Because our source of truth for the resulting model is a STEP file, you will always have STEP file contents when you list your generated models. Any other formats you request here will also be returned when you list your generated models.

This operation is performed asynchronously, the `id` of the operation will be returned. You can use the `id` returned from the request to get status information about the async operation from the `/async/operations/{id}` endpoint.

One thing to note, if you hit the cache, this endpoint will return right away. So you only have to wait if the status is not `Completed` or `Failed`.

The format the output file should be converted to.

output_format

Autodesk Filmbox (FBX) format. <https://en.wikipedia.org/wiki/FBX>

Binary glTF 2.0.

This is a single binary with .glb extension.

This is better if you want a compressed format as opposed to the human readable glTF that lacks compression.

glTF 2.0. Embedded glTF 2.0 (pretty printed).

Single JSON file with .gltf extension binary data encoded as base64 data URIs.

The JSON contents are pretty printed.

It is human readable, single file, and you can view the diff easily in a git commit.

The OBJ file format. <https://en.wikipedia.org/wiki/Wavefront_.obj_file> It may or may not have an an attached material (mtl // mtllib) within the file, but we interact with it as if it does not.

The PLY file format. <https://en.wikipedia.org/wiki/PLY_(file_format)>

The STEP file format. <https://en.wikipedia.org/wiki/ISO_10303-21>

The STL file format. <https://en.wikipedia.org/wiki/STL_(file_format)>

If we should output the kcl for the model.

Access-Control-Allow-Headers header. This is a comma-separated list of headers.

ID for this request. We return it so that users can report this to us and help us debug their problems.

Generate a CAD model from text.

For text-to-cad prompts, this will always return the STEP file contents as well as the format the user originally requested.

This endpoint requires authentication by a Zoo employee.

The ML prompts are returned in order of creation, with the most recently created ML prompts first.

Maximum number of items returned by a single call

limit

Token returned by previous call to retrieve the subsequent page

page_token

sort_by

Supported set of sort modes for scanning by created_at only.

Currently, we only support scanning in ascending order.

Sort in increasing order of "created_at".

Sort in decreasing order of "created_at".

List all ML prompts.

This endpoint requires authentication by a Zoo employee.

Get a ML prompt.

This endpoint is used to convert a proprietary CAD format to KCL. The file passed MUST have feature tree data.

A STEP file does not have feature tree data, so it will not work. A sldprt file does have feature tree data, so it will work.

Input filepaths will be normalized and re-canonicalized to be under the current working directory -- so returned paths may differ from provided paths, and care must be taken when handling user provided paths.

The options to run on the code. By default this is set to `execute`.

code_option

Code option for running and verifying kcl.

