chore(api): cache the setup status to cut down DB access

[cqjjjzr]

Important

1. Make sure you have read our contribution guidelines
2. Ensure there is an associated issue and you have been assigned to it
3. Use the correct syntax to link this PR: Fixes #<issue number>.

Summary

Fix #36965 . Not using @cache due to reasons mentioned in the comment.s

Checklist

• This change requires a documentation update, included: Dify Document
• I understand that this PR may be closed in case there was no previous discussion or issues. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly.
• I ran make lint && make type-check (backend) and cd web && pnpm exec vp staged (frontend) to appease the lint gods

[github-actions]

Pyrefly Diff

base → PR

--- /tmp/pyrefly_base.txt	2026-06-05 00:42:26.802000994 +0000
+++ /tmp/pyrefly_pr.txt	2026-06-05 00:42:15.570946821 +0000
@@ -2081,7 +2081,7 @@
 ERROR `SimpleNamespace` is not assignable to attribute `_current_tenant` with type `Tenant | None` [bad-assignment]
   --> tests/unit_tests/controllers/console/test_workspace_members.py:63:43
 ERROR `in` is not supported between `Literal['count']` and `None` [not-iterable]
-   --> tests/unit_tests/controllers/console/test_wraps.py:213:16
+   --> tests/unit_tests/controllers/console/test_wraps.py:220:16
 ERROR `SimpleNamespace` is not assignable to attribute `db` with type `SQLAlchemy` [bad-assignment]
   --> tests/unit_tests/controllers/files/test_image_preview.py:23:17
 ERROR `SimpleNamespace` is not assignable to attribute `request` with type `Request` [bad-assignment]

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	46.88%	46.89%	+0.00%
Strict coverage	46.38%	46.38%	+0.00%
Typed symbols	25,884	25,892	+8
Untyped symbols	29,644	29,650	+6
Modules	2819	2819	0

[asukaminato0721]

try use closure.

or class.

import functools

def once_true(func):
    has_success = False 
    
    @functools.wraps(func)
    def wrapper(*args, **kwargs):
        nonlocal has_success
        
        if has_success:
        
            return True
            
        result = func(*args, **kwargs)
        
        if result is True:
            has_success = True
            
        return result
        
    return wrapper

[asukaminato0721]

https://doc.rust-lang.org/std/cell/struct.OnceCell.html ref

[asukaminato0721]

@dosu how is this change

[dosubot]

This PR looks like a solid, well-thought-out optimization. Here's a quick review:

What it does well:

1. Smart caching strategy — The once_true decorator only memoizes True results, which is exactly right for a one-way state transition (setup goes from incomplete → complete, never back). This avoids the footgun of @cache permanently memoizing an initial False before setup has run .

2. Immediate cache warming — Calling mark_setup_completed() right after creating the DifySetup record avoids an unnecessary extra DB round-trip .

3. Good test coverage — Tests verify both that True is cached (only 1 DB call for multiple checks) and that False is not cached .

4. Real impact — Since setup_required decorates every protected console endpoint, this eliminates a SELECT on every single API request once the instance is set up.

Minor observations / suggestions:

• The implementation already uses the closure-based approach that @asukaminato0721 suggested, with the addition of mark_success() / reset_success() helper methods attached to the wrapper — nice for testability and the explicit mark_setup_completed() path.
• The cache is process-scoped (closure variable), so it naturally resets on server restart — no stale state concerns across deploys.
• One edge case worth noting: if someone resets their Dify instance (purges DifySetup from the DB) without restarting the server, the cached True would persist. This is acknowledged in issue [Refactor/Chore] Cut down a DB access by caching setup status for setup_required #36965 and is an acceptable trade-off since that scenario is extremely rare and a restart would fix it.

Overall, this is a clean, minimal change that removes unnecessary DB load in the hot path. Looks good to merge. 👍

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.