> ## Documentation Index
> Fetch the complete documentation index at: https://filament.getgalaxy.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Gong

> Read calls, transcripts, users, and reference data from Gong

The Gong source reads conversation data and the users, workspaces, and settings
associated with it. It uses an
[HTTP manifest](/pages/connectors/building-a-connector/http-manifests) and is
**alpha**. It has been checked against Gong's API docs and tested with mock
responses, but has not yet been tested against a live account.

## Configuration

| Field               | Scope      | Default | Description                                          |
| ------------------- | ---------- | ------- | ---------------------------------------------------- |
| `access_key`        | Connection |         | Required. Secret. Gong API access key.               |
| `access_key_secret` | Connection |         | Required. Secret. Secret paired with the access key. |
| `base_url`          | Connection |         | Required. API base URL from your Gong API settings.  |

A Gong technical administrator can create credentials under **Company settings
→ Ecosystem → API**. Copy the base URL shown there for your account, along with
the key and secret. See [Gong's API introduction](https://help.gong.io/apidocs/introduction-2)
for setup and access requirements.

The key needs access to each selected resource. Connection testing makes a
request to the users endpoint, so user-read access is also required.

| Resources                                    | Required scope                 |
| -------------------------------------------- | ------------------------------ |
| `users`                                      | `api:users:read`               |
| `calls`                                      | `api:calls:read:extensive`     |
| `transcripts`                                | `api:calls:read:transcript`    |
| `workspaces`                                 | `api:workspaces:read`          |
| `keyword_trackers`                           | `api:settings:trackers:read`   |
| `scorecards`                                 | `api:settings:scorecards:read` |
| `library_folders`, `library_folder_contents` | `api:library:read`             |

## Resources

| Resource                  | Endpoint                         | Parent            |
| ------------------------- | -------------------------------- | ----------------- |
| `users`                   | `GET /v2/users`                  |                   |
| `calls`                   | `POST /v2/calls/extensive`       |                   |
| `transcripts`             | `POST /v2/calls/transcript`      |                   |
| `workspaces`              | `GET /v2/workspaces`             |                   |
| `keyword_trackers`        | `GET /v2/settings/trackers`      |                   |
| `scorecards`              | `GET /v2/settings/scorecards`    |                   |
| `library_folders`         | `GET /v2/library/folders`        |                   |
| `library_folder_contents` | `GET /v2/library/folder-content` | `library_folders` |

Users, calls, and transcripts are selected by default. The other resources are
optional. Filament only writes the resources you select. To read folder contents,
it also lists folders, even when you haven't selected the folder resource.

`calls` includes call metadata, participants, extended CRM context, summaries,
topics, trackers, interaction statistics, and public comments. Nested data stays
together in JSON columns. Media download URLs are not requested. See
[Gong's call data reference](https://help.gong.io/apidocs/retrieve-detailed-call-data-by-various-filters-v2callsextensive-2)
for field availability.

`transcripts` has one row per call, keyed by `call_id`. Its JSON transcript array
keeps speakers, topics, sentences, and relative timestamps together. Transcripts
are read directly in bulk, without a separate request for each call.

`keyword_trackers` and `scorecards` contain definitions, including keyword rules
and scorecard questions. Scorecards do not include submitted answers or scores.

The library lists public, non-archived folders. `library_folder_contents` has
one row per folder, with a JSON array of its calls and snippets. A call can appear
more than once in that array, with a different snippet each time.

IDs are stored as strings. Optional fields are nullable, timestamps include
time zones, and the `raw` column keeps fields that aren't mapped to their own
columns.

## Modes

All resources support full reads. Calls and transcripts are requested without
date bounds. Use **full upsert** to refresh existing rows, or **full replace** to
also remove rows that are no longer returned by Gong.

Incremental reads are not supported. Gong's call date filters track when a call
took place, rather than when its transcript, summary, or CRM context last changed.
Using those dates as an update cursor could miss later edits to older calls.

## Behavior

* **Auth**: every request uses HTTP Basic authentication with the access key
  and secret, against your configured base URL.
* **Pagination**: users, calls, and transcripts follow `records.cursor` until
  it is missing or empty. The other endpoints do not use pagination.
* **Recovery**: interrupted top-level reads can resume from a saved page cursor.
  If that request fails, the existing HTTP engine retries the resource from the
  beginning. Library contents always restart from the folder listing.
* **Rate limiting**: three requests per second, shared across resources in this
  connection. Gong's default allowance is also 10,000 requests per company per
  day, shared with other API clients. Filament honors `Retry-After` on throttled
  requests but does not track the daily allowance.
* **Empty results**: calls and transcripts accept a `404` only when Gong's
  `errors` array contains exactly `No calls found corresponding to the provided filters`.
  Other errors still fail the read. Gong also documents empty-result `404`
  responses for scorecards and library folders. Those remain errors until their
  exact bodies have been verified.

This version covers conversation data and its reference tables. Activity and
coaching reports, answered scorecards, audit logs, Engage, CRM imports, and media
downloads are outside its scope.