Problem statement
With Knowhere's migration to vision-model-based, page-level document parsing, the current UI lacks a dedicated interface to browse parsed page results and connect them to inline citations in answers. Users have no way to inspect what was extracted from each page, navigate from a citation directly to the parsed source, or see a structured breakdown of referenced sources at the end of an answer.
Affected area
Other
Proposed solution
-
The section is titled "Parsed Results", with a subtitle displaying the source file name in the format "from [filename]".
-
The new Brain view retains only the "List" and "Tree" toggle in the top-right corner.
-
Each parsed page card is simplified to show only:
3.1 Card title: page number + parsing path
3.2 Page image rendering
3.3 Keywords
-
Inline citation chips are required in answers. Citation format: filename / page. Each chip has a max-width of 250px; if the filename exceeds this, it should be truncated. Each citation generates one chip corresponding to the specific referenced page. Clicking a chip navigates to the corresponding page in the Parsed Results view.
-
The source list at the end of an answer uses a numbered list. Each file is one list item, displayed as the filename followed by all page numbers cited from that file.
-
Rename "Knowhere Notebook" in the top-left corner to "Knowhere Brain".
-
Responses, code blocks, and tables can be copied directly. You can also download this conversation turn as Markdown or PDF (without citations).
Alternatives considered
No response
Compatibility and migration notes
No response
Additional context
No response
Problem statement
With Knowhere's migration to vision-model-based, page-level document parsing, the current UI lacks a dedicated interface to browse parsed page results and connect them to inline citations in answers. Users have no way to inspect what was extracted from each page, navigate from a citation directly to the parsed source, or see a structured breakdown of referenced sources at the end of an answer.
Affected area
Other
Proposed solution
The section is titled "Parsed Results", with a subtitle displaying the source file name in the format "from [filename]".
The new Brain view retains only the "List" and "Tree" toggle in the top-right corner.
Each parsed page card is simplified to show only:
3.1 Card title: page number + parsing path
3.2 Page image rendering
3.3 Keywords
Inline citation chips are required in answers. Citation format: filename / page. Each chip has a max-width of 250px; if the filename exceeds this, it should be truncated. Each citation generates one chip corresponding to the specific referenced page. Clicking a chip navigates to the corresponding page in the Parsed Results view.
The source list at the end of an answer uses a numbered list. Each file is one list item, displayed as the filename followed by all page numbers cited from that file.
Rename "Knowhere Notebook" in the top-left corner to "Knowhere Brain".
Responses, code blocks, and tables can be copied directly. You can also download this conversation turn as Markdown or PDF (without citations).
Alternatives considered
No response
Compatibility and migration notes
No response
Additional context
No response