Skip to content

fix(blocks): read the mapper operation from the tool-call context in j–z blocks - #10078

Merged
waleedlatif1 merged 8 commits into
stagingfrom
sweep2/m4
Nov 16, 2026
Merged

waleedlatif1 merged 8 commits into
stagingfrom
sweep2/m4

Conversation

@waleedlatif1

@waleedlatif1 waleedlatif1 commented Nov 16, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

Moves every block mapper for the integrations J–Z onto mapperOperation (added in #10070), so the operation is read from the tool-call context. These integrations were validated in earlier sweeps, so this PR covers the mapper fix plus any defects found while tracing it.

  • params.operation is undefined when a block runs as an Agent tool, in Pi or in Human-in-the-loop, so until now these mappers skipped their per-operation renames and coercions on those paths. Each newly live branch keeps the model's own args (via toolCallArg), accepts already-typed values, and never overwrites an arg with undefined.
  • Canvas requests for saved workflows are unchanged, except for the fixes listed below. No ids were renamed or removed.
  • The check:mapper-operation baseline shrinks for every block fixed here, and generated tool metadata and docs are regenerated.

Release notes

Jotform

  • As an Agent tool (or in Pi / Human-in-the-loop), Jotform now applies fields filled on the tool row under their block names (report title/type/fields, question text/order/name, new-form questions/properties/emails, bulk submissions/questions, history filters) to the matching tool params.

Kalshi

  • As an Agent tool (or in Pi / Human-in-the-loop), Kalshi now forwards order IDs, order tickers, order type, amend count/client order ID, status/ticker filters, and series tickers filled on the tool row.

Ketch

  • As an Agent tool (or in Pi / Human-in-the-loop), Ketch now parses purposes, topics, controls, and user data that arrive as JSON text into objects, and coerces collectedAt to a number, matching what the canvas sends.

LangSmith

  • As an Agent tool (or in Pi / Human-in-the-loop), Create Run no longer drops the model's run outputs (they were previously overwritten with an empty value).
  • Update Run, Create Feedback, and Create Runs Batch now parse JSON-text fields and coerce the feedback score the same way the canvas does.

LaunchDarkly

  • As an Agent tool (or in Pi / Human-in-the-loop), create_flag (LaunchDarkly and v2) now sends temporary: true when the model asks for a temporary flag. Before, a boolean true was turned into false.
  • update_flag now archives or restores when the model passes archive: true/false. Before, a boolean value was dropped and the update could fail with "At least one update field must be provided".
  • toggle_flag now also accepts a boolean enabled after the mapper runs. This was unchanged in practice, but it would have broken once the operation was resolved.

LinkedIn

  • As an Agent tool (or in Pi / Human-in-the-loop), requests are unchanged. The mapper now resolves the operation from the tool call, and the model's args are kept. Canvas requests for saved workflows are unchanged.

Linq

  • As an Agent tool (or in Pi / Human-in-the-loop), create_attachment now normalizes a picked file to a single file object, so a one-element file array no longer fails schema validation. The react_to_message add/remove choice made by the model is kept.

Loops

  • As an Agent tool (or in Pi / Human-in-the-loop), requests are unchanged. The mapper now resolves the operation from the tool call, and the model's args (for example, create_contact_property name/type) are kept. Canvas requests for saved workflows are unchanged.

ManageEngine ServiceDesk Plus

  • As an Agent tool (or in Pi / Human-in-the-loop), requests are unchanged. The mapper now resolves the operation from the tool call, and the model's args are kept. Canvas requests for saved workflows are unchanged.

Mem0

  • As an Agent tool (or in Pi / Human-in-the-loop), Search Memories and Get Memories are now validated and mapped for the operation that was actually called. Before, they were checked as Add, failed on missing messages, and fell back to unmapped params.

Mintlify

  • As an Agent tool (or in Pi / Human-in-the-loop), Mintlify now turns numeric strings (limit, offset, pageSize, scoreThreshold, retrievalPageSize) into numbers, drops values that are not numbers, and keeps the model's path, source, and status args.

Oracle NetSuite

  • As an Agent tool (or in Pi / Human-in-the-loop), NetSuite now parses JSON-string body and items, coerces and range-checks limit and offset, coerces expandSubResources, and drops roleId/roleExternalId unless the related type is contact. The model's taskId is kept for Get Async Status and Get Async Result.

Otter.ai

  • As an Agent tool (or in Pi / Human-in-the-loop), List Conversations now coerces includeShared to a boolean. The request sent is unchanged.

