Skip to content

Core concepts

CXGear uses a small set of ideas repeatedly. Understanding them avoids confusion later (for example why Records does not show your full CSV, or why a spinner cannot see a column).

Your company account. Billing, users, roles, API keys, vault secrets, and integrations all belong to the organization. Everyone you invite works inside the same org (subject to roles).

You do not create a new organization for every campaign — you create projects instead.

A workspace for one campaign or client. Tables, spinners, jobs, and outcomes for that work live here.

There are two main project types:

TypePurpose
OutreachContact lists, multi-channel automation, dispositions
Call AnalyzerRecordings → transcript → tags → QA score → summary
Screenshot needed
Projects list with at least one project visible.
Where: Projects
Save as: src/assets/screenshots/02-projects-list.png

A table of source rows — contacts for outreach, or call rows for analysis. You import CSV files, append more rows, or push rows through an ingest webhook.

Source data is always available under Data tables. The Records page is for results after a run, not for browsing the full import.

When you configure metadata fields on a data table, you choose which columns matter for automation (for example name, phone, account_id).

Only those fields:

  • Enter the spinner as params / payload
  • Appear as contact context on the Records page
  • Show up in column pickers in the spinner builder

Other CSV columns stay in the table for reference but do not enter the pipeline.

A pipeline: an ordered list of steps that run for each row (Log, AI agent, WhatsApp, Tag, and so on). You build it, publish it (freeze the definition), then run it.

Child (supporting) spinners can be linked for retries or callbacks; they usually start from a parent, not from Run now.

One execution of a published spinner. A job processes many records and tracks completed / failed counts. Open Jobs to browse runs and drill into individual records.

One contact’s result for a job:

  • Status — completed, failed, waiting, …
  • Metadata — the fields that entered the pipeline
  • Outcomes — what the pipeline produced (disposition, AI text, message ids, …)

Internal debug fields (log lines, model names) are hidden from the main outcome view; use Pipeline journey when you need step-by-step debug detail.

A connected path to the outside world: WhatsApp Business number, SMS sender, email, or voice. Channels are configured under Integrations and selected inside spinner steps.

A reusable provider configuration (LLM, speech-to-text, storage). You add credentials once; projects and spinners reference them instead of pasting keys everywhere.

Encrypted storage for secrets (tokens, passwords). Prefer Integrations for LLM keys; use Vault when a step needs a named secret.

A user who can take live work (calls today; chats in the conversations roadmap). Agents use presence (available / offline). Admins are not automatically agents unless they also have agent permissions and dialler setup.

A debug timeline for one record: each step’s result, log messages, and errors. Useful when something fails; not the primary ops screen.