Feature(aidp): Knowledge base details page by mew-boo · Pull Request #4096 · ModelEngine-Group/nexent

mew-boo · 2026-10-14T09:03:33Z

AIDP knowledge base details page

Changes

The File tab is renamed as Uploaded Files.
The function of deleting files in batches is added to the Uploaded Files tab. A maximun of 10 files can be deleted at a time.
The Upload Task tab is added.
The Details tab is added.

Screenshots

cj2026-bit

I verified the detail page against the deploy worktree and its AIDP mock. The inline comments below describe reproducible behavior and places where the mock diverges from the AIDP response contract previously observed.

cj2026-bit · 2026-10-14T09:45:15Z

+            server_url,
+            api_key,
+            channel["fs_id"],
+            f"/{kds_id}",


The selected channel includes its source directory, but this request constructs /{kds_id} instead. I reproduced this on the PR commit: the Upload Tasks tab showed 0 rows; the backend sent /aidp-kb-faq, which the mock logged as outside the KB and returned empty. Posting History with the channel src_dir (/aidp/knowledge/aidp-kb-faq) returned 10 rows (22 total). Please pass the selected channel's actual source directory here.

cj2026-bit · 2026-10-14T09:45:15Z

+        )
+    raw_items = history.get("value") if isinstance(history, dict) else None
+    rows = [item for item in raw_items if isinstance(item, dict)] if isinstance(raw_items, list) else []
+    raw_total = history.get("total_count") if isinstance(history, dict) else None


The AIDP History response contract provided for this integration uses total_count for the current page and total_record_count for the global total (for example, 3 vs. 317). Marking the page-local value as reliable makes AidpPagination treat 3 as the global total, so it can disable next-page navigation even when next_link is present. The included mock returns a filtered all-pages count here, which masks the mismatch. Please use the AIDP paging signal without treating total_count as a global total.

cj2026-bit · 2026-10-14T09:45:16Z

+        "extracting": _history_count("processing_record_count"),
+        "failed": _history_count("failed_record_count"),
+        "success": _history_count("success_record_count"),
+        "queued": _history_count("queued_record_count"),


The observed AIDP History payload exposes waiting_record_count, while this adapter reads queued_record_count. On the real response the queued summary will therefore stay at zero when files are waiting. The PR mock emits queued_record_count, so it currently hides this mismatch. Please map the actual AIDP field to the UI's queued statistic.

cj2026-bit · 2026-10-14T09:45:16Z

+            lambda: client.post(
+                retry_url,
+                headers=headers,
+                json={"file_uuids": file_uuids},


The AIDP Retry contract previously tested for this integration requires retry_file_info: [{"file_uuid": "..."}]; sending file_uuids returns HTTP 422 because retry_file_info is missing. The PR mock accepts file_uuids, so local success would not catch the live failure. Please align both the adapter and mock with AIDP's request body.

cj2026-bit · 2026-10-14T09:45:16Z

+                lane="control-io",
+                owner="config",
+            ),
+            return_exceptions=True,


GET /knowledge-bases/{kds_id}/documents already exists with the same page, page_size, and keyword parameters and calls Count. This new /files endpoint duplicates that path, but return_exceptions=True turns a Count failure into HTTP 200 with an estimated total. That hides the request failure and can show a misleading total. Please reuse the existing documents endpoint (or move the needed changes there) and let Count failures fail the request.

cj2026-bit · 2026-10-14T09:45:16Z

+            {formatDateTime(detail.updated_at)}
+          </AidpDetailField>
+          <AidpDetailField label={t("aidpKnowledge.detailCreator")}>
+            {detail.created_by || detail.user_name || AIDP_UNKNOWN_VALUE}


created_by is the Nexent permission record's owner_user_id; the API also provides creator_name for display. Prioritizing created_by makes this field show an opaque ID -- I reproduced the detail page displaying the literal user_id -- instead of the Nexent user name. Please render creator_name, falling back to -- when it is absent, rather than showing the ID.

cj2026-bit

The PR checks currently fail on two actionable issues: changed files are missing the required copyright header, and the updated unavailable-KB test expects HTTP 200 although the handler still calls Count after the detail request fails. Inline comments identify both.

cj2026-bit · 2026-10-14T10:25:12Z

 from __future__ import annotations

 import asyncio
+import json


The copyright-check gate reports this file and 22 other changed files as missing the required copyright header. Please add the standard header to all 23 files reported by the check; otherwise this required PR gate continues to fail.

cj2026-bit · 2026-10-14T10:25:12Z

+        # The detail view degrades instead of failing: the count fetch only
+        # runs for ACTIVE knowledge bases, so a downed AIDP detail + count no
+        # longer escalates into a 5xx detail response.
+        assert response.status_code == HTTPStatus.OK


This new expectation does not match the handler: after the detail call fails, get_knowledge_base still calls count_aidp_docs_impl unconditionally. With this test mocked Count failure, that exception escapes and the endpoint returns 502, which is the CI failure. Please either skip Count when the KB is already marked UNAVAILABLE, so the expected 200/no-count contract is real, or change the test to assert the intended failure behavior.