Parallel AI

  • As an Agent tool (or in Pi / Human-in-the-loop), Search now fills search_queries from the objective when only an objective is given, coerces max_results and max_chars_per_result to numbers, and keeps the model's include_domains and exclude_domains.
  • Extract always sends full_content (false when not set) and maps a stored extract_objective. Deep Research maps stored research_input, processor, and output_format.

People Data Labs

  • As an Agent tool (or in Pi / Human-in-the-loop), People Data Labs now keeps the model's name, website, profile, location, min_likelihood, and titlecase on Company Enrich, Person Enrich, Person Identify, and Company Clean. Before, the mapper took its no-operation path and cleared these fields to undefined. Autocomplete also keeps the model's size.

Perplexity

  • As an Agent tool (or in Pi / Human-in-the-loop), Chat now sends an explicit temperature of 0, which was dropped before. Search keeps a domain filter sent as an array.

Persona

  • As an Agent tool (or in Pi / Human-in-the-loop), Import Accounts now normalizes the stored CSV file the same way the canvas does.

PitchBook

  • As an Agent tool (or in Pi / Human-in-the-loop), profile and lookup operations now forward the model's pbId. Before, it was always cleared, so every ID-based operation failed on tool calls. Search filters on the owning search operation keep the model's values.

Plane

  • As an Agent tool (or in Pi / Human-in-the-loop), Plane now coerces number and boolean string params for the operation that was called (for example, create_cycle sort_order "5" becomes 5). Before, they were always handled with the List Projects field map.

Polymarket

  • As an Agent tool (or in Pi / Human-in-the-loop), Polymarket now reads the operation from the tool call. Requests are unchanged for normal model arguments.

Power BI

  • As an Agent tool (or in Pi / Human-in-the-loop), List Workspaces and Get Refresh History now turn a top/skip value sent as a string into a number. Before, the tool rejected it. When the model omits them, the block defaults apply (top 100, skip 0; refresh history top 60).
  • Execute Query and Refresh Dataset now get the block's defaults for includeNulls (true) and notifyOption (NoNotification).

Profound

  • As an Agent tool (or in Pi / Human-in-the-loop), Profound now reads the operation from the tool call. Requests are unchanged for normal model arguments.

Reddit

  • As an Agent tool (or in Pi / Human-in-the-loop), Reddit now runs the selected operation's parameter mapping instead of always the Get Posts mapping. Get Controversial, Search, and Get User Posts/Comments keep the model's time window instead of dropping it. Submit Post sends the model's boolean nsfw/spoiler/send_replies values and its text or URL. Every other operation keeps the model's own arguments (IDs, text, limits, sort, flags).

RevenueCat

  • As an Agent tool (or in Pi / Human-in-the-loop), RevenueCat now reads the operation from the tool call. Requests are unchanged for normal model arguments.

Rocketlane

  • As an Agent tool (or in Pi / Human-in-the-loop), Rocketlane (legacy and v2) now applies the block's coercions: IDs become numbers, comma lists become arrays, and JSON is parsed. Fields renamed between block and tool (task/time-off type, time-entry source IDs, search_time_entries projectEq/taskEq, list_users and list_invoices status filters, document templateId, placeholder userEmailId) now keep the model's value. Before, create_time_entry could drop the project, phase, or activity the model named.
  • Canvas: when a number (for example a resolved number reference) reaches a string-typed text field or filter (such as External Reference ID Equals, user ID filters, project name, or page token), it is now sent as text. Before, it was silently dropped, so filtered list calls returned unfiltered results. String input is unchanged.

Rootly

  • As an Agent tool (or in Pi / Human-in-the-loop), Rootly (legacy and v2) now maps values saved on the tool row (for example the Incident ID field) to the tool. Model args are still kept wherever the row field is empty. Before, values saved on the row were ignored on tool calls.

Sendblue

  • As an Agent tool (or in Pi / Human-in-the-loop), Sendblue now splits a newline-separated recipient list saved on the tool row into an array, coerces max_duration_ms, and keeps the model's typing state.

Sixtyfour

  • As an Agent tool (or in Pi / Human-in-the-loop), Enrich Lead and Enrich Company now send a struct saved on the tool row, and still keep a struct the model sends.

Smartlead

  • As an Agent tool (or in Pi / Human-in-the-loop), Smartlead now forwards the model's leads (Add Leads to Campaign), sequences (Save Campaign Sequences), and custom fields (Update Lead), plus the model's email (Get Lead by Email / Update Lead) and name (Create Campaign / Upsert Webhook). Before, these were overwritten with empty values.

