fix(smart-mind-map): prevent OWUI v0.10+ frontend crash on action early-return

[Fu-Jie]

Closes #101

Problem

On OpenWebUI v0.10+, clicking the Smart Mind Map action icon crashes the frontend:

Chat.svelte:1796 Uncaught (in promise) TypeError: Cannot read properties of undefined (reading 'content')
    at Se (Chat.svelte:1796:37)
    at async V (Messages.svelte:427:3)

The Docker log shows the action itself returns HTTP 200 in ~38 ms (no LLM call), yet the browser crashes.

Root cause (confirmed against OWUI main source)

I fetched the current OWUI main source to trace the exact failure path instead of guessing.

Frontend (Chat.svelte → chatActionHandler, Chat.svelte:1791-1802):

if (res !== null && res.messages) {
    for (const message of res.messages) {
        history.messages[message.id] = {
            ...history.messages[message.id],
            ...(history.messages[message.id].content !== message.content   // ← line 1796
                ? { originalContent: history.messages[message.id].content }
                : {}),
            ...message
        };
    }
}

The frontend looks up every returned message by id in history.messages. If history.messages[message.id] is undefined, accessing .content throws.

Plugin (before this fix) — all three early-return paths appended a new message dict without an id:

body["messages"].append({"role": "assistant", "content": "❌ ..."})
return body

When the frontend iterates that appended message, message.id is undefined, history.messages[undefined] is undefined, and the crash follows. This is exactly the 38 ms / HTTP 200 / no-LLM-call scenario in the issue: an early-return path was hit (empty/short content), the plugin appended an id-less message, and the frontend crashed while processing the otherwise-successful response.

Fix

Update the existing last message's content (it already carries a valid id) instead of appending a new id-less message. The OWUI frontend preserves the original content as originalContent before applying the update, so nothing is lost. The error is still surfaced to the user via the notification emitter (_emit_notification).

• New _set_last_message_content(body, content) helper with safe fallbacks (no-op when body has no messages, handles non-dict last entry, preserves id and other fields).
• Rewired all three early-return paths (no messages / no extractable content / text too short) to use it.
• Path 希望可以统一一下图标风格 #1 (messages is not a list) now simply returns body unchanged (cannot safely touch body["messages"]).
• Path Build MkDocs documentation portal with Material theme and GitHub Pages deployment #3 (text too short) preserves the original short text alongside the warning, mirroring the existing exception handler.
• Bumped version 1.0.1 → 1.0.2.

Tests

New regression test suite (tests/plugins/actions/smart-mind-map/test_issue_101_regression.py, 11 cases). Verified the tests fail on the old code (11 failures, including assert 3 == 2 showing the extra appended message, and Message without id would crash OWUI v0.10 frontend) and pass on the fix.

Key invariant enforced: every message in the returned body["messages"] must carry an id.

tests/plugins/actions/smart-mind-map/test_issue_101_regression.py ...... [ 54%]
.....                                                                    [100%]
11 passed

Issue #97 regression tests still pass (no break to markdown_normalizer):

tests/plugins/filters/markdown_normalizer/test_issue_97_regression.py ..   [ 33%]
....                                                                       [100%]
6 passed

Notes for the issue reporter

• The "No content found to export!" message you saw is the plugin's error_no_content notification — it means the last assistant message had no extractable text content at the moment the action ran. After this fix the crash is gone; if you still see the notification (without a crash), that's a separate content-availability question (e.g. the model response being too short, or stored in a structured-output format). Please retry on v1.0.2 and report back.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[github-actions]

✅ Plugin Version Check / 插件版本检查

版本更新检测通过！PR 包含版本变化和更新说明。

Version check passed! PR contains version changes and update description.

---

版本变化 / Version Changes

Plugin Updates

• Smart Mind Map: v1.0.1 → v1.0.2 | 📖 README

---

This comment was generated automatically. / 此评论由自动生成。

[gemini-code-assist]

Code Review

This pull request addresses a frontend crash (issue #101) on OpenWebUI v0.10+ by updating the last message's content in-place instead of appending a new id-less message, and adds a regression test suite. The feedback recommends making the message update helper more robust by ensuring an id field is always present (with a fallback), updating the regression tests to verify this fallback, and syncing the version bump to 1.0.2 across all required documentation files per the repository style guide.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

If last is not a dictionary, it is replaced with {"role": "assistant"}. However, this new dictionary does not contain an id field. Since the core issue (#101) is that the frontend crashes when a message lacks an id field, returning a message without an id will still trigger the same crash.

Furthermore, if last is a dictionary but somehow lacks an id field, it will also cause a crash.

To make this robust and defensive, we should ensure that the updated message always has an id field, falling back to body.get("id") (which represents the current assistant message ID) if it is missing.

Suggested change

	last = messages[-1]
	if not isinstance(last, dict):
	messages[-1] = {"role": "assistant"}
	messages[-1]["content"] = content
	last = messages[-1]
	if not isinstance(last, dict):
	last = {"role": "assistant"}
	messages[-1] = last
	if "id" not in last and isinstance(body, dict) and "id" in body:
	last["id"] = body["id"]
	last["content"] = content

[gemini-code-assist]

According to the repository style guide's Documentation Sync rule, when changing a plugin and bumping its version, you must update the version and release notes in all associated documentation files.

Currently, smart_mind_map.py has been bumped to 1.0.2, but README.md and README_CN.md still reference 1.0.1. Please ensure the following files are updated to maintain consistency:

1. plugins/actions/smart-mind-map/README.md
2. plugins/actions/smart-mind-map/README_CN.md
3. docs/plugins/actions/smart-mind-map.md
4. docs/plugins/actions/smart-mind-map.zh.md
5. docs/plugins/actions/index.md
6. docs/plugins/actions/index.zh.md

References

1. Must update all documentation files (README, README_CN, docs, index) with the new version when changing a plugin. ^(link)

[gemini-code-assist]

To align with the defensive fallback id logic added to _set_last_message_content, we should update this test to verify that the fallback id is correctly assigned when the last message is not a dictionary.

Suggested change

	def test_handles_non_dict_last_message(self):
	action = _make_action()
	body = {"messages": ["not-a-dict"]}
	action._set_last_message_content(body, "recovered")
	assert body["messages"][-1]["content"] == "recovered"
	def test_handles_non_dict_last_message(self):
	action = _make_action()
	body = {"id": "fallback-id", "messages": ["not-a-dict"]}
	action._set_last_message_content(body, "recovered")
	assert body["messages"][-1]["content"] == "recovered"
	assert body["messages"][-1]["id"] == "fallback-id"