<details><summary>JSON schema</summary>

```json { "title": "CodeOption", "description": "Code option for running and verifying kcl.", "type": "string", "enum": [ "parse", "execute", "cleanup", "mock_execute" ] } ``` </details>

Converts a proprietary CAD format to KCL.

A request to generate KCL code completions.

Generate code completions for KCL.

Even if you give specific ranges to edit, the model might change more than just those in order to make the changes you requested without breaking the code.

You always get the whole code back, even if you only changed a small part of it.

This operation is performed asynchronously, the `id` of the operation will be returned. You can use the `id` returned from the request to get status information about the async operation from the `/async/operations/{id}` endpoint.

This endpoint will soon be deprecated in favor of the `/ml/text-to-cad/multi-file/iteration` endpoint. In that the endpoint path will remain but it will have the same behavior as `ml/text-to-cad/multi-file/iteration`.

Iterate on a CAD model with a prompt.

This endpoint can iterate on multi-file models.

Even if you give specific ranges to edit, the model might change more than just those in order to make the changes you requested without breaking the code.

You always get the whole code back, even if you only changed a small part of it. This endpoint will always return all the code back, including files that were not changed. If your original source code imported a stl/gltf/step/etc file, the output will not include that file since the model will never change non-kcl files. The endpoint will only return the kcl files that were changed.

This operation is performed asynchronously, the `id` of the operation will be returned. You can use the `id` returned from the request to get status information about the async operation from the `/async/operations/{id}` endpoint.

Input filepaths will be normalized and re-canonicalized to be under the current working directory -- so returned paths may differ from provided paths, and care must be taken when handling user provided paths.

Body for iterating on models from text prompts.

A response from a text to CAD multi-file iteration.

Iterate on a multi-file CAD model with a prompt.

This will always return the STEP file contents as well as the format the user originally requested.

This endpoint requires authentication by any Zoo user. It returns the text-to-CAD models for the authenticated user.

The text-to-CAD models are returned in order of creation, with the most recently created text-to-CAD models first.

If we should return the model file contents or just the metadata.

no_models

List text-to-CAD models you've generated.

This endpoint requires authentication by any Zoo user. The user must be the owner of the text-to-CAD model.

Get a text-to-CAD response.

This can be a text-to-CAD creation or iteration.

This endpoint requires authentication by any Zoo user. The user must be the owner of the ML response, in order to give feedback.

feedback

Give feedback to a specific ML response.

Machine learning to generate CAD models and other things.

Geometry Engine

Zoo CAD Engine Overview

KCL: A Programming Language for Parametric CAD

File Formats

A Practical Overview of CAD File Formats

Text-to-CAD

Text-to-CAD: Generating Editable, Parametric B-Rep CAD Models from Natural Language

Automation

Automating Your Workflow

The response from the `AddHoleFromOffset` command.

If the offset path splits into multiple paths, this will contain the UUIDs of the new paths. If the offset path remains as a single path, this will be empty, and the resulting ID of the (single) new path will be the ID of the `AddHoleFromOffset` command.

The email address of the user to add to the org.

An ISO-3166 alpha-2 country code. Always uppercase.

The country component. This is a two-letter ISO country code.

Edge info struct (useful for maintaining mappings between edges and faces and adjacent/opposite edges).

The size of the angle, measured in the chosen unit.

How to style the end of the annotation line.

How to style the start of the annotation line.

How to style the start and end of the line

2D annotation type (screen or planar space)

A response for a query on the API call table that is grouped by something.

The email of the user that requested the API call.

The user ID of the user that requested the API call.

The origin of the API call. This is parsed from the `Origin` header.

The IP address of the user making the API call.

The async API call was uploaded to be converted.

An API call with the price.

This is a join of the `ApiCall` and `ApiCallPrice` tables.

The date and time the API call completed billing.

DateTime

The date and time the API call was created.

int64

String

The Request Method (VERB)

This type also contains constants for a number of common HTTP methods such as GET, POST, etc.

Currently includes 8 variants representing the 8 methods defined in [RFC 7230](https://tools.ietf.org/html/rfc7231#section-4.1), plus PATCH, and an Extension variant for all extensions.

The HTTP method requested by the API call.

The number of minutes the API call was billed for.

The organization ID of the API call if it is billable through an organization.

double

The request query params sent by the API call.

The response body returned by the API call. We do not store this information if it is above a certain size.

The date and time the API call started billing.

The status code returned by the API call.

int32

The Stripe invoice item ID of the API call if it is billable.

The date and time the API call was last updated.

The ID of the user that made the API call.

token used to fetch the next page of results (if any)

The type of error sent by the KittyCAD API.

An API token.

These are used to authenticate users with Bearer authentication.

The date and time the API token was created.

If the token is valid. We never delete API tokens, but we can mark them as invalid. We save them for ever to preserve the history of the API token.

An auth token. A uuid with a prefix of api-

The date and time the API token was last updated.

The ID of the user that owns the API token.

Information about a third party app client.

The number of times we've attempted to process this job.

The time and date the async API call was completed.

The time and date the async API call was created.

The unique identifier of the async API call.

This is the same as the API call ID.

The JSON input for the API call. These are determined by the endpoint that is run.

The JSON output for the API call. These are determined by the endpoint that is run.

The time and date the async API call was started.

The time and date the async API call was last updated.

The user ID of the user who created the async API call.

The worker node that is performing or performed the async API call.

The time and date the API call was completed.

The time and date the API call was created.

The unique identifier of the API call.

This is the same as the API call ID.

The output format of the file conversion.

The output format options of the file conversion.

The converted files (if multiple file conversion), if completed, base64 encoded. The key of the map is the path of the output file.

