Hello"}]
)
4. Anthropic Integration
Automatically track all your Anthropic Claude calls.
from agentbasis.llms.anthropic import instrument
# Enable Anthropic instrumentation
instrument()
# Now just use the Anthropic client as normal
from anthropic import Anthropic
client = Anthropic()
response = client.messages.create(
model="claude-3-opus-20240229",
max_tokens=1024,
messages=[{"role": "user", "content": "Hello"}]
)
5. LangChain Integration
Track chains, tools, retrievers, and LLM calls in LangChain with full parent-child span relationships.
from agentbasis.frameworks.langchain import get_callback_handler
# Create a callback handler
handler = get_callback_handler()
# Pass it to your LangChain calls
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="gpt-4")
response = llm.invoke("Hello world", config={"callbacks": [handler]})
For chains and agents, pass the callback handler in the config:
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from agentbasis.frameworks.langchain import get_callback_config
# Use get_callback_config() for convenience
chain = LLMChain(llm=llm, prompt=PromptTemplate.from_template("{query}"))
result = chain.invoke({"query": "What is AI?"}, config=get_callback_config())
6. Pydantic AI Integration
Track Pydantic AI agents with built-in OpenTelemetry support.
from agentbasis.frameworks.pydanticai import instrument
# Enable global instrumentation for all Pydantic AI agents
instrument()
# Your agents are now automatically traced
from pydantic_ai import Agent
agent = Agent("openai:gpt-4")
result = agent.run_sync("Hello!")
For per-agent control with user context:
from agentbasis.frameworks.pydanticai import create_traced_agent
# Create an agent pre-configured with tracing and context
agent = create_traced_agent(
"openai:gpt-4",
system_prompt="You are a helpful assistant."
)
# Set user context - it will be included in traces
agentbasis.set_user("user-123")
result = agent.run_sync("Hello!")
7. Track Users & Sessions (Optional)
Associate traces with specific users and sessions to debug issues and see per-user analytics.
# Set the current user (from your auth system)
agentbasis.set_user(current_user.id)
# Optionally set session and conversation IDs
agentbasis.set_session("session-abc")
agentbasis.set_conversation("conv-123")
# All subsequent LLM calls will be tagged with this context
response = client.chat.completions.create(...)
Or use the context manager for scoped context:
from agentbasis import context
with context(user_id="user-123", session_id="session-abc"):
# All traces in this block include the context
response = client.chat.completions.create(...)
Core Concepts
- OpenTelemetry: We use OTel under the hood for maximum compatibility.
- Spans: Every action (function call, LLM request) is recorded as a Span.
- Transport: Data is batched and sent asynchronously to the AgentBasis backend.
Documentation
For full documentation, visit docs.agentbasis.co.
Contributing
See CONTRIBUTING.md for development setup and guidelines.
License
MIT License - see LICENSE for details.
Metadata
Release files for agentbasis 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| agentbasis-0.1.0.tar.gz | 23.2 kB | Without provenance | Details for agentbasis-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| agentbasis-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for agentbasis-0.1.0-py3-none-any.whl |
Total release size: 48.6 kB
Release files / agentbasis-0.1.0.tar.gz
| Download URL | agentbasis-0.1.0.tar.gz |
|---|---|
| Size | 23.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
85d33c5e8edb4f950e47d27faa126de27bf679378ae420a967633a8a342dc262
|
|
BLAKE2b-256 checksum How to use checksums |
e0c82a371533c1cd7fe34aebb4516cc7292f77621b70f18f83054b68b7e31323
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.4
|
Release files / agentbasis-0.1.0-py3-none-any.whl
| Download URL | agentbasis-0.1.0-py3-none-any.whl |
|---|---|
| Size | 25.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0707b3340a416f7df1aff75512f9e60b3a21a21e0ef9d9a4b82f23dad91adbf2
|
|
BLAKE2b-256 checksum How to use checksums |
dc3276544adedafb7ec926d0ae8ba47277357f97fb1006708a26084cb3e00448
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.4
|
Security for agentbasis 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.