Skip to content

:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.

feat(data-warehouse): sync gong call media urls (#129685)

Tyler Crandall33ac0a3da409
diff --git a/products/warehouse_sources/sources/gong/canonical_descriptions.py b/products/warehouse_sources/sources/gong/canonical_descriptions.py
index 211209495a4e..47fa00995ae7 100644
--- a/products/warehouse_sources/sources/gong/canonical_descriptions.py
+++ b/products/warehouse_sources/sources/gong/canonical_descriptions.py
@@ -65,6 +65,25 @@ CANONICAL_DESCRIPTIONS: CanonicalDescriptions = {
"content": "Gong's AI-generated summary of the call — the Spotlight `brief`, `keyPoints`, `highlights`, `callOutcome`, and `outline`. For the verbatim conversation rather than the summary, see `transcripts`.",
},
},
+ "calls_media": {
+ "description": "Download URLs for the audio and video recording of a Gong call, sourced from the /v2/calls/extensive endpoint. Gong keeps each URL valid for 8 hours, so the table is rewritten on every sync and only URLs from the latest sync work.",
+ "docs_url": "https://gong.app.gong.io/settings/api/documentation#post-/v2/calls/extensive",
+ "columns": {
+ "id": "Unique identifier for the call. Join to `calls` or `calls_extensive` on this.",
+ "title": "Title of the call.",
+ "url": "URL of the call within the Gong web app.",
+ "started": "Time at which the call started (ISO 8601).",
+ "duration": "Duration of the call in seconds.",
+ "direction": "Direction of the call (Inbound, Outbound, Conference, or Unknown).",
+ "scope": "Whether the call was internal or external.",
+ "media": "Media type of the call (Audio or Video).",
+ "language": "Detected primary language of the call.",
+ "workspaceId": "Identifier of the workspace the call belongs to.",
+ "primaryUserId": "Identifier of the main Gong user (host) on the call.",
+ "scheduled": "Scheduled start time of the call, if applicable.",
+ "mediaUrls": "Download URLs for the call recording: `audioUrl`, and `videoUrl` when the call has video. Each URL is valid for 8 hours after the sync that wrote it.",
+ },
+ },
"transcripts": {
"description": "The transcript of a recorded Gong call — what each speaker said, grouped into monologues and sentences.",
"docs_url": "https://gong.app.gong.io/settings/api/documentation#post-/v2/calls/transcript",
diff --git a/products/warehouse_sources/sources/gong/gong.py b/products/warehouse_sources/sources/gong/gong.py
index 44ac943a2911..471d655f673b 100644
--- a/products/warehouse_sources/sources/gong/gong.py
+++ b/products/warehouse_sources/sources/gong/gong.py
@@ -136,18 +136,22 @@ def _extensive_body(
return body
-def _flatten_extensive_call(call: dict[str, Any], row_keys: tuple[str, ...]) -> dict[str, Any]:
+def _flatten_extensive_call(
+ call: dict[str, Any], row_keys: tuple[str, ...], row_columns: Optional[dict[str, str]] = None
+) -> dict[str, Any]:
"""Lift the `metaData` block of an extensive call row to the top level.
`/v2/calls/extensive` nests the core call fields under `metaData` and returns the requested
enrichment blocks as siblings. Flattening `metaData` up keeps `id` and `started` available at
the top level for primary-key merge and datetime partitioning, while preserving each
- enrichment block as a nested column.
+ enrichment block as a nested column, renamed through `row_columns` where its key clashes with
+ a `metaData` field.
"""
meta = call.get("metaData") or {}
flattened = dict(meta)
+ columns = row_columns or {}
for key in row_keys:
- flattened[key] = call.get(key)
+ flattened[columns.get(key, key)] = call.get(key)
return flattened
@@ -341,7 +345,8 @@ def _iter_call_rows(
json_body=_extensive_body(config, window_start, window_end, cursor),
)
rows = [
- _flatten_extensive_call(row, config.extensive_row_keys) for row in data.get(config.response_key, [])
+ _flatten_extensive_call(row, config.extensive_row_keys, config.extensive_row_columns)
+ for row in data.get(config.response_key, [])
]
else:
params: dict[str, Any] = {
diff --git a/products/warehouse_sources/sources/gong/settings.py b/products/warehouse_sources/sources/gong/settings.py
index a7e0062adc20..89857030a698 100644
--- a/products/warehouse_sources/sources/gong/settings.py
+++ b/products/warehouse_sources/sources/gong/settings.py
@@ -31,6 +31,10 @@ class GongEndpointConfig:
# Sibling keys of `metaData` to keep as nested columns on the flattened row. Must line up with
# what `extensive_content_selector` requests, or the columns arrive permanently null.
extensive_row_keys: tuple[str, ...] = ()
+ # Column to store an enrichment block under, for blocks whose key clashes with a `metaData`
