Intake spinner & schedules

Post Call Analyzer does not magically watch your storage forever. Call rows land in a data table (recording URL plus metadata), then an intake spinner processes them: download, transcribe, tag, score, summarize.

You can run intake manually when testing, or on a schedule (commonly nightly) so yesterday’s recordings are ready for supervisors in the morning.

What the intake spinner is

When you create a Call Analyzer project, CXGear seeds an intake pipeline (and supporting child steps) wired to that project’s analyzer config. Conceptually it:

1. Picks up rows that need analysis
2. Reads the recording (via your storage / URL setup)
3. Transcribes speech to text (STT provider)
4. Applies tags
5. Scores QA
6. Writes a summary
7. Stores results on the analyzed record (and rolls up analytics)

You do not paste API keys into the spinner — providers are selected on Setup and credentials live under Integrations.

Screenshot needed

Spinners list in a Call Analyzer project showing the intake spinner.

Where: Call Analyzer project → Spinners

Save as: src/assets/screenshots/49-pca-intake-spinner.png

Why intake is a spinner

Spinners are CXGear’s universal “run work over rows” engine. PCA uses the same job, record, and journey model as outreach:

• Jobs show each run’s progress
• Records show per-call results
• Pipeline journey debugs a single failure

That means your ops team learns one mental model for automation and analysis.

Prerequisites before the first run

Use the setup checklist in full. In short:

1. AI providers (LLM, STT, storage) under Integrations
2. Project Setup selects those providers and score bands
3. Tags and QA configured
4. Data table with recording URLs and metadata fields
5. Intake spinner published
6. Runtime service running (schedules and jobs need it)

If the runtime is down, schedules never fire and manual runs stall.

Manual run (how)

1. Load or append call rows into the project data table (CSV or ingest webhook).
2. Open Spinners, select the intake spinner.
3. Confirm it is published.
4. Click Run now (or start from Jobs, depending on your UI habit).
5. Watch Jobs for completed / failed counts.
6. Open Records for scores, grades, tags, and summaries.

Manual runs are perfect for validating prompts and scorecards on a small sample before you trust a nightly schedule.

Schedules (how)

A schedule trigger uses a cron-style expression and timezone (for example every night at 01:00 in Asia/Kolkata).

Where to configure

Schedule lives on the spinner’s trigger settings (not only on the Setup page). Setup may remind you that nightly / hourly cadence is configured on the spinner.

1. Open the intake spinner.
2. Enable a schedule trigger (you can combine schedule with manual).
3. Set cron and timezone.
4. Publish.
5. Keep the runtime running — the runtime evaluates schedules on an interval; the API alone is not enough.

Screenshot needed

Spinner trigger settings with schedule enabled and cron fields.

Where: Call Analyzer project → Spinners → intake → Triggers

Save as: src/assets/screenshots/50-pca-schedule-trigger.png

Why nightly is common

• Recordings often land in storage throughout the day
• Transcription and LLM cost is easier to budget in a batch
• Supervisors review a stable morning report on Analytics and Records

You can run hourly or other cadences if your compliance team needs faster turnaround — just watch provider rate limits and cost.

How rows get into the table

Intake only analyzes rows that exist in the data table:

Method	When to use
CSV import / append	Bulk historical loads
Ingest webhook	CRM or dialler posts new call rows as they finish
API	Custom integrations with an org API key

After rows land, either wait for the schedule or run intake manually.

What success looks like

• Sample manual run produces transcripts, tags, scores, and summaries on Records.
• Scheduled job appears under Jobs at the expected time.
• Analytics totals move after the run (calls analyzed, grade mix, top tags).
• Failures are visible per record with a journey you can open — not a silent skip.

Common problems

Symptom	Likely cause	What to try
Schedule never runs	Runtime not running	Start runtime; confirm health
Job runs but zero analyzed	No new rows / wrong table	Check data table and metadata fields
Transcription failures	STT provider or recording URL	Setup providers; verify URL access / storage
GenAI tags empty	LLM not selected or key missing	Integrations + project Setup
Old scorecard on new runs only	Expected for history	Snapshots freeze QA per analysis — see Tags & QA
“No intake spinner” style errors	Project not fully seeded	Recreate from Call Analyzer template or contact admin to repair config

Intake vs outreach spinners

	PCA intake	Outreach spinner
Input	Call rows with recordings	Contact rows
Goal	Transcript, tags, QA, summary	Messages, dispositions, journeys
Typical trigger	Nightly schedule	Manual, webhook, or campaign schedule

Same platform, different project type.

Next

• PCA setup
• Analyzed records & scores
• Analytics
• Triggers