[connectivity-ai]

P023: fix guidance corrected in the rule

Updated P023's full description and prescription to keep whole-block offloading for wholly synchronous work while explicitly allowing streaming async producers with sequential per-chunk offloading of synchronous sink calls (await run_in_thread(handle.write, chunk)). The clarification preserves streaming and avoids moving the async producer into sync code or buffering the whole response. Followed the Metabase callable-offload pattern, with the Tableau streaming case as the documented exception. The detector was unchanged and still reports 86 Tableau findings, 8 Metabase findings, and 0 for MySQL/OpenAPI. Ruff passed; the focused documentation/prescription tests passed (568 passed, 1 xfailed). The import check is unavailable in this rule-fix task.

• Evidence: the remediation lane's reviewer judged the rule's guidance wrong in 1 repo(s); the rule's detection was right
• Detection unchanged: atlanhq/atlan-tableau-app → 86, atlanhq/atlan-mysql-app → 0, atlanhq/atlan-metabase-app → 8, atlanhq/atlan-openapi-app → 0
• Not verified: the full conformance suite and the other 190-odd rules; CI is the gate for those.

Reviewer evidence

• atlanhq/atlan-tableau-app: At app/extracts/api/workbook_content.py:199, 209-211, and 225, the code incrementally consumes async response.aiter_bytes() and preserves a size cap plus rollback. P023's prescription says to offload the whole block owning a file handle; here that would require moving the async producer into synchronous code or materializing the response and changing the documented streaming behavior. Sequential await run_in_thread(f.write, chunk) calls remove the blocking writes while preserving that behavior. The prescription should explicitly allow per-chunk offloading at async boundaries when a synchronous sink consumes an async producer.

Opened by the conformance remediation lane (agent-sdk). Never merged automatically.

Tracking: FND-4530

[linear]

FND-4530

[cmgrote]

/lens

[atlan-app-fleet]

lens summary · updated after round 2 (re-review · only commits since e65c06f)

✅ Ready to merge — every finding is resolved

Open findings: 🔴 0 critical · 🟠 0 high · 🟡 0 medium · ⚪ 0 low (nit)

lens reads this PR as — P023 correctly detects blocking synchronous sink calls in async code, but its whole-block offloading guidance did not fit an async producer streaming into a synchronous sink: following it could force moving the producer into sync code or buffering the response, changing streaming behavior. How: The PR preserves whole-block offloading for wholly synchronous work while documenting a streaming exception: keep the async producer awaited and sequentially offload each sink operation, including open and close. It updates the prescription sources and rule description, with focused tests that check the on-loop write is reported and the offloaded streaming form is not; detection itself is unchanged.

Approach check — sound.

✔️ Resolved (1)

id	severity	where	finding	found	resolved	how
F-94d09d	🟡 medium	packages/conformance/conformance/programs/areas/prescriptions.prose.md:972	The offload example cannot bound a stuck file operation	round 1	round 2	verified fixed

🕘 Round history (2)

round	run	commits	new	resolved	blocking open	cost	log
1	first review	3f373cd..e65c06f	1	0	0	$0.022	run
2	re-review · only commits since e65c06f	e65c06f..4e25c21	0	1	0	$0.015	run

_{round 2 · $0.037 of $1.00 · 35 model calls · 0 failed requests · 85% prompt cache hits · gpt-6-luna}

[cmgrote]

/lens

[atlan-app-fleet]

✅ lens · round 2 · re-review · only commits since e65c06f — ready to merge — every finding resolved · $0.04

Open findings: 🔴 0 critical · 🟠 0 high · 🟡 0 medium · ⚪ 0 low (nit)
Resolved this round: 1

Approach check — ✅ sound

• Problem: P023 correctly detects blocking synchronous sink calls in async code, but its whole-block offloading guidance did not fit an async producer streaming into a synchronous sink: following it could force moving the producer into sync code or buffering the response, changing streaming behavior.
• How the PR solves it: The PR preserves whole-block offloading for wholly synchronous work while documenting a streaming exception: keep the async producer awaited and sequentially offload each sink operation, including open and close. It updates the prescription sources and rule description, with focused tests that check the on-loop write is reported and the offloaded streaming form is not; detection itself is unchanged.

_{$0.037 of $1.00 · 35 model calls · 0 failed requests}

Full summary · Run log

[atlan-ci]

lens: ready to merge — every finding at every level is resolved (round 2).