Repository navigation
fix(integrations): close follow-ups from the integration validation audits - #10183
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
[Critical impact] The PR appears safe to merge; no new blocking issues were found. SummaryThis PR updates integration requests, tightens input checks, and keeps retired operations out of new selections while preserving saved workflows.
As documented by waleedlatif1, saved Greptile Query calls still fail because the endpoint was retired; a replacement remains a product decision. No new actionable issues were found.
|
286d1f3 to
0e24630
Compare
|
@cubic-dev-ai review this PR |
There was a problem hiding this comment.
Review completed against the latest diff
Reply with feedback, questions, or to request a fix.
Turn on auto-fix | Re-trigger cubic
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
All reported issues were addressed across 94 files
Reply with feedback, questions, or to request a fix.
Turn on auto-fix | Re-trigger cubic
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
All reported issues were addressed across 93 files
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Turn on auto-fix | Re-trigger cubic
…i, address review
…-first order writes
ec87f50 to
29a2ac6
Compare
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
No issues found across 93 files
Confidence score: 5/5
- Automated review surfaced no issues in the provided summaries.
- No files require special attention.
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Turn on auto-fix | Re-trigger cubic
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
No issues found across 93 files
Confidence score: 5/5
- Automated review surfaced no issues in the provided summaries.
- No files require special attention.
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Re-trigger cubic
Summary
Closes follow-ups that independent audits found in the already-merged integration validation PRs (#10164-#10181), plus two items the audits flagged as broken in production: Kalshi Create/Cancel/Amend Order and Greptile Query/Search. Each fix is small and stays backward compatible. No block type, subBlock, operation, tool, param or output id was renamed or removed, and no OAuth scope was added. Saved workflows send the same request as before, except where the old request was a bug (listed below). Every behavior fix has a test, and each test was checked to fail on the pre-fix code.
namespaceis required (the block already required it, and the URL cannot be built without it), the param type is no longer widened,rerank.rank_fieldsis required again as in the 2025-01 spec, and one shared rerank-query helpervec-1) is fetched instead of failing "ids must be valid JSON". JSON arrays still worknodeVersionoutside Vercel's documented enum is rejected before the project is created, so no stray project is left behindtext,highlights,subpagesorexcludeSourceDomainis no longer sent asnull/0. An explicitfalse/0is still senttools.config.tool(serialization) intotools.config.params, scoped to Get Activities, and keeps the agent'sleadIdarg/containschunk gets the same array check as later chunks, through one shared helper/portfolio/orderswrite endpoints (410deprecated_v1_order_endpoint)hiddenoperation, as it already skipped permission-denied ones) (kept as a hidden option so saved workflows and v2 state writes stay valid), and the Query/Search tools are marked retired<logs-{now/d}>work again (encoded as Elasticsearch documents), and dot segments are still rejected. This was a regression against mainsafeUrlPathSegmentTSDoc@returnsupdatedblockFieldOrToolArghelper replaces the copies in SQS and SSM (codepipeline keeps its own, whose null/'' semantics differ). The stale-input test now asserts the executor mergefind_rowstransformResponsewith a bare-array bodydeleteFathomWebhookhelper for both delete paths; typed test context instead ofas never; Ashby test comment trimmedtoRecord; Ketchvendorsdescription; 1PasswordUNSUPPORTEDvault type typed from the SDK enum; Infisical approval fields marked nullableparseOptionalJsonInput(keeping its static error message, which never quotes the input); Gamma test renamedcursorinput, raw activity start/end, placeholders; SFTP connector reuses the sharedjoinRemotePath(same output for its normalized paths)query/authorName; Parallel logs "completed" only after the failed/cancelled check; Linkup cost lookup table (same prices); Quiver retired-modelSetLeft for a product decision:
/v2/queryand/v2/search. Index Repository and Check Status are unchanged.Release notes
Pinecone
vec-1) as well as a JSON array.Vercel
22instead of22.x) before creating the project. Before, the project was created and then the version update failed, leaving an extra project behind. A blank version still means no version.Exa
0, and untouched switches sentnull.Tavily
Lemlist
Polymarket
0means the full history and a value that is not a timestamp gets a clear error.Spotify
Kalshi
Create Order, Cancel Order and Amend Order work again. Kalshi removed the old
/portfolio/orderswrite endpoints, so they now use/portfolio/events/orders. Saved blocks keep their fields; this is how they translate, per Kalshi's OpenAPI spec:0.5600), and contracts as a fixed-point count. Setting both a yes and a no price that disagree is an error.good_till_canceled. Without an expiration that rests until canceled, which is what the old API did. Expiration is sent asexpiration_time.taker_at_cross, the value in Kalshi's own request example.orderobject (as null when Kalshi no longer reports it). Kalshi now returns only the order ID, client order ID and fill counts from a create, cancel or amend. The order output reports those, plus the direction, price and settings that were sent; fields Kalshi no longer reports are null. A count-less or price-less amend also reports the order as it was before the amend inold_order. The V2 amend order output gainsyes_price_dollars,no_price_dollars,fill_count_fpandremaining_count_fp.market_ticker. Without it the cancel goes to shard 0, and an order on another shard (Combos, Crypto and Commodities, Sports) answers 404 and stays open. Cancel Order therefore has a new optional Market Ticker field, sent asmarket_ticker. When it is empty, the order is read first to find its ticker. Set Market Ticker when cancelling an order you just created; without it, Sim reads the order first, and Kalshi's order reads can briefly lag after a write. The canceled count is unchanged, andordercarriesorder_id,client_order_idandticker.Greptile
Agent tool picker
Elasticsearch
<logs-{now/d}>work again.PostHog
groupsarray. Before, the rollout was silently dropped.Wealthbox
Ketch, 1Password, Infisical, arXiv
vendorsis the deprecated list of opted-out vendor IDs; 1Password vaulttypecan beUNSUPPORTED; Infisical approval IDs and timestamps can be null; arXiv lists thequeryandauthorNameoutputs.