Errors and how to fix them

The platform shows these errors while you build, deploy, or run a pipeline. Each appears here exactly as shown, with its cause and fix. Search this page for the exact text on your screen, and note that <name>, <n>, and similar placeholders stand for the parts of a message that change.

In the Issues window

You meet these when you deploy. The Issues window collects them: errors block the deploy, and warnings can be overridden. See Self-check for how the gate sorts them. Search the families below for your exact message. If it is not here, it restates a setting’s documented bounds. The Other pre-deploy messages section points to the page that owns it.

Graph wiring

Message	Cause	How to fix	Where the rule lives
Missing connection for output handle "<handleId>" in node "<name>"	A node output, or a Match branch, connects to nothing.	Connect every output on the path from Start to End, or remove the node.	This table (graph wiring has no owning page yet).
Infinite cycle: <node> → <node> → …	The graph loops, so an item could never leave it.	Remove or re-route one edge in the listed loop.	This table (graph wiring has no owning page yet).
Handle "<handleId>" of node "<name>" references unknown field "<fieldProxyId>"	The node still points at a field that was deleted.	Re-bind the node’s fields, or ask the agent to fix the reference.	Fields and data
Required field "<name>" is not available at input handle "<handleId>" of node "<name>"	No upstream node provides this required field on that path.	Produce the field upstream, mark the input optional, or re-route the edge.	Fields and data
Required field "<name>" is only optionally provided at input handle "<handleId>" of node "<name>"	The field is required here, but the upstream node may leave it empty.	Make the upstream output required, or mark this input optional.	Fields and data
Optional field "<name>" is not available at input handle "<handleId>" of node "<name>"	An optional input names a field no upstream node provides.	Remove the optional input, or produce the field upstream.	Fields and data

Task interface sample data

Message	Cause	How to fix	Where the rule lives
Code is required	The labeling interface has not been built yet.	Ask the agent to build it.	Generation node
Data is not valid JSON	The interface’s sample data does not parse.	Fix the sample-data JSON.	Fields and data
Data must be a JSON object	The sample data parses but is not an object.	Make the sample data an object keyed by input name.	Fields and data
Data is missing required input "<name>"	The sample data has no value for a required input.	Add a sample value for the input.	Fields and data
Data for input "<name>" does not match its schema	The sample value has the wrong type or shape.	Fix the sample value to match the field’s type.	Fields and data
Data has unexpected key "<key>" that does not match any input	The sample data names a field the node does not read.	Remove the stray key.	Fields and data
<component> has minDurationSec (<n>) greater than maxDurationSec (<m>). minDurationSec must be less than or equal to maxDurationSec	A video control in the interface sets a minimum duration longer than its maximum, on a data-collection node.	Lower the minimum or raise the maximum so the minimum is not longer, or ask the agent to fix the control.	Generation node

Field type schema

Message	Cause	How to fix	Where the rule lives
Invalid JSON schema: not valid JSON	The field’s type schema does not parse as JSON.	Fix the schema JSON, or ask the agent to regenerate the field type.	Fields and data
Invalid JSON schema: must be a JSON object	The type schema is JSON but not an object.	Replace it with a JSON-object schema.	Fields and data
Invalid JSON schema: <details>	The type schema does not compile. <details> names the problem.	Fix the reported schema problem.	Fields and data

Node names and settings

Message	Cause	How to fix	Where the rule lives
The name must be between 4 and 32 characters long	The node’s name is shorter than 4 or longer than 32 characters.	Rename the node to between 4 and 32 characters.	This table (no owning page yet).
The name must start with a letter, can contain only lowercase letters, numbers and underscores, and must end with a letter or number	The node’s name uses characters the format forbids, or does not start and end as required.	Rename it to lowercase_with_underscores form: start with a letter, use only lowercase letters, numbers, and underscores, and end with a letter or number.	This table (no owning page yet).
The name must be unique across all nodes	Another node already uses this name.	Rename one of the two nodes so every node name is different.	This table (no owning page yet).
The description exceeds the maximum of 150000 characters	The node’s description is longer than 150000 characters.	Shorten the description to 150000 characters or fewer.	This table (no owning page yet).
This setting exceeds the maximum of 150000 characters	The total text in one of the node’s settings exceeds 150000 characters.	Shorten that setting’s content to 150000 characters or fewer.	This table (no owning page yet).

