Hugging Face Tracing And Monitoring

See the Quick Start guide on how to install and configure Graphsignal.

Agents

Graphsignal automatically instruments, traces and monitors Transformers Agents. Traces provide execution details for each run, LLM, and tool. These insights include prompts, completions, parameters, latency, and exceptions.

Pipelines

To monitor any call, including Hugging Face pipelines, use a @graphsignal.trace_function decorator or graphsignal.start_trace context manager:

from transformers import pipeline

pipe = pipeline(task="text-generation")

with graphsignal.start_trace('predict'):
    output = pipe('some text')

Useful links

• Tracing Hugging Face Transformers Agents
• OpenAI API Cost Tracking: Analyzing Expenses by Model, Deployment, and Context