Monitoring and QA
Your call traces belong on your servers.
Dograh grades calls after they end and keeps a full trace of each. Self-host Dograh and Langfuse, and run the QA model yourself, and transcripts, traces and QA results never leave your servers.
01 / WHAT YOU SEE
What a finished call leaves behind
A QA review
Tags such as DEAD_AIR, a 1-10 score, sentiment and a short summary for each step of the call.
A full trace
What speech-to-text heard, the exact prompt and every tool request and response.
Slow turns stand out
The Call Transcript shows a reasoning delay per agent turn.
Results in your systems
In Run Detail and the webhook's annotations field.
02 / HOW MANY CALLS
Review every call at first, then a sample
Sample Rate (%) starts at 100. Drop it to 5-10% once the agent is stable, and raise it again after each flow change.
03 / VS HOSTED PLATFORMS
Monitoring on your own servers
04 / HOW IT WORKS
Set it up in four steps
- 01
Add the QA node
It takes no edges and reviews calls after they end.
- 02
Set the sample
Leave it at 100% through launch, then lower it to 5-10%.
- 03
Read the flagged calls
Start with the lowest scores, then open the trace for the failing layer.
- 04
Connect Langfuse
Add your keys per organisation. Traces stay private by default.
05 / WHERE THE DATA LIVES
The data stays where your models run
Dograh is the orchestration layer. Self-host it with open-weight models you run, and call data, QA results included, stays inside your network. Your own keys to a hosted model provider move the contract, but the audio still reaches it.
Questions
Does QA change what the agent says on a live call?+
No. The QA node runs after the call ends, on the transcript. It has no edges in the workflow, so it measures the call without changing it.
What does the default QA review return?+
For each step of the call: tags such as DEAD_AIR and USER_FRUSTRATED, a 1-10 score, a sentiment and a short summary. You can replace the prompt with your own checks.
Is the sentiment label the customer's satisfaction?+
No. It is the review's reading of how the call went, useful for sorting calls. To learn how customers felt, ask them and record the answer.
How many calls should QA review?+
Every call at launch and for a week after any flow change. Once the agent is stable, set Sample Rate (%) to 5-10%.
Can I send traces to a self-hosted Langfuse?+
Yes. Enter your Langfuse host and keys in Platform Settings for your organisation. It works on managed and self-hosted Dograh, and traces are private by default.
What happens if I make a trace public in Langfuse?+
A public trace exposes the full call transcript, the prompts and the tool payloads. Keep traces private unless you mean to share one.
Can I see how long each stage of a turn took?+
The Call Transcript shows a reasoning delay per agent turn, only part of the turn. View Trace shows the speech-to-text, model and voice time on each turn; export to Noveum or BigQuery to analyse many calls.
Does any call data leave my servers?+
Not if you self-host Dograh and Langfuse and run every model yourself, the QA model included. A hosted model provider, even on your own keys, still gets the audio.
Find the bad calls. Keep every trace.
Run QA and tracing on a voice agent you host.