@@ -11,43 +11,6 @@ For human-friendly, themed highlights see [Release Notes](docs/RELEASE_NOTES.md)
 
 ## [Unreleased]
 
-### Added
-
-- **JetBrains: parity for index freshness, always-on retrieval, and multi-session
-  history** (plugin 0.4.14): auto re-index `@workspace` after a save (debounced;
-  toggle in Settings ▸ Tools ▸ CodeSetu), an "always retrieve every turn" option,
-  and a **Chat history…** switcher (composer menu) backed by saved sessions —
-  New chat now starts a fresh session and the old one is kept. Settings UI gains
-  a **@workspace** group (embedding base URL/model + the two toggles). Brings
-  JetBrains level with VS Code 0.4.14.
-- **Auto re-index `@workspace` on save** (VSCode 0.4.14): when an index already
-  exists, CodeSetu re-indexes incrementally a few seconds after you save a file,
-  so retrieval stays fresh without re-running the command. Toggle with
-  `codesetu.workspaceIndex.autoReindex` (default on). It never auto-builds — only
-  refreshes an index you've already created.
-- **Always-on `@workspace` retrieval** (VSCode 0.4.14): set
-  `codesetu.workspaceIndex.alwaysRetrieve` to pull relevant indexed code into
-  every chat turn, not only when you type `@workspace`. Retrieves silently when
-  an index exists (never triggers a build).
-- **JetBrains: full composer parity with VS Code** (plugin 0.4.13): Agent Mode
-  is now the **default** for new chats; a **version badge** (`v<x.y.z>`) sits by
-  the model picker; typing `@` offers a dedicated **@workspace** entry (inserts
-  the literal mention instead of pinning a file); and the slash palette includes
-  **`/edit`**, which runs the Edit with CodeSetu diff flow (per-hunk
-  accept/reject) on the active editor using the text after the command.
-- **JetBrains: Context & activity panel** (plugin 0.4.12): the per-turn panel
-  (renamed from "Context sent to AI") now shows the **provider · model ·
-  endpoint**, whether **Agent Mode** is on, an expandable **Tools available (N)**
-  list, and the **@workspace** outcome (retrieved N chunks with the file hits, or
-  the failure reason) plus pinned/retrieved counts — matching VS Code. (Tool
-  calls actually _used_ during a turn already stream inline as `🔧` lines.)
-- **JetBrains: `@workspace` auto-builds the index on first use** (plugin 0.4.11):
-  typing `@workspace …` when no index exists now builds it automatically before
-  retrieving (off the EDT), instead of silently returning nothing — matching the
-  VS Code behaviour. A down/misconfigured embeddings endpoint surfaces an error
-  in the chat rather than breaking the turn. The **Index Workspace** command
-  remains for explicit rebuilds.
-
 ### Planned for v0.1
 
 - Inline FIM completions via Sarvam-30B
@@ -59,6 +22,47 @@ For human-friendly, themed highlights see [Release Notes](docs/RELEASE_NOTES.md)
 
 ---
 
+## [0.4.14] - 2026-07-01
+
+VSCode and JetBrains are aligned at 0.4.14.
+
+### Added
+
+- **Auto re-index `@workspace` on save** (VSCode + JetBrains): when an index
+  already exists, CodeSetu re-indexes incrementally a few seconds after you save
+  a file, so retrieval stays fresh without re-running the command. Toggle:
+  `codesetu.workspaceIndex.autoReindex` (VSCode) / Settings ▸ Tools ▸ CodeSetu
+  (JetBrains); default on. Never auto-builds — only refreshes an existing index.
+- **Always-on `@workspace` retrieval** (VSCode + JetBrains):
+  `codesetu.workspaceIndex.alwaysRetrieve` pulls relevant indexed code into every
+  chat turn, not only when you type `@workspace`. Retrieves silently when an
+  index exists (never triggers a build).
+- **Multi-session chat history in JetBrains**: a **Chat history…** switcher
+  (composer menu) backed by saved sessions; New chat starts a fresh session and
+  keeps the old one. (VSCode already had this.)
+- **Full composer parity in JetBrains**: Agent Mode is the **default** for new
+  chats; a **version badge** (`v<x.y.z>`) sits by the model picker; typing `@`
+  offers a dedicated **@workspace** entry (inserts the literal mention instead of
+  pinning a file); the slash palette includes **`/edit`**, which runs the Edit
+  with CodeSetu diff flow (per-hunk accept/reject) on the active editor.
+- **Context & activity panel in JetBrains**: the per-turn panel shows the
+  **provider · model · endpoint**, whether Agent Mode is on, an expandable
+  **Tools available (N)** list, and the **@workspace** outcome (retrieved files,
+  or the failure reason) plus pinned/retrieved counts — matching VS Code.
+- **JetBrains: `@workspace` auto-builds the index on first use**: typing
+  `@workspace …` with no index builds it automatically before retrieving (off the
+  EDT) instead of silently returning nothing; a bad embeddings endpoint surfaces
+  an error rather than breaking the turn. The **Index Workspace** command remains
+  for explicit rebuilds.
+
+### Fixed
+
+- **JetBrains: version badge no longer uses an internal API** — read the version
+  from the shipped `plugin.xml` instead of `PluginManagerCore.getPlugin`, clearing
+  the Plugin Verifier "internal method usage" finding.
+
+---
+
 ## [0.4.10] - 2026-06-24
 
 ### Added
 
@@ -2,8 +2,6 @@ package ai.codesetu.toolwindow
 
 import ai.codesetu.settings.CodeSetuSettingsState
 import ai.codesetu.skills.loadBuiltinSkills
-import com.intellij.ide.plugins.PluginManagerCore
-import com.intellij.openapi.extensions.PluginId
 import java.net.URI
 import java.nio.charset.StandardCharsets
 
@@ -33,9 +31,19 @@ object ChatWebviewHtml {
       .replace("__VERSION__", escapeHtml(pluginVersion()))
   }
 
-  /** The installed plugin version, shown as a badge so a stale build is obvious. */
-  private fun pluginVersion(): String =
-    PluginManagerCore.getPlugin(PluginId.getId("ai.codesetu.codesetu"))?.version ?: ""
+  /**
+   * The plugin version, read from the shipped plugin.xml (patched with the real
+   * version at build time). Avoids the internal `PluginManagerCore.getPlugin`
+   * API flagged by the Plugin Verifier; shown as a badge so a stale build is
+   * obvious.
+   */
+  private val version: String by lazy {
+    val xml = ChatWebviewHtml::class.java.getResourceAsStream("/META-INF/plugin.xml")
+      ?.use { String(it.readAllBytes(), StandardCharsets.UTF_8) } ?: return@lazy ""
+    Regex("<version>([^<]+)</version>").find(xml)?.groupValues?.get(1)?.trim() ?: ""
+  }
+
+  private fun pluginVersion(): String = version
 
   private fun speechConfigJson(state: CodeSetuSettingsState.State): String =
     "{\"sttProvider\":${jsonString(state.speechSttProvider)}," +
 
@@ -57,6 +57,15 @@
   ]]></description>
 
   <change-notes><![CDATA[
+    <h3>0.4.14</h3>
+    <ul>
+      <li><b>@workspace — semantic codebase search</b> — index your repo and retrieve code by meaning. Auto-builds on first use, auto re-indexes after you save, and an optional "always retrieve every turn" mode. Runs on any OpenAI-compatible embeddings endpoint.</li>
+      <li><b>Agent Mode is now the default</b>, and works with models that lack native tool-calling (e.g. Gemma) via prompted tool-calls.</li>
+      <li><b>Context &amp; activity panel</b> — every turn shows the provider · model · endpoint, the tools available, and the @workspace results; a version badge sits by the model picker.</li>
+      <li><b>/edit</b> — rewrite a selection/file from an instruction with a diff preview and per-hunk accept/reject (command and <code>/edit</code> slash).</li>
+      <li><b>@folder pinning</b> and a <b>multi-session chat history</b> switcher.</li>
+      <li><b>On-prem Docker installer</b> — self-host chat + embeddings fully inside your network (air-gapped friendly).</li>
+    </ul>
     <h3>0.4.1</h3>
     <ul>
       <li><b>Inline FIM completions</b> — ghost-text suggestions as you type, debounced and cached so re-triggering at the same spot avoids a redundant model call.</li>