> For the complete documentation index, see [llms.txt](https://docs.datasentinel.io/manual/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.datasentinel.io/manual/changelog/2026.md).
# 2026
All releases are cumulative. Upgrading to the latest version includes all previous changes.It's sufficient to install or run only the most recent release.
{% hint style="info" %}
Please refer to the documentation to [**upgrade Datasentinel**](https://docs.datasentinel.io/implementation/upgrade/platform)**.**
{% endhint %}
{% updates format="full" %}
{% update date="2026-11-02" tags="feature,improvement,fix,changes" %}
## V2026.10
This release introduces [**AI Connector**](https://docs.datasentinel.io/ai-connector), a new way to work with Datasentinel. Next to the web console and the REST API, you can now use AI assistants: ask in plain language and get findings backed by your metrics, through an MCP server and a Skill.
A new way to work with Datasentinel
This release also brings a deep rework of the [**PDF reports**](https://docs.datasentinel.io/features/key-features/reporting) and a set of bug fixes.
### :sparkles: AI Connector
> [**AI Connector**](https://docs.datasentinel.io/ai-connector) opens Datasentinel metrics to AI assistants. Ask a question in plain language: the assistant analyzes your PostgreSQL workload, explains what it finds, and turns the evidence into concrete recommendations.
>
> It is made of two parts: an **MCP server**, running on each platform server, which exposes Datasentinel metrics over the [Model Context Protocol (MCP)](https://modelcontextprotocol.io/), and a **Skill**, which tells the assistant which tools to call and how to turn the results into analyses and recommendations. Any assistant that supports MCP and Skills can be used ([**connect your assistant**](https://docs.datasentinel.io/ai-connector/getting-started/connect-a-compatible-assistant)): Claude (desktop, Claude Code), OpenAI Codex and ChatGPT, GitHub Copilot (CLI and VS Code), Microsoft Copilot Studio, Cursor, Mistral Vibe, Google Antigravity, and more. The LLM behind the assistant is your choice.
>
> * **What you can do**: performance analysis in the past (workload, top queries, wait events, blocking scenarios, tables and indexes, WAL, replication, host metrics), real-time troubleshooting through [**Live360**](https://docs.datasentinel.io/features/key-features/live360), workload comparison, actionable recommendations, and [**PDF reports**](https://docs.datasentinel.io/ai-connector/generate-reports). The Skill adapts its communication style to the user's profile (**Data admin** or **Developer**) and replies in the user's language. See [**What you can analyze**](https://docs.datasentinel.io/ai-connector/what-you-can-analyze) and the [**Skill commands reference**](https://docs.datasentinel.io/ai-connector/skill-commands).
> * **Security**: **OAuth 2.1 sign-in** with the web interface credentials, the same [**RBAC**](https://docs.datasentinel.io/implementation/platform-usage/configuration/users-and-roles) rights as the web interface, **read-only** access, and the assistant never knows any PostgreSQL credential. Scripts and scheduled jobs use a [**user token**](https://docs.datasentinel.io/implementation/platform-usage/api-reference/access-token). See [**How authentication works**](https://docs.datasentinel.io/ai-connector/getting-started/how-authentication-works), [**Session administration**](https://docs.datasentinel.io/ai-connector/getting-started/manage-assistant-sessions) and the [`mcp.log`](https://docs.datasentinel.io/ai-connector/troubleshooting/read-the-mcp-log) audit.
> * **Licensing**: AI Connector is an **add-on**, enabled for the whole installation by a licence token that includes the AI Connector option. Contact [**Datasentinel**](mailto:contact@datasentinel.io) to upgrade your licence.
> * **Deployment**: the MCP server runs as a new `datasentinel_mcp` service on each platform server, behind nginx on the `/mcp` path. No maintenance is required. See [**MCP Server Setup**](https://docs.datasentinel.io/ai-connector/getting-started/mcp-server-setup).
>
> For the most precise analyses, upgrade local agents to [**Agent v4.0.0**](https://docs.datasentinel.io/implementation/agent-usage/release-notes) or later and install the [**internal function**](https://docs.datasentinel.io/getting-started/postgresql-clusters/internal-function) in monitored databases.
### Documentation MCP server
> The Datasentinel documentation is published with GitBook, and it now comes with its own MCP server, hosted by GitBook. Register it in your assistant next to your platform, and the assistant answers product questions from the manual while it works on your databases: what a setting does, how to configure an alert, what an API route returns.
>
> * **URL**: `https://docs.datasentinel.io/manual/~gitbook/mcp`, Streamable HTTP transport.
> * **No sign-in**: the documentation is public, and the server is read-only and never sees your data.
> * **Name it `docs-datasentinel`**: the `datasentinel-` prefix is reserved by the Skill for monitoring platforms.
>
> Refer to [**Explore the documentation from your assistant**](https://docs.datasentinel.io/ai-connector/explore-the-documentation).