Field structure

Message	Cause	How to fix	Where the rule lives
Entity type must be one of the supported core entity types: <list>	A field’s type is missing or is not one the platform supports. <list> names the supported types.	Give the field one of the listed types, or ask the agent to fix it.	This table (no owning page yet).
Relations must reference existing entities	A field relation points at a field that no longer exists.	Re-link the relation to an existing field, or remove it.	This table (no owning page yet).
Output entities must have at least one relation to another entity	On a node with more than one field, an output field has no relation to another field.	Relate the output to the field it builds on.	This table (no owning page yet).

Onboarding examples

These are warnings, so you can override them and deploy anyway.

Message	Cause	How to fix	Where the rule lives
Example check paths do not match any output entity jsonPath of the node: <paths>	An onboarding worked-example’s checks target locations that none of the node’s output fields use. <paths> lists them.	Point the checks at the node’s output locations, or override the warning.	This table (no owning page yet).
Examples gated only by llm checks, with no golden-backed deterministic check: <ids>	An onboarding worked-example is scored only by an LLM check, with no exact or code check to grade it deterministically. <ids> lists the examples.	Add a golden-backed check (an exact or code check) to the example, or override the warning.	This table (no owning page yet).

Other pre-deploy messages

Most other messages in the Issues window restate a setting’s documented bounds, such as a name length, a step count, or a price floor. The message names the setting. For node settings, see Code node, Generation node, Quorum node, Match node, and Start and End nodes. For quality and offboarding settings, see Quality control. See Self-check for check cases and their run states, Fields and data for field descriptions, and Label with an LLM instead of experts for synthetic-only settings.

On a check card

A check card shows these in its Error section after a run fails. See Self-check for what each check kind does and when to run it.

Message	Cause	How to fix	Where the rule lives
The check expired before it was submitted	The opened labeling screen was never submitted within about an hour.	Rerun the User Interface check and complete the screen.	Self-check
No quality criteria configured on this node	An LLM Quality Assurance check ran with no criteria on the node.	Add quality criteria, then rerun the check.	Self-check
LLMQA check requires LLM QA to be enabled on this node (evaluation_result proxies are missing)	LLM review is turned off on the node.	Turn on LLM review, then rerun the check.	Self-check
Invalid check input: <parse error>	A Code Test case’s input JSON stopped parsing.	Fix the case’s JSON input.	Self-check
Invalid check input JSON	An LLM Quality Assurance case’s input JSON stopped parsing.	Fix the case’s JSON input.	Self-check
Execution output must be an object	A Code Test run returned something other than a dictionary.	Make the code return a dictionary.	Code node
Missing required field: "<name>" / Unknown field: "<name>" / Field "<name>" exceeds the maximum size of 32KB (actual: <n>KB) / Field "<name>": <schema detail>	The check’s input or returned output breaks the node’s field contract or the 32 KB per-value cap.	Fix the case input, or fix the node’s code and fields so the output matches its declared contract.	Code node

On the Items tab

The run’s Items tab records these as an item’s failure reason, on the item and in its trace.

Message	Cause	How to fix	Where the rule lives
This item is missing a value for the required input "<name>".	The item entered the run with no value for a required Start input.	Fill the value in the dataset, or mark the input optional. A later run labels the item only if its own dataset, filter, and limit still select it — a failed status is not a retry trigger.	Fields and data
This item is missing values for required inputs: "<name>", "<name>".	Several required Start inputs have no value for the item.	Fill the missing values in the dataset, or mark those inputs optional. A later run labels the item only if its own dataset, filter, and limit still select it — a failed status is not a retry trigger.	Fields and data
Insufficient funds to reserve the node cost	The team balance ran out mid-run, and the run paused.	Top up the balance, then press Resume.	Pricing and billing
Missing required field "<name>" (<fieldProxyId>) / Unknown field proxy (<fieldProxyId>) / Field "<name>" (<fieldProxyId>) exceeds the maximum size of 1024KB (actual: <n>KB) / Field "<name>" (<fieldProxyId>): <schema detail>	The Code node’s returned output breaks its declared field contract or the 1 MB per-value cap.	Fix the code to return every required declared field, only declared fields, matching types, under 1 MB each.	Code node
Dynamic entity "<name>" (<fieldProxyId>) has <n> instances, exceeding the max of <max>	A repeating input arrived with more values than the field’s maximum.	Raise the field’s maximum (up to 50), or reduce the upstream values.	Generation node
Required field proxy '<fieldProxyId>' is missing	A Match test read a required field with no recorded value.	Mark the tested input optional, or guarantee the value upstream.	Match node
Invalid JSONPath '<path>': <detail>	A Match test’s location does not evaluate.	Fix the test’s location expression.	Match node
JSONPath '<path>' resolved to <n> values, expected a single value	A Match test’s location matches more than one value.	Narrow the location so it resolves to at most one value.	Match node