+ # field — the `media` URLs block would otherwise overwrite the `media` type string. Keys not
+ # listed keep their own name.
+ extensive_row_columns: dict[str, str] = field(default_factory=dict)
# `POST /v2/calls/transcript` answers with `callId` and `transcript` and nothing else. Endpoints
# that set this are driven off the `/v2/calls` list for the same window instead: each page of
# calls supplies the `callIds` filter for one transcript request, and the call's `started`
@@ -168,6 +172,30 @@ GONG_ENDPOINTS: dict[str, GongEndpointConfig] = {
},
],
),
+ # Download URLs for each call's audio and video recording, from `/v2/calls/extensive` with the
+ # `media` block of the selector requested. Gong signs these URLs for 8 hours only, so the table
+ # is full refresh only: an incremental sync would fetch each call once and keep its URL long
+ # after it stopped working, while a full refresh rewrites every URL each run. Kept as its own
+ # default-off table because the URLs open the recording to anyone who reads them, and because
+ # asking for media needs the `api:calls:read:media-url` scope: a key without it must not break
+ # `calls_extensive`.
+ "calls_media": GongEndpointConfig(
+ name="calls_media",
+ path="/v2/calls/extensive",
+ response_key="calls",
+ primary_key="id",
+ partition_key="started",
+ uses_date_window=True,
+ uses_extensive=True,
+ capture_http_samples=False,
+ should_sync_default=False,
+ extensive_content_selector={
+ "context": "None",
+ "exposedFields": {"media": True},
+ },
+ extensive_row_keys=("media",),
+ extensive_row_columns={"media": "mediaUrls"},
+ ),
# Transcript text for the same calls, from `POST /v2/calls/transcript`, which returns only
# `callId` and `transcript`. Each row is stamped with the `started` of the call it belongs to,
# read from the `/v2/calls` page that drove the request — without it a transcript carries no
diff --git a/products/warehouse_sources/sources/gong/source.py b/products/warehouse_sources/sources/gong/source.py
index 9283e102a962..6eb1633a3941 100644
--- a/products/warehouse_sources/sources/gong/source.py
+++ b/products/warehouse_sources/sources/gong/source.py
@@ -63,6 +63,10 @@ To also sync the `calls_extensive` table (call participants and CRM associations
To also sync the `transcripts` table (what was said on each call), additionally grant:
- `api:calls:read:transcript`
+To also sync the `calls_media` table (audio and video download URLs for each call, which Gong keeps valid for 8 hours), additionally grant:
+- `api:calls:read:extensive`
+- `api:calls:read:media-url`
+
To also sync these tables, additionally grant the scope listed next to each:
- `trackers` (keyword tracker definitions): `api:settings:trackers:read`
- `answered_scorecards` (scorecard reviews of calls): `api:stats:scorecards`
@@ -119,7 +123,11 @@ To also sync these tables, additionally grant the scope listed next to each:
incremental_fields=endpoint_config.incremental_fields,
default_incremental_lookback_seconds=endpoint_config.default_incremental_lookback_seconds,
should_sync_default=endpoint_config.should_sync_default,
- description="Only syncs the last 365 days on initial sync"
+ description=(
+ "Only syncs the last 365 days on initial sync"
+ if endpoint_config.supports_incremental
+ else "Syncs the last 365 days on every sync"
+ )
if endpoint_config.uses_date_window
else None,
)
diff --git a/products/warehouse_sources/sources/gong/tests/test_gong.py b/products/warehouse_sources/sources/gong/tests/test_gong.py
index ae6b9d63dae5..4f4338534355 100644
--- a/products/warehouse_sources/sources/gong/tests/test_gong.py
+++ b/products/warehouse_sources/sources/gong/tests/test_gong.py
@@ -259,6 +259,32 @@ class TestExtensiveCalls:
},
{"content": {"brief": "Buyer wants SSO.", "keyPoints": [{"text": "Asked about SSO"}]}},
),
+ (
+ "calls_media_keeps_media_type_and_stores_urls_in_their_own_column",
+ "calls_media",
+ {
+ # `metaData` also has a `media` key, so the URLs must not land on it.
+ "metaData": {
+ "id": "c1",
+ "title": "Discovery",
+ "started": "2026-03-01T00:00:00Z",
+ "media": "Video",
+ },
+ "media": {
+ "audioUrl": "https://media.example.com/c1/audio.mp3",
+ "videoUrl": "https://media.example.com/c1/video.mp4",
+ },
+ "parties": [{"emailAddress": "[email protected]"}],
+ },
+ {"context": "None", "exposedFields": {"media": True}},
+ {
+ "media": "Video",
+ "mediaUrls": {
+ "audioUrl": "https://media.example.com/c1/audio.mp3",
+ "videoUrl": "https://media.example.com/c1/video.mp4",
+ },
+ },
+ ),
]
)
def test_posts_selector_and_flattens_metadata(