Run Audio Quality Testing

Set up and launch a Coval audio-quality testing sweep end-to-end: one voice agent, one test set, the audio-robustness persona pack, shared metrics, and a multi-run report grouped by Persona. Mirrors the public cookbook at https://docs.coval.dev/guides/testing-across-audio-qualities.

When the runs complete this skill hands off to analyze-audio-quality-report for the recommended-fixes write-up.

When To Use

• The user wants to follow the "Testing Across Audio Qualities" cookbook but does not want to hand-launch each persona run.
• The user has a voice agent (PSTN, SIP, WebSocket, LiveKit, Vapi, OpenAI Realtime, Gemini Realtime, etc.) and a test set they trust.
• The user wants a multi-run report URL grouped by Persona that they can save, share, or pass to the analysis skill.

Do not use this skill for chat-only agents. Audio-quality testing exercises STT, TTS, audio timing, and background-noise handling, none of which a chat agent runs.

Prerequisites

coval --version    # CLI must be installed (brew install coval-ai/tap/coval)
coval whoami       # must be authenticated to the right org

If coval whoami fails, run coval login first.

Audio-Robustness Persona Pack

The cookbook uses these seven Coval personas. Each persona configures a different real-world audio condition. The skill expects all seven to exist in the user's organization with these exact names.

Persona Name	Audio Condition
Standard Customer	Baseline clean-call behavior
Impatient Customer	Short answers and lower patience
Confused Customer	Clarification handling
Interruptive Speaker	Overlap and interruption handling
Super Fast Speaker	Fast speech
High Background Noise Speaker	Background noise robustness
Low Volume Speaker	Quiet speaker audio

If any persona is missing in the user's org, see the Missing personas section below before launching.

Workflow

Step 1: Confirm The Voice Agent

If $ARGUMENTS includes an agent name or ID, look it up. Otherwise list voice agents and ask the user to pick one:

coval agents list --format json | jq '.[] | select(.model_type | test("VOICE|REALTIME|WEBSOCKET")) | {id, display_name, model_type}'

Pick exactly one. The cookbook is explicit: keep the agent configuration fixed across all runs. Confirm with the user before continuing.

If the agent emits OpenTelemetry traces, recommend including trace-based timing metrics in step 3. If not, recommend setting up traces (see Coval traces docs) or proceeding with transcript-and-audio-only metrics.

Step 2: Confirm The Test Set

coval test-sets list

Pick one test set. The same test set must be used across every persona run. Subsampling is allowed as long as the same sampled cases run across all seven personas — use a fixed --sub-sample-seed (e.g. 42) and the same --sub-sample-size on every run.

Step 3: Confirm Metrics

The cookbook recommends separating task success from audio-path behavior. Help the user pick at least one metric from each of these groups; reuse any the agent already has attached when they fit.

Goal	Useful Metrics
Task outcome	Composite evaluation, task-completion LLM judges, or scenario-specific pass/fail metrics
Responsiveness	Latency, time to first audio, trace TTFB, provider response-time
Speech recognition	STT Word Error Rate for traced agents, or Transcription Error for audio-backed conversations
Generated voice quality	Voice Quality; Speech Artifact Score and artifact-specific metrics
Conversation flow	Interruption rate, silence, sentiment, turn-level timing

Do not include manager_audio_frequency — it measures signal characteristics, not perceived audio quality.

Do not recommend STT Word Error Rate (Audio Upload) for this general workflow. It only applies when the test set uses uploaded audio with reference transcripts, which is a different setup from the normal voice-simulation sweep.

coval metrics list

Collect the chosen metric IDs into a comma-separated string for the launch step. If the user wants to use the agent's default metrics, omit --metric-ids entirely on the launch call.

Step 4: Resolve Persona IDs

For every persona in the audio-robustness pack, resolve its ID by exact-name match:

for name in "Standard Customer" "Impatient Customer" "Confused Customer" \
            "Interruptive Speaker" "Super Fast Speaker" \
            "High Background Noise Speaker" "Low Volume Speaker"; do
  coval personas list --filter "name:\"$name\"" --format json \
    | jq -r --arg n "$name" '.[] | select(.name == $n) | "\(.id)\t\(.name)"'
done

If any persona returns no results, jump to Missing personas before launching.

Step 5: Launch The Runs

Launch seven runs — one per persona — with the same agent, test set, metrics, and (if subsampling) the same --sub-sample-seed. Use shared tags so the runs are easy to find later.