When you deploy or launch

You meet these when you confirm a deploy, or when you set what a run will launch on the Start node.

Message	Cause	How to fix	Where the rule lives
Insufficient balance to deploy this pipeline. Add funds to your team and restart the deploy.	The team balance is below the deploy minimum at deploy time.	Add funds, then deploy again. A failed deploy frees the pipeline.	Pricing and billing
Your team’s balance (<amount>) is too low to deploy this pipeline. Add funds to your team to continue.	The balance is below the deploy minimum when you confirm.	Add funds, then reopen the deploy.	Pricing and billing
Cannot create run for pipeline <pipelineId> because it already has an active run.	The pipeline already has a run whose deploy did not fail (a run that failed at run time still counts).	Clone the pipeline and deploy the clone.	How the platform works
Cannot deploy pipeline (<n> validation error<s>): <messages>	The server revalidated at deploy and found blocking errors.	Fix the listed issues. Each also appears in the Issues window.	Self-check
In "<node>" you tried to reuse audience from a pipeline that wasn't deployed. Please select a different node from this project or choose "Do not reuse audience".	Audience reuse points at a node with no recorded deployment.	Point reuse at a deployed node, or turn reuse off.	Generation node
In "<node>" you tried to reuse audience from a pipeline you don't have access to. Please select a node from this project or choose "Do not reuse audience".	Audience reuse points at another team’s pipeline.	Point reuse at one of your team’s nodes, or turn reuse off.	Generation node
Cannot disable onboarding when reusing WFR from a node that was deployed with onboarding (<pipelineId>/<nodeId>)	A reusing node turned onboarding off while its source deployment had it on.	Turn onboarding back on, or stop reusing that audience.	Generation node
Only <n> item<s> match the current filter. Lower the limit or widen the filter.	The item limit is higher than the number of matching items.	Lower the limit, or widen the filter.	Label a large dataset in batches
This run would launch <n> items, more than the maximum of 10,000 per run. Set “Max items to launch” to 10,000 or fewer, or narrow the filter.	The launch count is over the per-run limit of 10,000 items.	Set Max items to launch to 10,000 or fewer, or narrow the filter.	Label a large dataset in batches
Pipeline input field<s> not present in this dataset: <names>	Bound inputs name fields the current dataset lacks.	Refresh fields and re-bind, or switch to a dataset that has them.	Start and End nodes
The item filter references dataset fields that no longer exist. Clear the filter to recompute item counts correctly.	The filter references removed dataset fields.	Clear the filter, then rebuild it.	Start and End nodes
Dataset already has field(s) "<name>", … with an incompatible JSON schema. Rename the field proxies, edit the dataset fields' schemas, or remove the conflicting dataset fields, then retry.	A pipeline field (usually an output) has the same name as an existing dataset field but an incompatible type, so the deploy fails instead of overwriting it.	Rename the pipeline field, or change the dataset field’s type to match, then deploy again.	Label a large dataset in batches
The provided `itemFilter` references field IDs that do not exist on the selected dataset.	The agent relays this when a requested filter names fields the new dataset lacks after a dataset switch.	Ask the agent to rebuild the filter against the new dataset’s fields.	Start and End nodes

On the dataset page

You meet these on a dataset’s own page, when you upload a file, derive fields, append items, or filter the table. See Create and manage datasets for the tasks behind them.

