You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Vector indexes in CockroachDB now support table backfills, making it easy to add vector indexes to existing tables without requiring manual data rewrites or service interruption. This update enhances the usability of vector search features by allowing seamless indexing of pre-existing data. Additionally, support for both cosine distance and inner product has been added, giving users more flexibility in selecting similarity metrics that best suit their machine learning or AI workloads. Together, these improvements make it easier to build and scale intelligent applications directly on CockroachDB.
48
+
<p class="feature-summary">Vector Indexing Improved and Generally Available</p>
49
+
<p class="feature-description">
50
+
CockroachDB's vector indexing capabilities are now generally available, promoting production-ready similarity search for AI and machine learning workloads from Preview to GA. This release brings critical operational improvements while maintaining the distributed architecture advantages that distinguish CockroachDB from specialized vector databases and PostgreSQL with pgvector.
51
+
</p>
52
+
<p class="feature-description">
53
+
The standout v25.4 improvement is <strong>online table backfills</strong>: adding vector indexes to tables with existing data no longer requires taking the table offline during the backfill process. This eliminates downtime when adopting vector search capabilities, enabling seamless integration of AI features into live production systems without service interruption.
<li>Online table backfills: Add vector indexes without service interruption</li>
58
+
<li>Distributed horizontal scaling: Automatically partition vector indexes across nodes as data grows, unlike single-node PostgreSQL pgvector</li>
59
+
<li>Incremental index maintenance: No background rebuilds required when inserting or updating vectors</li>
60
+
<li>Multiple distance metrics: L2 distance, cosine distance, and inner product for diverse ML workloads</li>
61
+
<li>pgvector compatibility: Standard PostgreSQL vector operations and SQL semantics</li>
62
+
<li>Unified architecture: Combine vector search with operational data, transactions, and consistency</li>
63
+
</ul>
64
+
<p class="feature-description">
65
+
Vector indexing is now enabled by default (<code>feature.vector_index.enabled = true</code>), ready for production AI workloads including semantic search, retrieval-augmented generation, recommendation systems, and content similarity matching.
51
66
</p>
52
67
</td>
53
68
<td>25.4</td>
@@ -59,7 +74,7 @@ This section summarizes the most significant user-facing changes in v25.4.0 and
59
74
<tr>
60
75
<td>
61
76
<p class="feature-summary">Stable SQL API and Controlled Access to Unsupported crdb_internal Tables</p>
62
-
<p class="feature-description" markdown="1">
77
+
<p class="feature-description">
63
78
CockroachDB introduces a stable, well-defined SQL API for system introspection. This API enables customers and third-party tools to safely and reliably query cluster metadata and performance data in production environments.
64
79
</p>
65
80
</td>
@@ -74,7 +89,7 @@ This section summarizes the most significant user-facing changes in v25.4.0 and
@@ -92,13 +107,13 @@ This section summarizes the most significant user-facing changes in v25.4.0 and
92
107
<tr>
93
108
<td>
94
109
<p class="feature-summary">Transaction Diagnostic Bundles for Comprehensive SQL Tracing</p>
95
-
<p class="feature-description" markdown="1">
110
+
<p class="feature-description">
96
111
CockroachDB now supports Transaction Diagnostic Bundles, enabling Cockroach Labs Support to capture and analyze a complete trace of an entire SQL transaction not just individual statements.
97
112
98
113
This feature extends the existing Statement Diagnostic Bundle capability to provide a holistic view of multi-statement transactions, simplifying the diagnosis of complex customer issues during support incidents.
<li>A Jaeger trace and text trace of the full transaction execution path.</li>
103
118
<li>Individual Statement Diagnostic Bundles for each statement executed within the transaction.</li>
104
119
</ul>
@@ -112,17 +127,17 @@ This section summarizes the most significant user-facing changes in v25.4.0 and
112
127
<tr>
113
128
<td>
114
129
<p class="feature-summary">Structured and Consistent Logging Channels</p>
115
-
<p class="feature-description" markdown="1">
130
+
<p class="feature-description">
116
131
CockroachDB has updated structured and consistent logging channels across the system to improve log discoverability, production observability, and support troubleshooting.
@@ -154,13 +169,13 @@ This section summarizes the most significant user-facing changes in v25.4.0 and
154
169
<tr>
155
170
<td>
156
171
<p class="feature-summary">Metrics Viewer Role for CockroachDB Cloud</p>
157
-
<p class="feature-description" markdown="1">
172
+
<p class="feature-description">
158
173
This release introduces the Metrics Viewer role in CockroachDB Cloud, which provides dedicated permissions for collecting metrics and viewing insights. The role enhances security by enabling granular access to cluster performance data.
159
174
160
175
The Metrics Viewer role provides read-only access to cluster metrics and insights, without administrative or data modification privileges.
<li>Improved Security: Enables the principle of least privilege by allowing users or services to monitor cluster performance without broader access to sensitive data or configuration.</li>
165
180
<li>Targeted Access: Provides focused access to operational metrics, empowering teams to monitor health and performance efficiently.</li>
166
181
<li>Enhanced Collaboration: Facilitates secure collaboration by allowing different teams (e.g., SREs, developers) to access relevant metrics without over-provisioned permissions.</li>
0 commit comments