SUB_SAMPLE_SEED=42
SUB_SAMPLE_SIZE=0   # 0 = all cases; set to a positive integer to subsample
CONCURRENCY=5
TAGS="audio-quality-testing,cookbook"
metric_args=()
if [ -n "${METRIC_IDS:-}" ]; then
  metric_args=(--metric-ids "$METRIC_IDS")
fi

for persona_id in "$STANDARD_ID" "$IMPATIENT_ID" "$CONFUSED_ID" \
                  "$INTERRUPTIVE_ID" "$SUPER_FAST_ID" \
                  "$HIGH_NOISE_ID" "$LOW_VOLUME_ID"; do
  coval runs launch \
    --agent-id "$AGENT_ID" \
    --persona-id "$persona_id" \
    --test-set-id "$TEST_SET_ID" \
    "${metric_args[@]}" \
    --concurrency "$CONCURRENCY" \
    --sub-sample-size "$SUB_SAMPLE_SIZE" \
    --sub-sample-seed "$SUB_SAMPLE_SEED" \
    --tags "$TAGS" \
    --name "Audio Quality — $persona_id" \
    --format json
done

Capture every run_id from each launch response.

Tip: the public API also supports a batch launch (POST /v1/runs/batch) if you prefer one HTTP call. The CLI does not expose batch launch today.

Step 6: Watch For Completion

for run_id in "${RUN_IDS[@]}"; do
  coval runs watch "$run_id"
done

coval runs watch blocks until the run reaches a terminal status (COMPLETED, FAILED, or CANCELLED). Run these in parallel shells or background jobs if you want all seven watched at once.

If any run finishes as FAILED with total_test_cases=0, the run-setup worker crashed before sub-sampling. Re-launch that single run with --sub-sample-size 1 (or omit subsampling) and report the failure as a Coval issue — this has been observed in some orgs.

Step 7: Build The Multi-Run Report URL

Concatenate the seven run IDs and produce the report URL:

RUN_IDS_CSV=$(IFS=,; echo "${RUN_IDS[*]}")
ORG_SLUG=$(coval whoami --format json | jq -r '.organization.slug')
echo "https://app.coval.dev/${ORG_SLUG}/reports/new?run_ids=${RUN_IDS_CSV}"

Tell the user to:

1. Open the URL.
2. Set Compare by to Persona.
3. Save the report with a descriptive name (this preserves the grouping for future viewers).
4. If they want to share without a Coval login, use the report's Share button to publish a shareable link.

Step 8: Hand Off Analysis

Once the report is saved, point the user at analyze-audio-quality-report for a structured next-fix write-up. The analysis skill turns the grouped report into prompt, tool-handling, STT/TTS, or coverage recommendations.

Suggest pasting the saved report URL into:

Use the Coval `analyze-audio-quality-report` skill on this report:
<paste saved report URL>

Missing Personas

If any persona in the audio-robustness pack does not exist in the user's organization, you have two options:

1. Recommended — create them via the Coval app. The audio-robustness personas ship as templates in the Coval persona library. Open Personas → New Persona in the Coval app and pick the matching template, then re-run Step 4. The templates include the correct voice_name, background_sound, voice_volume, and voice_speed for each condition.

2. CLI fallback — create with explicit config. Each persona needs a clear persona_prompt describing the caller role and one or more audio-condition fields (background_sound, voice_volume, voice_speed, etc.). See the persona creation skill and the Coval personas reference for the full field list. Hand-crafted personas may not match the template behavior exactly — prefer option 1 when possible.

Do not invent persona IDs or skip a persona to get the runs launching. Missing one audio condition means the cookbook's grouped-comparison breaks down.

Output Format

When the skill finishes, return a short, actionable summary:

## Audio Quality Testing — Launch Summary

**Agent:** <display_name> (<id>)
**Test set:** <display_name> (<id>)
**Subsample:** <sub_sample_size> cases (seed <sub_sample_seed>) — or "full test set"
**Metrics:** <comma-separated metric names>

**Runs:**
| Persona | Run ID | Status | Run link |
|---|---|---|---|
| Standard Customer | … | … | https://app.coval.dev/<org>/runs/<id> |
| Impatient Customer | … | … | … |
| Confused Customer | … | … | … |
| Interruptive Speaker | … | … | … |
| Super Fast Speaker | … | … | … |
| High Background Noise Speaker | … | … | … |
| Low Volume Speaker | … | … | … |

**Multi-run report:**
https://app.coval.dev/<org>/reports/new?run_ids=<csv> — open and set **Compare by → Persona**, then save.

**Next step:** run the `analyze-audio-quality-report` skill on the saved report.

Guardrails

• One agent, one test set, same metrics, same subsample seed across all seven runs. Any divergence breaks the grouped comparison the report depends on.
• Do not change the agent's configuration mid-sweep. If you must, launch a fresh sweep so the comparison stays apples-to-apples.
• Do not use chat-only agents. The cookbook explicitly does not apply.
• Do not skip a persona because the org "doesn't use" that condition. The full pack is what makes the comparison interpretable.
• Do not present manager_audio_frequency as a perceived audio-quality score.
• Do not invent persona, agent, test set, or metric IDs — always resolve them from the user's org.