…ctionality in detailed information section

…he issues raised in the review

cj2026-bit

Rechecked the eight previous review comments against c5700e3. Six are addressed; the pagination comment and the duplicated /files-path comment still need the two corrections below. The two backend test modules pass locally (422 tests), but the isolated regression probes below expose behavior those tests do not cover.

cj2026-bit · 2026-10-15T09:30:06Z

+    raw_global_total = history.get("total_record_count") if isinstance(history, dict) else None
+    total_reliable = isinstance(raw_global_total, int) and not isinstance(raw_global_total, bool)
+    total = int(raw_global_total) if total_reliable else (page - 1) * page_size + len(rows)


[P2] Keep the global History statistic out of filtered pagination

This fixes the page-local total_count problem for the unfiltered list, but total_record_count is also used as a reliable total after keyword/status filtering. The PR mock computes this statistic before applying either filter. AidpPagination ignores hasMore when totalReliable is true, so a filtered last page can still expose many empty pages.

Reproduced against c5700e3 with an isolated History response: one matching row, total_record_count=317, next_link=null. The endpoint returns total_count=317, total_reliable=true, has_more=false; rendering the actual pagination component produces 32 pages with Next enabled.

For filtered queries, please use total_reliable=false and the upstream paging signal unless AIDP supplies a confirmed filtered total. Keep total_record_count in the summary statistics. Add a regression where one filtered match and no next_link leave Next disabled.

Revised, please check if it meets expectations.

cj2026-bit · 2026-10-15T09:30:07Z

+    if len(rows) < page_size:
+        has_more = False


[P2] Do not override an explicit next_link just because the page is short

The /files route is still present and used by listIngestedFiles. Count failures now propagate, but this remaining branch makes its paging behavior differ from /documents: a short page unconditionally clears has_more even when AIDP explicitly supplies the next-page link.

Reproduced against c5700e3 using the same isolated upstream response for both routes: page_size=10, two rows, next_link="/next-page", keyword search. /files returns has_more=false, while /documents returns true. The detail page therefore prevents users from reaching the remaining search results.

Please reuse the existing /documents paging implementation, or infer the last page from row count only when no explicit paging signal is available. The test test_file_list_suppresses_phantom_next_link_on_short_page currently asserts this truncation as correct; change it to protect the upstream next_link instead.

…into develop_st

…into develop_st Pull the latest code to resolve the issue that the automatic test case fails

…am next_link and stop short-page override in /files

codecov · 2026-10-15T12:45:31Z

Codecov Report

❌ Patch coverage is 78.75648% with 41 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ckend/ext_components/aidp/services/aidp_service.py	58.00%	21 Missing ⚠️
backend/ext_components/aidp/apps/aidp_mgmt_app.py	86.01%	14 Missing and 6 partials ⚠️

📢 Thoughts on this report? Let us know!

feat(aidp): Knowledge base details page

449655a

mew-boo requested review from Dallas98, WMC001, YehongPan, hhhhsc701 and jeffwu-1999 as code owners October 14, 2026 09:03

cj2026-bit reviewed Oct 14, 2026

View reviewed changes

cj2026-bit self-assigned this Oct 15, 2026

mew-boo added 2 commits October 15, 2026 15:31

feat(aidp): Knowledge base details: Added knowledge graph editing fun…

1002704

…ctionality in detailed information section

Feat: Knowledge base details: 1. Added the copyright header. 2. Fix t…

c5700e3

…he issues raised in the review

cj2026-bit reviewed Oct 15, 2026

View reviewed changes

mew-boo added 3 commits October 15, 2026 19:12

Merge branch 'develop' of https://github.com/ModelEngine-Group/nexent …

61e431a

…into develop_st

Merge branch 'develop' of https://github.com/ModelEngine-Group/nexent …

66c495c

…into develop_st Pull the latest code to resolve the issue that the automatic test case fails

Fix(aidp): Knowledge base details: keep filtered pagination on upstre…

622555f

…am next_link and stop short-page override in /files

Conversation

mew-boo commented Oct 14, 2026

Uh oh!

cj2026-bit left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cj2026-bit left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cj2026-bit left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov Bot commented Oct 15, 2026

Codecov Report

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants