datasets

Dataset-kind repo extras — episode index, frames index, per-episode playback payload, dataset meta

4 operations · 9 schemas

GET `/v1/datasets/{owner}/{repo}/episodes`

List dataset episodes

listDatasetEpisodes · scope none

Episode index of a dataset-kind repo (LeRobotDataset layout; parsed from repo files — the parser lands in M6, the contract is fixed now). 409 problem if the repo is not a dataset.

Parameters

listDatasetEpisodes parameters
Name	In	Type	Description
`owner`required	path	`Handle`	Repo owner handle (username or org slug).
`repo`required	path	`Slug`	Repo slug.
`limit`	query	`integer` default `20` · ≥ `1` · ≤ `100`	Page size.
`cursor`	query	`string` length ≤ `512`	Opaque cursor from the previous page's `next_cursor`.

Responses

listDatasetEpisodes responses
Status	Description	Body
`200`	Page of episodes.	`EpisodePage`
`404`	Resource not found (or hidden from the caller).	`Problem`application/problem+json
`409`	State conflict (duplicate handle/slug, wrong repo kind, terminal job, ...).	`Problem`application/problem+json

GET `/v1/datasets/{owner}/{repo}/episodes/{episode_index}/frames-index`

Frames index for one episode

getEpisodeFramesIndex · scope none

Maps an episode to the repo files that store its frames (parquet data slices, per-camera videos) so clients can range-read instead of downloading the whole dataset.

Parameters

getEpisodeFramesIndex parameters
Name	In	Type	Description
`owner`required	path	`Handle`	Repo owner handle (username or org slug).
`repo`required	path	`Slug`	Repo slug.
`episode_index`required	path	`integer` ≥ `0`	Zero-based episode index.

Responses

getEpisodeFramesIndex responses
Status	Description	Body
`200`	The frames index.	`FramesIndex`
`404`	Resource not found (or hidden from the caller).	`Problem`application/problem+json
`409`	State conflict (duplicate handle/slug, wrong repo kind, terminal job, ...).	`Problem`application/problem+json

GET `/v1/datasets/{owner}/{repo}/episodes/{episode_index}/series`

Playable payload for one episode

getEpisodeSeries · scope none

Everything a viewer needs to play one episode in a single round trip: presigned per-camera video URLs, the downsampled observation.state / action channels as {t, name, value} points, fps and the episode's task strings. getEpisodeFramesIndex says where the frames live; this says what they are. The server range-reads the parquet columns it needs — it never downloads the whole file — and decimates each channel to at most max_points samples with a min/max algorithm that preserves spikes. 409 problem if the repo is not a dataset, or if its files are not a readable LeRobotDataset.

Parameters

getEpisodeSeries parameters
Name	In	Type	Description
`owner`required	path	`Handle`	Repo owner handle (username or org slug).
`repo`required	path	`Slug`	Repo slug.
`episode_index`required	path	`integer` ≥ `0`	Zero-based episode index.
`max_points`	query	`integer` default `2000` · ≥ `2` · ≤ `2000`	Maximum samples returned per channel. The series is decimated to fit (min/max per bucket, so spikes survive); a shorter episode is returned whole.

Responses

getEpisodeSeries responses
Status	Description	Body
`200`	The episode payload.	`EpisodeSeries`
`404`	Resource not found (or hidden from the caller).	`Problem`application/problem+json
`409`	State conflict (duplicate handle/slug, wrong repo kind, terminal job, ...).	`Problem`application/problem+json

PUT `/v1/datasets/{owner}/{repo}/meta`

Set dataset meta

putDatasetMeta · scope key:write

Full replace of the dataset meta block on a dataset-kind repo (written by uploaders/workers). 409 problem if the repo kind is not dataset.

Parameters

putDatasetMeta parameters
Name	In	Type	Description
`owner`required	path	`Handle`	Repo owner handle (username or org slug).
`repo`required	path	`Slug`	Repo slug.

