Skip to main content

By agent type · Video

Ten stages, one summary, and no idea which stage was wrong.

A video agent segments, transcribes, samples frames, describes, retrieves and then summarizes. By the time the output is wrong it has passed through six models, and the only artifact anyone kept is the last one. Netra keeps all of it as one trace.

The pipeline, stage by stage

Every stage as a span

Segmentation, transcription, frame analysis, retrieval and the final summarisation, in order, under one trace for the asset.

Transcript and reasoning together

Video work is mostly a speech and text problem wearing a timeline. The transcript the model reasoned over sits next to what it concluded.

Latency where the queue builds

Per-stage timing across a long-running job, so the bottleneck is a number rather than an argument between two teams.

Cost per asset

Minutes transcribed, frames sampled and tokens spent attributed to the video and the route that processed it.

Output, graded

Score the summary, the tags or the extracted moments against a reference set — faithfulness to the transcript, correct structure, nothing invented.

Custom spans for your own stages

OpenTelemetry-native instrumentation means a shot detector or a human QA step joins the trace with a decorator, not a rewrite.

The failures worth naming

Two hops back

A bad segment boundary makes a perfectly good summariser produce nonsense. Without the intermediate spans, the blame lands on the last model every time.

Sampled past it

The moment that mattered fell between two sampled frames. Seeing which frames were actually sent is the difference between fixing the sampler and re-prompting forever.

Cost that scales with duration

A pipeline that is fine on a two-minute clip is a budget event on a two-hour recording. Per-asset attribution catches it on the first long file.

Regression-test the pipeline, not the prompt

Keep the assets that broke as a fixed set and grade every model swap, sampler change or prompt edit against them before it ships. The same evaluators keep scoring production, so drift shows up as a trend rather than a complaint.

Start today

Trace the whole pipeline, not just the last step

Every stage as a span, cost attributed per asset. Start free.