Message	Cause	How to fix	Where the rule lives
Please upload a JSON, JSONL or CSV file.	The dropped or chosen file is none of JSON, JSONL or CSV.	Upload a .json, .jsonl or .csv file.	Create and manage datasets
File is empty	The file is 0 bytes.	Upload a file that has content.	Create and manage datasets
File is too large (max 1024 MB).	The file is over the 1024 MB per-file limit.	Split the data into files of 1024 MB or under.	Create and manage datasets
Please select only 1 file.	More than one file was dropped at once.	Drop one file at a time.	Create and manage datasets
File size exceeds maximum allowed size of 1024MB	The uploaded file is over the 1024 MB limit. The server rejected it.	Upload a file of 1024 MB or under.	Create and manage datasets
Another upload is already in progress. Complete or cancel it before starting a new one.	The dataset already has an upload running. It takes one at a time.	Wait for it to finish, or cancel it, then upload again.	Create and manage datasets
Cannot ingest into existing field "uploaded": its schema differs from the upload field schema	The dataset already has a field named uploaded with a custom type, so the upload cannot store into it.	Uploads cannot use that field. Create a new dataset and upload there.	Create and manage datasets
File is malformed (could not parse <JSON|JSONL|CSV>).	The file does not parse as the format it was read as — JSON, JSONL or CSV, or a JSON root that is not an array. For JSONL the message names the offending line.	Repair the file in the format it was read as: a JSON array of objects, one JSON object per line, or a CSV with a header row and one row per item.	Create and manage datasets
File must contain at least one item.	The file’s array is empty.	Add at least one item to the file.	Create and manage datasets
Validation failed with <n> error(s):, followed by one line per bad item: Item at index <n> exceeds the maximum allowed size of 1.00 MiB (actual: <size>) / Item at index <n> does not match field "<name>" JSON Schema	One or more items failed validation, so nothing was written. The list stops at 50.	Fix the listed items (shrink any over 1 MB), then upload again.	Create and manage datasets
Dataset item cap reached; remaining rows were not imported.	The upload hit the 150,000-item cap. The items up to the cap were kept.	Upload fewer items, or split the data across datasets.	Create and manage datasets
Schema incompatible with existing upload (<n> issue(s)):, followed by one line per issue: Field "<name>" (<jsonPath>) is present in existing upload but missing in new upload / <path>: type "<type>" is not present in existing schema / <path>: existing schema has enum constraint but new schema does not / <path>: enum value "<v>" is not present in existing schema / <path>: existing schema has array items constraint but new schema does not / <path>.<key>: property exists in existing schema but is missing in new schema	An appended file’s shape does not match the first upload’s fields.	Make the appended file’s fields match the first upload’s names and types, then append again.	Create and manage datasets
Dataset has reached the maximum of 500 fields	The dataset already has 500 fields.	Use a new dataset, or derive fewer fields.	Create and manage datasets
Adding <n> field(s) would exceed the maximum of 500 fields (currently <m>)	The batch of new fields would push the dataset over 500.	Create fewer fields, or use a new dataset.	Create and manage datasets
Field "<name>" has reached the maximum of 500 revisions per item	A field on one item already holds 500 recorded values.	Point new derivations at a different target field, or use a new dataset.	Create and manage datasets
Value-based filter is not supported on field(s) <names>: revisions exceed the 32KB preview limit. Use the "no revision" filter or filter by source instead.	The field’s largest value is over 32 KB, so it cannot be filtered by value.	Filter that field by source, or by missing value, instead.	Create and manage datasets
Invalid JSONPath: <detail>	The filter’s location expression does not parse.	Fix the location. Use $ for the whole value.	Create and manage datasets
Choose a value filter, source filter, or “No revision for this field”.	Apply was clicked with no filter set.	Set a value, a source, or the no-value option, then apply.	Create and manage datasets
Derived field validation failed:, followed by one line per bad entry: - "<name>": name already exists in dataset / - "<name>": duplicate name in batch / - "<label>": source field <id> not found / - target field <id> not found in dataset	A request to derive fields has invalid entries, so nothing was created.	Fix the names and sources in the request, then retry, or ask the agent to.	Create and manage datasets

Related

• Self-check — the Issues gate that shows most of these errors, and how it sorts errors from warnings.
• Code node — the run-time limits and field contract behind the Code Test and Code-node item failures.
• Start and End nodes — the dataset, item filter, and launch-limit rules behind the launch-scope errors.
• Label a large dataset in batches — the 10,000-item per-run cap and the batch cycle behind the launch-count errors.
• Create and manage datasets — the upload contract, field derivation, and filters behind the dataset-page errors.