From bc17e55899793956d61525297fd086f1be88d599 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Thu, 9 Jul 2026 00:34:34 +0000
Subject: [PATCH] =?UTF-8?q?=F0=9F=A6=8B=20New=20version=20release?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .changeset/tool-activity-blocks.md |  9 ---------
 CHANGELOG.md                       | 10 ++++++++++
 package.json                       |  2 +-
 3 files changed, 11 insertions(+), 10 deletions(-)
 delete mode 100644 .changeset/tool-activity-blocks.md

diff --git a/.changeset/tool-activity-blocks.md b/.changeset/tool-activity-blocks.md
deleted file mode 100644
index 417c271..0000000
--- a/.changeset/tool-activity-blocks.md
+++ /dev/null
@@ -1,9 +0,0 @@
----
-'@smooai/chat-widget': minor
----
-
-Add an opt-in `showToolActivity` option that renders the agent's tool activity (grep / read_file / bash / knowledge_search…) as inline chips interleaved with its prose, mirroring the smooth daemon SPA's `blocks` model.
-
-- Defaults to **`false`** — a customer-facing support widget stays byte-for-byte unchanged (prose bubble only). Enable via the `show-tool-activity` HTML attribute or `showToolActivity: true` in the config.
-- When enabled, an assistant turn that invokes a tool renders as an ordered strip of prose bubbles + tool chips (running… / done ✓ / error), in the order the model produced them. Tool activity is read from `state.rawResponse.toolCall` / `state.rawResponse.toolResult` (the correct nested path — reading `state.toolResult` leaves chips stuck "running…").
-- Tool name/args are rendered via `textContent` (never `innerHTML`), so a tool payload can't inject markup. New public `MessageBlock` / `ToolCall` types.
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 949faf2..8c14d1c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,15 @@
 # Changelog
 
+## 0.13.0
+
+### Minor Changes
+
+- 92b31ad: Add an opt-in `showToolActivity` option that renders the agent's tool activity (grep / read_file / bash / knowledge_search…) as inline chips interleaved with its prose, mirroring the smooth daemon SPA's `blocks` model.
+
+  - Defaults to **`false`** — a customer-facing support widget stays byte-for-byte unchanged (prose bubble only). Enable via the `show-tool-activity` HTML attribute or `showToolActivity: true` in the config.
+  - When enabled, an assistant turn that invokes a tool renders as an ordered strip of prose bubbles + tool chips (running… / done ✓ / error), in the order the model produced them. Tool activity is read from `state.rawResponse.toolCall` / `state.rawResponse.toolResult` (the correct nested path — reading `state.toolResult` leaves chips stuck "running…").
+  - Tool name/args are rendered via `textContent` (never `innerHTML`), so a tool payload can't inject markup. New public `MessageBlock` / `ToolCall` types.
+
 ## 0.12.0
 
 ### Minor Changes
diff --git a/package.json b/package.json
index 87e70b0..22a47d3 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
     "name": "@smooai/chat-widget",
-    "version": "0.12.0",
+    "version": "0.13.0",
     "description": "Embeddable AI chat as a framework-light web component — the Aurora Glass design, streaming replies, grounded sources, and per-brand theming. Speaks the smooth-operator WebSocket protocol.",
     "license": "MIT",
     "author": "SmooAI",