Skip to main content

Deploy, observe, and auto-heal your AI agents

Project description

agentbase-sdk

Deploy. Observe. Heal. The Python SDK for AgentBase - the operating system for AI agents.

Install

pip install agentbase-sdk

Quick Start (2 lines to instrument)

import agentbase_sdk as ab

ab.init(api_key="ab_live_...", agent_name="my-agent")
ab.instrument_openai()  # auto-trace every LLM call

# That's it. Every OpenAI call is now traced with:
# - Token counts & costs
# - Latency
# - Error detection
# - Auto-heal on failures

Manual Tracing

import agentbase_sdk as ab

ab.init(api_key="ab_live_...", agent_name="support-bot")

with ab.trace("handle-ticket") as t:
    t.set_input(user_question)

    with ab.span("classify", span_type="llm") as s:
        s.set_model("gpt-4o")
        intent = classify(user_question)
        s.set_output(intent)
        s.set_tokens(prompt_tokens=200, completion_tokens=20)

    with ab.span("search-kb", span_type="tool") as s:
        s.set_tool("knowledge_base")
        results = search(intent)
        s.set_output(str(results))

    with ab.span("respond", span_type="llm") as s:
        s.set_model("gpt-4o")
        answer = generate(intent, results)
        s.set_output(answer)
        s.set_tokens(prompt_tokens=800, completion_tokens=200)

    t.set_output(answer)

Decorators

from agentbase_sdk import observe, heal

@observe(as_trace=True)
def handle_support_ticket(question: str) -> str:
    return my_agent.run(question)

@heal(retry=3, fallback=lambda: "Sorry, please try again.")
def risky_api_call(query: str) -> str:
    return external_api.search(query)

Auto-Instrumentation

ab.instrument_openai()      # OpenAI SDK
ab.instrument_anthropic()   # Anthropic SDK
ab.instrument_langchain()   # LangChain
ab.instrument_crewai()      # CrewAI

Auto-Heal

When your agent encounters errors, AgentBase automatically:

Error Fix
Hallucination (404, missing endpoint) Corrects endpoint, retries
Context overflow Chunks input with sliding window
Infinite loop Circuit breaker
Tool timeout Retry with exponential backoff
Wrong tool selected Re-routes with task context
JSON parse error Retries with format hint

Dashboard

View all traces, costs, and heal events at your AgentBase dashboard.

Links

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

agentbase_autoheal-0.1.0.tar.gz (9.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

agentbase_autoheal-0.1.0-py3-none-any.whl (10.2 kB view details)

Uploaded Python 3

File details

Details for the file agentbase_autoheal-0.1.0.tar.gz.

File metadata

  • Download URL: agentbase_autoheal-0.1.0.tar.gz
  • Upload date:
  • Size: 9.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.9

File hashes

Hashes for agentbase_autoheal-0.1.0.tar.gz
Algorithm Hash digest
SHA256 1a3178ab422b4a6973a5ec07b1d24a50bcdeed1006bd5488da0f6cf9c598378c
MD5 fb09da90e6bc0da02a306f5666a728fe
BLAKE2b-256 427396bf539437c41091a57636ee0aa5ec2575aba352005450cafbf66a6c021d

See more details on using hashes here.

File details

Details for the file agentbase_autoheal-0.1.0-py3-none-any.whl.

File metadata

File hashes

Hashes for agentbase_autoheal-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 e8a741aa07c7f6f8306f9f07b18e6a4d5916d17760a5d36bc39d0aa2756f3026
MD5 16d0db2a45763ad2b9e4fe4f3c0d5a7b
BLAKE2b-256 315ab47c2e6c240da6c1742cef98ad53afd1f5af9bd07c8d1c6d519d7ae3c8de

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page