Stripe

  • As an Agent tool (or in Pi / Human-in-the-loop), Stripe no longer fails in the block mapper when the model passes metadata, address, items, images, or recurring as objects or arrays. Optional Yes/No fields and pagination are now coerced on those calls too.
  • Canvas workflows that reference an object into these fields no longer error with "Invalid JSON input".

TikTok

  • As an Agent tool (or in Pi / Human-in-the-loop), TikTok now runs the selected operation's mapping. Before, every call was treated as Get User, which dropped video IDs, max count, cursor, file, and publish ID. Query Videos now accepts video IDs as an array.

Vanta

  • As an Agent tool (or in Pi / Human-in-the-loop), Vanta now applies operation-specific normalization (page size, file upload, filter dropdowns). Model-supplied fileName, description, name, statusMatchesAny, q, and type are kept. No change for saved canvas workflows.

Wealthbox

  • As an Agent tool (or in Pi / Human-in-the-loop), Read Contact, Read Task, and Write Task without a contact or task ID no longer error in the mapper (the tools list or create in that case), so the OAuth credential mapping is kept. Canvas validation is unchanged.

Zendesk

  • No behavior change. The mapper's operation read moved to the shared helper.

Zep

  • As an Agent tool (or in Pi / Human-in-the-loop), Zep (legacy and v2) now validates and maps params for the operation actually called. Before, every call was treated as "create thread", the mapper threw for calls without both a thread ID and a user ID, and its coercions (for example the numeric limit) were skipped.
  • Add Messages calls now also accept messages as a JSON string.

Review follow-ups

  • Rocketlane Create Time Entry from an Agent tool row with a saved source no longer also sends the model's conflicting source.

Testing

  • Each integration (or batch of blocks) was fixed by one agent and then reviewed by independent adversarial reviewers until it came back clean. Integrations got a vendor-spec/auth lens and a canvas-vs-tool-call/backward-compat lens; batches got the canvas-vs-tool-call lens.
  • Regression tests call each mapper the way the Agent path does, params(modelArgs, { operation }) without an operation in params, and were confirmed to fail on the pre-fix code.
  • Ran the touched test files, biome, check:mapper-operation, check:explicit-any, check:tool-metadata, check:docs, check:docs-manifest, check:block-successors, check:deployment-config, check:canvas-sentences, check:test-patterns, check:utils, check:comment-hygiene, check:source-conventions and check:integration-catalog locally. CI runs type-check and the full suite.

🤖 Generated with Claude Code

@vercel

vercel Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
docs Ready Ready Preview Nov 16, 2026 3:29am UTC

Request Review

@greptile-apps

greptile-apps Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

[High impact] The PR appears safe to merge; the Rocketlane source collision is fixed.

Summary

The PR makes J–Z block mappers read the selected operation from tool-call context and keeps model arguments during field mapping.

  • Jotform through Linq map tool calls using the chosen operation.
  • Loops through Mem0 map tool calls with the chosen operation.
  • Mintlify through Profound map tool-call arguments for the selected action.
  • Reddit and Rocketlane map tool calls to the chosen operation and expected input shapes.
  • Rootly maps each tool call with its selected operation and keeps model arguments.
  • Sendblue through Smartlead keep tool-call values while mapping each action.
  • Stripe through Vanta map tool calls and convert values for the chosen action.
  • Wealthbox through Zep use the operation selected by each tool call.
  • Mapper check baselines drop entries covered by this PR.

Reviews (2) · Last reviewed commit: "fix(rocketlane): send only the saved tim..." · Reviewed by Greptile

Comment thread apps/sim/blocks/blocks/rocketlane.ts Resolved

@cubic-dev-ai cubic-dev-ai Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All reported issues were addressed

Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic

Comment thread apps/sim/blocks/blocks/rocketlane.ts Resolved
@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@greptile

@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@cubic-dev-ai review this PR

@cubic-dev-ai

cubic-dev-ai Bot commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

@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 cubic-dev-ai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No issues found across 63 files

Confidence score: 5/5

  • Automated review surfaced no issues in the provided summaries.
  • No files require special attention.

Turn on auto-fix | Re-trigger cubic

@waleedlatif1
waleedlatif1 merged commit 5729416 into staging Nov 16, 2026
31 of 32 checks passed
@waleedlatif1
waleedlatif1 deleted the sweep2/m4 branch November 16, 2026 04:00

This branch was successfully deployed

1 active deployment
Preview — 3e19ee53 Deployed Nov 16, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant