agent-healthkit 0.2.0

Platform-agnostic health monitoring SDK for AI agents

Agent HealthKit

Agent HealthKit is a platform-agnostic health monitoring SDK for AI agent frameworks. It provides a lightweight adapter interface that any agent system can implement to expose runtime metrics (memory usage, context size, error rates, uptime), built-in detectors for common failure modes like context bloat and resource exhaustion, and a unified run_full_check API that aggregates detector results into actionable health reports — all with zero required dependencies.

Install

pip install agent-healthkit

# Optional: system resource monitoring (CPU/RAM)
pip install agent-healthkit[system]

Quick Start

from healthkit_sdk import GenericAdapter, run_full_check

adapter = GenericAdapter()
adapter.set_metrics(memory_mb=512, context_tokens=45000, error_rate=0.01, uptime_seconds=3600)
report = run_full_check(adapter)
print(report.status)        # "healthy" | "warning" | "critical"
print(report.issues)        # list of detected issues

Built-in Detectors

Detector	Description	Triggers When
ContextBloatDetector	Monitors token/context window usage	Context > 80% of max (warning) or > 95% (critical)
SystemResourcesDetector	Monitors CPU and RAM usage via psutil	RAM > 80% (warning) or CPU > 90% (critical)

Extending

Implement BaseAdapter to connect your own agent framework:

from healthkit_sdk import BaseAdapter, AgentMetrics

class MyAdapter(BaseAdapter):
    def collect_metrics(self) -> AgentMetrics:
        return AgentMetrics(memory_mb=..., context_tokens=..., ...)

License

MIT © JakebotLabs