### Internal function
> The new optional [**`datasentinel_analytics` function**](https://docs.datasentinel.io/getting-started/postgresql-clusters/internal-function) gives Datasentinel read-only access to planner statistics and execution plans, without any SELECT privilege on application tables. It enables table and index bloat estimates, column and extended statistics, and live execution plans. It never returns application data. It is recommended for AI Connector analyses.
>
> Install it in each application database with the `deploy_datasentinel_function.sh` script shipped with the agent. Local agents must be upgraded to [**Agent v4.0.0**](https://docs.datasentinel.io/implementation/agent-usage/release-notes) or later.
### Reporting
> * **New layout**: all [**PDF reports**](https://docs.datasentinel.io/features/key-features/reporting) are more compact, with harmonized colors and charts.
> * **Workload report**: the header shows the **load profile** and the **top wait events**.
> * **Blocking scenarios** and **wait events** are now included in the reports.
> * **Tables and indexes**: more per-table details (bloat, XID age, options), **column** and **extended statistics**, and rankings by several criteria.
> * **Instance report**: lists the **collected parameters** in addition to the [**changed parameters**](https://docs.datasentinel.io/features/other-features/changed-parameters).
> * **Performance**: lower memory usage and faster generation on large fleets.
### Inactive instances
> Instances with no data left in the metrics database can now be purged automatically. An instance is inactive when it stopped sending metrics for longer than the retention period. When the purge is enabled in the [**backend configuration**](https://docs.datasentinel.io/implementation/platform-usage/configuration/backend#purge-of-inactive-instances), the platform removes these instances from the inventory once a day. The purge is disabled by default.
>
> The new `purge_inactive_instances.sh` script lists these instances and can purge them right away. Refer to [**Tooling**](https://docs.datasentinel.io/implementation/platform-usage/tooling#inactive-instances).
### Agentless monitoring: query filters
> The query monitoring filters can now be set once for all agentless instances, in the `agent` section of the backend configuration file:
>
> * `query-monitoring-min-calls`: minimum number of calls (default **2**, lowest value 2);
> * `query-monitoring-min-time`: minimum total execution time, in seconds (default **1**, lowest value 0).
>
> A query of `pg_stat_statements` is collected if it reaches either value. Set `query-monitoring-min-time` to **0** to collect every query. Restart the dispatcher after a change.
>
> Refer to [**Backend configuration**](https://docs.datasentinel.io/implementation/platform-usage/configuration/backend#agentless-agents).
### Diagnostic scripts
> The diagnostic file now includes a **backend configuration** section.\
> \
> Refer to [**Diagnostic Tools**](https://docs.datasentinel.io/implementation/troubleshooting/platform-diagnostics-tools) section of the documentation.
### Embedded PostgreSQL
> The PostgreSQL instance embedded in the platform (the Grafana metadata database) moves from **15.5** to [**15.19**](https://www.postgresql.org/docs/15/release-15-19.html), published by the PostgreSQL community on 2026-08-13. It includes the bug and security fixes of all the 15.x minor releases since 15.5.
>
> * **New installations** come with PostgreSQL 15.19.
> * **Existing platforms**: the patch detects the older version and asks you to run the new `update_postgresql.sh` script, shipped in the patch archive, as the Datasentinel owner. It is a minor update: the binaries are replaced, the data is kept as is, and the previous binaries are restored automatically if the new instance does not start. Grafana is unavailable for about a minute; metrics collection and alerting are not affected. You choose when to run it.
>
> Refer to [**Upgrade the platform**](https://docs.datasentinel.io/implementation/upgrade/platform).
>
> The full list of fixes is in the official [**PostgreSQL 15.19 release notes**](https://www.postgresql.org/docs/15/release-15-19.html), and the related CVEs on the [**PostgreSQL security page**](https://www.postgresql.org/support/security/).
### Alerting
> The global alerting settings are now kept in memory for 60 seconds, which reduces the load on the metrics database. A change to the settings takes effect within a minute. Each change is recorded in `alerting.log`, with the user who saved it.
### Security
> Third-party dependencies have been updated to address known security vulnerabilities.
### :wrench: Fixes
> * **Agentless monitoring: platform self-monitoring**\
> After an unexpected restart of its agent, the platform server's own host metrics were no longer collected. They now resume automatically.
> * **Agentless monitoring: host metrics after an agent restart**\
> When an agent of the pool restarted, the connections moved to the new agent lost their host (OS) monitoring until the pool was restarted. It is now restored automatically. A disabled connection carrying the host monitoring also left its host without OS metrics; the host is now monitored again as soon as the connection is enabled.
> * **Agentless monitoring: empty `agent` section**\
> An `agent` section left empty in the backend configuration file prevented the agentless agents from starting. The section is now ignored and the default values are used.
> * **Live sessions: query texts**\
> Sessions without a query (idle in transaction blockers, background processes) triggered a useless query text lookup in the metrics database. It is now skipped.
