Observability & Resilience (maticlib.observability & maticlib.resilience)

Tracing

Easily log execution time, token usage, and errors across the entire lifecycle.

from maticlib.observability.trace import PipelineTrace
from maticlib.observability.callbacks import LoggingCallbackHandler

trace = PipelineTrace(pipeline_name="My_RAG")
handler = LoggingCallbackHandler()
handler.on_pipeline_start(trace)
# ... run steps ...
handler.on_pipeline_end(trace)

Retry Policies

Handle flaky external LLM APIs with robust, exponential backoff. - Use RetryPolicy manually or the @with_retry(max_retries=3) decorator on any function.

from maticlib.resilience.retry import with_retry

@with_retry(max_retries=3, initial_delay=1.0)
def call_external_api():
    # Will automatically retry on exceptions
    pass