analytics-python

name	analytics-python
summary	Couchbase Analytics Service (CBAS) for Python — run OLAP queries, window functions, and external data links without impacting operational workloads (Server 6.x–7.x; use columnar-analytics for Server 8+)
description	Couchbase Analytics Service (CBAS) for Python — run OLAP queries, window functions, and external data links without impacting operational workloads (Server 6.x–7.x; use columnar-analytics for Server 8+)
compatibility	Python SDK 4.x. Requires couchbase>=4.0.
metadata	{"last_verified":"2026-05","min_server_version":"6.0","deprecated_by":"columnar-analytics","deprecated_since":"8.0","max_server_version":"7.6","handoff":[{"condition":"user is upgrading to Server 8.x and wants the columnar engine","skill":"columnar-analytics"},{"condition":"user asks about SQL++ queries","skill":"server-querying-python"},{"condition":"user wants to write results back to operational collections transactionally","type":"variant","skill":"transactions-python"}]}

Couchbase Analytics — Python

Version disambiguation: This skill covers Couchbase Server 6.x–7.x (legacy CBAS). If the user mentions Server 8, Couchbase 8, or the columnar engine, route to columnar-analytics instead. If the server version is unknown, ask before proceeding. Analytics runs OLAP queries on a shadow copy of your data, isolated from KV and Query Service workloads. Use it for aggregations, reports, and joins across large datasets.

---

Setup — Create an Analytics Collection

-- Link the default bucket (once, in the Analytics workbench or cbsh)
CREATE DATAVERSE myapp IF NOT EXISTS;
USE myapp;
CREATE ANALYTICS COLLECTION orders ON `myapp`.`_default`.`orders` IF NOT EXISTS;

---

SDK Query

from datetime import timedelta
from couchbase.options import AnalyticsOptions

result = cluster.analytics_query(
    "SELECT country, COUNT(*) AS cnt FROM airlines GROUP BY country ORDER BY cnt DESC",
    AnalyticsOptions(timeout=timedelta(minutes=5))
)
for row in result:
    print(row)

Parameterized Query

result = cluster.analytics_query(
    "SELECT * FROM orders WHERE status = $status AND total > $min_total",
    AnalyticsOptions(named_parameters={"status": "pending", "min_total": 100})
)

Window Functions

SELECT name, country, stars,
       RANK() OVER (PARTITION BY country ORDER BY stars DESC) AS rank_in_country
FROM hotels
WHERE stars IS NOT MISSING;

Analytics vs Query Service: See shared/server/analytics-vs-query.md for the full comparison, setup steps, and common errors.

Writing Results Back

Use transactions for atomic write-back, or individual KV upserts for non-atomic bulk writes.

result = cluster.analytics_query('SELECT id, total FROM orders WHERE status = "complete"')
collection = cluster.bucket('myapp').scope('_default').collection('summaries')

for row in result.rows():
    collection.upsert(f"summary::{row['id']}", {'total': row['total']})

For atomic multi-document write-back, wrap in a transaction — see transactions-python.

When to Use Analytics vs Query Service

Scenario	Use
Full collection scan + complex aggregation	Analytics
OLTP — point lookups, small result sets	Query Service
Report that would need many GSI indexes	Analytics
Real-time data (< 1s staleness)	Query Service

Analytics data lags the operational store by seconds to minutes depending on ingestion rate.