> * **Top Queries: missing query texts**\
> Some query texts retrieved from the PostgreSQL instance (with double quotes or a trailing backslash) could not be stored. They are now stored correctly.
> * **Full report generation**\
> On a fresh installation, generating a report could fail with `'>' not supported between instances of 'NoneType' and 'int'` when index sizes were not collected yet.
> * **Top Queries summary**\
> The summary failed when no data was found for the selected period.
> * **Query details: wrong query text**\
> When no `pg_stat_statements` metrics were stored for the selected period, the header of the query details dashboards could display the text of another query. It now displays the text of the selected query, or no text if none is stored.
> * **Negative sizes**\
> Negative size values are no longer displayed in reports.
> * **LDAP users**\
> Fixed an error affecting [**users authenticated through LDAP**](https://docs.datasentinel.io/implementation/platform-usage/configuration/ldap).
> * **Backend configuration**\
> The `apis.max_requests` setting of `datasentinel.yml` was ignored and replaced by its default value.
> * **Diagnostic scripts**\
> Instances with no reported query limit were counted as having reached their daily limit.
> {% endupdate %}
{% update date="2026-07-10" tags="fix,changes" %}
## V2026.07
### Security
> * Updated bundled OpenSSL 3.0.14 → 3.0.21, addressing CVE-2026-45447 and related upstream CVEs (defense-in-depth; not exploitable in normal Datasentinel usage).
> * Third-party dependencies have also been updated to address known security vulnerabilities.
### :wrench: Fixes
> * **Agentless monitoring**\
> Connections pointing to an unreachable instance can now be deleted from the UI and the API. Previously the dispatcher's delete request timed out against an agent blocked on hung database I/O. The agent completed the deletion while the dispatcher aborted, leaving its instance pool out of sync and the connection stuck. The dispatcher now cleans up its own state whether the agent times out or answers 404 – Connection not found.
> {% endupdate %}
{% update date="2026-06-25" tags="changes" %}
## V2026.06
### Alerting
> **Faster parsing**
### Agent v3.9.3
> This release integrates [**Agent v3.9.3**](https://docs.datasentinel.io/implementation/agent-usage/release-notes) which includes two fixes related to tag handling.
> {% endupdate %}
> {% endupdates %}
{% updates format="full" %}
{% update date="2026-05-21" tags="improvement,fix,changes,fixes" %}
## V2026.05
### Diagnostic scripts
> The diagnostic file now includes internal InfluxDB performance metrics from the last 24 hours, providing deeper insight for troubleshooting and support analysis.\
> \
> Refer to [**Diagnostic Tools**](https://docs.datasentinel.io/implementation/troubleshooting/platform-diagnostics-tools) section of the documentation.
### Top queries
> In rare cases, a query text may not be loaded into the platform. Datasentinel now automatically attempts to retrieve any missing query texts directly from the source PostgreSQL instance when displaying the [**Top Queries**](https://docs.datasentinel.io/features/key-features/top-queries) dashboard
### Agent v3.9.2
> This release integrates [**Agent v3.9.2**](https://docs.datasentinel.io/implementation/agent-usage/release-notes) which includes two fixes related to tag handling.
### :wrench: Fixes
> * **Full Report generation**\
> Generating a [**Report** ](https://docs.datasentinel.io/features/key-features/reporting)could occasionally fail with a `list index out of range` error.
> * **Top Queries export**\
> Exporting [**Top Queries**](https://docs.datasentinel.io/features/key-features/top-queries) data in JSON format did not honor the **per execution** flag.\
> Resolved an error that occurred when exporting data in CSV format.
> {% endupdate %}
{% update date="2026-02-17" tags="improvement,changes" %}
## V2026.02
### UI Layout Enhancement
> Improved layout structure to completely fill the screen and enhance high-resolution display support.
### Agent v3.9.1
> This release integrates [**Agent v3.9.1**](https://docs.datasentinel.io/implementation/agent-usage/release-notes)
> {% endupdate %}
{% update date="2026-01-13" tags="feature,improvement,fix,changes,fixes" %}
## V2026.01
### Backend pool
> Backend process count and restart limits are now configurable.\
> Refer to [**Backend configuration**](https://docs.datasentinel.io/implementation/platform-usage/configuration/backend) section for more details.
### Diagnostic Scripts
> Enhanced diagnostic scripts with additional system and process metrics.\
> Added a new script to compress Datasentinel log files for support purposes.\
> \
> Refer to [**Diagnostic Tools**](https://docs.datasentinel.io/implementation/troubleshooting/platform-diagnostics-tools) section of the documentation.
### Alerting
> A dedicated database is now used for alerts to improve performance and avoid sharing resources with the metrics database.
### Libraries
> Updated third-party backend libraries.
### :wrench: Fixes
> * Enhanced error and exception handling for metrics database operations.
> {% endupdate %}
> {% endupdates %}
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.datasentinel.io/manual/changelog/2026.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