The source format of the file conversion.

The source format options of the file conversion.

The time and date the API call was started.

The time and date the API call was last updated.

The user ID of the user who created the API call.

The material density as denoted by the user.

The material mass as denoted by the user.

The code for the model. This is optional but will be required in the future once we are at v1.

The output of the model in the given file format the user requested, base64 encoded. The key of the map is the path of the output file.

The original source code for the model, previous to the changes.

The prompt for the overall changes. This is optional if you only want changes on specific source ranges.

The source ranges the user suggested to change.

The version of kcl to use. If empty, the latest version will be used.

The output files. Returns a map of the file name to the file contents. The file contents are not encoded since kcl files are not binary.

The project name. This is used to tie the prompt to a project. Which helps us make our models better over time.

The prompt for the overall changes. This is optional if you only want changes on specific source ranges. This will apply to all the files.

The authentication callback from the OAuth 2.0 client. This is typically posted to the redirect URL as query params after authenticating.

For Apple only, a JSON web token containing the user’s identity information.

The state that we had passed in through the user consent URL.

For Apple only, a JSON string containing the data requested in the scope property. The returned data is in the following format: `{ "name": { "firstName": string, "lastName": string }, "email": string }`

Co-ordinate axis specifier.

See [cglearn.eu] for background reading.

[cglearn.eu]: https://cglearn.eu/pub/computer-graphics/introduction-to-geometry#material-coordinate-systems-1

Specifies the sign of a co-ordinate axis.

Specifies which direction the axis is pointing.

Response sent when a request did not succeed.

Errors that occurred during the modeling command.

Websocket responses can either be successful or unsuccessful. Slightly different schemas in either case.

The user is missing a payment method and has exceeded their free API call credits for the month.

The response from the 'BooleanIntersection'.

If the operation produced just one solid, then its ID will be the ID of the modeling command request. But if any extra solids are produced, then their IDs will be included here.

The response from the `CameraDragEnd` command.

Camera rotate (spherical camera revolve/orbit)

Camera rotate (trackball with 3 degrees of freedom)

Camera zoom (increase or decrease distance to reference point center)

The response from the `CameraDragMove` command. Note this is an "unreliable" channel message, so this data may need more data like a "sequence"

The response from the `CameraDragStart` endpoint.

A type of camera movement applied after certain camera operations

Adjusts the camera position during the camera operation

Camera settings including position, center, fov etc

Camera's look-at center (center-pos gives viewing vector)

Camera's field-of-view angle (if ortho is false)

The Camera's orientation (in the form of a quaternion)

Whether or not the camera is in ortho mode

The camera's ortho scale (derived from viewing distance if ortho is true)

Card brand.

Can be `amex`, `diners`, `discover`, `jcb`, `mastercard`, `unionpay`, `visa`, or `unknown`.

Checks on Card address and CVC if provided.

Two-letter ISO code representing the country of the card.

Two-digit number representing the card's expiration month.

Four-digit number representing the card's expiration year.

Uniquely identifies this particular card number.

Card funding type.

Can be `credit`, `debit`, `prepaid`, or `unknown`.

ClientMetrics contains information regarding the state of the peer.

The height of the inbound video stream in pixels.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-frameheight

The width of the inbound video stream in pixels.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-framewidth

Counter of the number of WebRTC frames that the client has decoded from the inbound video stream.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-freezecount

Counter of the number of WebRTC frames the client has dropped from the inbound video stream.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-framesdropped

Current number of frames being rendered in the last second. A good target is 60 frames per second, but it can fluctuate depending on network conditions.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-freezecount

Counter of the number of WebRTC frames that the client has received from the inbound video stream.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-freezecount

Number of times the inbound video playback has frozen. This is usually due to network conditions.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-freezecount

Amount of "jitter" in the inbound video stream. Network latency is the time it takes a packet to traverse the network. The amount that the latency varies is the jitter. Video latency is the time it takes to render a frame sent by the server (including network latency). A low jitter means the video latency can be reduced without impacting smooth playback. High jitter means clients will increase video latency to ensure smooth playback.

https://www.w3.org/TR/webrtc-stats/#dom-rtcreceivedrtpstreamstats-jitter

Number of "key frames" decoded in the inbound h.264 stream. A key frame is an expensive (bandwidth-wise) "full image" of the video frame. Data after the keyframe become -- effectively -- "diff" operations on that key frame. The Engine will only send a keyframe if required, which is an indication that some of the "diffs" have been lost, usually an indication of poor network conditions. We like this metric to understand times when the connection has had to recover.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-keyframesdecoded

Amount of packets lost in the inbound video stream.

https://www.w3.org/TR/webrtc-stats/#dom-rtcreceivedrtpstreamstats-packetslost

Count of the total number of video pauses experienced by this receiver.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-pausecount

Count the total number of Picture Loss Indication (PLI) packets.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-plicount

Total duration of pauses in seconds.

This is the "ping" between the client and the STUN server. Not to be confused with the E2E RTT documented [here](https://www.w3.org/TR/webrtc-stats/#dom-rtcremoteinboundrtpstreamstats-roundtriptime)

https://www.w3.org/TR/webrtc-stats/#dom-rtcicecandidatepairstats-currentroundtriptime

Number of seconds of frozen video the user has been subjected to.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-totalfreezesduration

Count of the total number of video pauses experienced by this receiver.

https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-totalpausesduration

The response from the `ClosePath` command.

The UUID of the lone face of the resulting solid2D.

The language code is written in.

<details><summary>JSON schema</summary>

```json { "description": "The language code is written in.", "oneOf": [ { "description": "The `go` programming language.", "type": "string", "enum": [ "go" ] }, { "description": "The `python` programming language.", "type": "string", "enum": [ "python" ] }, { "description": "The `node` programming language.", "type": "string", "enum": [ "node" ] } ] } ``` </details>

Output of the code being executed.

<details><summary>JSON schema</summary>

```json { "description": "Output of the code being executed.", "type": "object", "properties": { "output_files": { "description": "The contents of the files requested if they were passed.", "type": "array", "items": { "$ref": "#/components/schemas/OutputFile" } }, "stderr": { "description": "The stderr of the code.", "default": "", "type": "string" }, "stdout": { "description": "The stdout of the code.", "default": "", "type": "string" } } } ``` </details>

The contents of the files requested if they were passed.

Output file contents.

<details><summary>JSON schema</summary>

```json { "description": "Output file contents.", "type": "object", "properties": { "contents": { "description": "The contents of the file. This is base64 encoded so we can ensure it is UTF-8 for JSON.", "type": "string" }, "name": { "description": "The name of the file.", "default": "", "type": "string" } } } ``` </details>

Struct to contain the edge information of a wall of an extrude/rotate/loft/sweep.

Every edge that shared one common vertex with the original edge.

The opposite edge has no common vertices with the original edge. A wall may not have an opposite edge (i.e. a revolve that touches the axis of rotation).

Container that holds a translate, rotate and scale. Defaults to no change, everything stays the same (i.e. the identity function).

How a property of an object should be transformed.

Rotate component of the transform. The rotation is specified as an axis and an angle (xyz are the components of the axis, w is the angle in degrees).

Rotate component of the transform. The rotation is specified as a roll, pitch, yaw.

Describes the file to convert (src) and what it should be converted into (output).

Amount (in the `currency` specified) that will be taken off the subtotal of any invoices for this customer.

Name of the coupon displayed to customers on, for instance invoices, or receipts.

By default the `id` is shown if `name` is not set.

APIML

Endpoints