UI layout and Chunking issues

[Yeon1007]

I have the following questions regarding UI layout and chunking issues and would like assistance:

1. Does RAGflow support modifying the current UI, such as changing places where avatars cannot be uploaded manually?

2. Regarding chunking, if I manually add chunks to a document's chunk result, and subsequent chats can retrieve from these manually added chunks (i.e., the embedding model embeds these manually added chunks) and will the TOC feature still work if it's enabled? @dosu

[dosubot[bot]]

1. RAGflow’s UI is built with React and supports manual avatar uploads. The AvatarUpload component lets users select, crop, and save avatar images via a file input. You can move or customize where avatars are uploaded by editing the React components and hooks in the frontend code—there are no backend restrictions on manual avatar upload, so you have full control over the UI layout (source, source, source).

2. When you manually add chunks to a document (via the API or UI), those chunks are immediately embedded using the dataset’s embedding model and indexed for retrieval. Subsequent chats and search operations can access these chunks just like any others (source, source). The TOC (Table of Contents) feature is generated during the initial document parsing and stored as a special chunk. If you add chunks manually after parsing, the TOC chunk will not automatically update to reflect the new structure; it only represents the chunk layout at the time of parsing. To update the TOC, you’d need to re-parse the document with TOC extraction enabled (source). The retrieval API’s TOC enhancement works with all indexed chunks as long as the TOC chunk exists (source).

If you want to change UI layout or chunking logic, you’ll need to edit the relevant frontend or backend code and rebuild. Let me know if you want pointers to specific files or functions!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[Magicbook1108]

Q1. We don't support UI modification. You might have to edit the code yourself.
Q2. I guess not, since toc is built during parsing. And the additional chunks are not considered in re-parsing. But you can have a try.

[Yeon1007]

After manually adding additional chunks, I encountered a problem generating the knowledge graph and it failed. Below is the log. How can I resolve this?
2025-12-01 15:18:17,857 WARNING 41 Error processing chunk 8/86: Task c959e672ce8111f08c0956d8ec84a7f7 was cancelled
2025-12-01 15:18:17,923 ERROR 41 set_progress(c959e672ce8111f08c0956d8ec84a7f7), progress: -1, progress_msg: 15:18:17 Error processing chunk 8/86: Task c959e672ce8111f08c0956d8ec84a7f7 was cancelled [Canceled], got exception
Traceback (most recent call last):
File "/ragflow/graphrag/general/extractor.py", line 129, in worker
await self._process_single_content(chunk_key_dp, idx, total, out_results, task_id)
File "/ragflow/graphrag/light/graph_extractor.py", line 95, in _process_single_content
glean_result = await trio.to_thread.run_sync(self._chat, "", history, gen_conf, task_id)
File "/ragflow/.venv/lib/python3.10/site-packages/trio/_threads.py", line 447, in to_thread_run_sync
return msg_from_thread.unwrap()
File "/ragflow/.venv/lib/python3.10/site-packages/outcome/_impl.py", line 213, in unwrap
raise captured_error
File "/ragflow/.venv/lib/python3.10/site-packages/trio/_threads.py", line 373, in do_release_then_return_result
return result.unwrap()
File "/ragflow/.venv/lib/python3.10/site-packages/outcome/_impl.py", line 213, in unwrap
raise captured_error
File "/ragflow/.venv/lib/python3.10/site-packages/trio/_threads.py", line 392, in worker_fn
ret = context.run(sync_fn, *args)
File "/ragflow/common/connection_utils.py", line 59, in wrapper
raise result
File "/ragflow/common/connection_utils.py", line 43, in target
result = func(*args, **kwargs)
File "/ragflow/graphrag/general/extractor.py", line 78, in _chat
raise TaskCanceledException(f"Task {task_id} was cancelled")
common.exceptions.TaskCanceledException: Task c959e672ce8111f08c0956d8ec84a7f7 was cancelled
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/ragflow/rag/svr/task_executor.py", line 158, in set_progress
raise TaskCanceledException(msg)
common.exceptions.TaskCanceledException: 15:18:17 Error processing chunk 8/86: Task c959e672ce8111f08c0956d8ec84a7f7 was cancelled [Canceled]
@dosu @Magicbook1108

[dosubot[bot]]

The "Task was cancelled" error during knowledge graph generation usually means the task was flagged as cancelled due to a timeout, manual cancellation, or exceeding the allowed error threshold during chunk processing. This is checked frequently in the async pipeline, and if triggered, all ongoing chunk workers are stopped immediately, which matches your log output (source, source, source).

Here’s what you can check and try:

• Make sure your manually added chunks have all required fields (like content_with_weight, kb_id, doc_id, etc.) and are formatted the same as chunks created by the standard pipeline. Malformed or missing fields can cause processing errors or slowdowns.
• If your manual chunks are very large or complex, they may slow down processing and cause the task to exceed its timeout (which scales with chunk count). Try reducing chunk size or complexity.
• Check if the task was manually cancelled or if there were resource issues (like concurrency limits or distributed lock contention) that could have triggered cancellation.
• If you see repeated errors on certain chunks, the process will abort if the error count exceeds the threshold (default is 3). Review logs for earlier errors before the cancellation message.

If you can share how you added the chunks (API, UI, or direct DB), or if you have more logs from before the cancellation, I can help narrow it down further.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[Magicbook1108]

It looks like you manually close the generating process.