Observability for LLM apps
See inside
every answer.
Spectra splits each model call into the signals behind it — latency, tokens, cost, retrieval and eval scores — so you can see why it answered the way it did.
Observability for LLM apps
Spectra splits each model call into the signals behind it — latency, tokens, cost, retrieval and eval scores — so you can see why it answered the way it did.
Move your cursor up and down to tilt the beam.
Light model: Snell's law + Cauchy dispersion, ×6 for clarity.
Trace values are an example.