Request body

application/json · required · DatasetMetaPut

putDatasetMeta request body
Field	Type	Description
`episode_count`required	`integer` ≥ `0`
`fps`	`number \| null` > `0`
`robot_type`	`string \| null` length ≤ `128`
`total_frames`required	`integer` ≥ `0`

Responses

putDatasetMeta responses
Status	Description	Body
`200`	Stored dataset meta.	`DatasetMeta`
`401`	Missing or invalid credentials.	`Problem`application/problem+json
`403`	Authenticated but not allowed (visibility, membership or scope).	`Problem`application/problem+json
`404`	Resource not found (or hidden from the caller).	`Problem`application/problem+json
`409`	State conflict (duplicate handle/slug, wrong repo kind, terminal job, ...).	`Problem`application/problem+json
`422`	Request failed validation.	`Problem`application/problem+json

Schemas (9)

The schemas these operations reach before any other tag’s do. A type that links elsewhere is rendered on that tag’s page.

EpisodePage

object

EpisodePage fields
Field	Type	Description
`items`required	`array of Episode`
`next_cursor`required	`string \| null`

FramesIndex

object

Where one episode's frames live, for range reads (v0 of the index — M6's parser fills it from the LeRobotDataset layout).

FramesIndex fields
Field	Type	Description
`episode_index`required	`integer` ≥ `0`
`fps`	`number \| null` > `0`
`num_frames`required	`integer` ≥ `0`
`files`required	`array of FrameFile`

EpisodeSeries

object

One episode, ready to play: video URLs plus decimated numeric channels on a shared time axis (seconds from episode start).

EpisodeSeries fields
Field	Type	Description
`episode_index`required	`integer` ≥ `0`
`fps`	`number \| null` > `0`
`num_frames`required	`integer` ≥ `0`	Frames in the episode (rows read from the data file).
`duration_s`	`number \| null` ≥ `0`
`tasks`required	`array of string`
`sampled_points`required	`integer` ≥ `0`	Samples returned per channel after decimation.
`downsample`required	`string` one of `none` · `minmax`	How the series was reduced. `minmax` keeps the minimum and the maximum of every bucket, so a one-frame spike survives; `none` means the episode was short enough to return whole.
`channels`required	`array of EpisodeChannel`
`points`required	`array of EpisodeSeriesPoint`	Flat `{t, name, value}` samples across every channel.
`videos`required	`array of EpisodeVideo`
`warnings`required	`array of string`	Human-readable notes about what could not be produced (a camera with no file, a layout feature this version does not map).

DatasetMetaPut

object

DatasetMetaPut fields
Field	Type	Description
`episode_count`required	`integer` ≥ `0`
`fps`	`number \| null` > `0`
`robot_type`	`string \| null` length ≤ `128`
`total_frames`required	`integer` ≥ `0`

Episode

object

Episode fields
Field	Type	Description
`index`required	`integer` ≥ `0`
`num_frames`required	`integer` ≥ `0`
`tasks`	`array of string`	Natural-language task strings attached to the episode.

FrameFile

object

FrameFile fields
Field	Type	Description
`path`required	`FilePath`
`role`required	`string` one of `data` · `video`	What this file holds for the episode.
`format`	`string \| null` length ≤ `32`	File format hint (e.g. `parquet`, `mp4`).
`camera_key`	`string \| null` length ≤ `128`	Camera stream key for `video` files.
`row_start`	`integer \| null` ≥ `0`	First row of this episode inside a shared data file.
`row_end`	`integer \| null` ≥ `0`	Exclusive end row of this episode inside a shared data file.

EpisodeChannel

object

One plottable scalar track inside a vector feature.

EpisodeChannel fields
Field	Type	Description
`name`required	`string` length ≤ `128`	Channel label — the joint/motor name from `info.json` when the feature declares one, else `{feature}[{index}]`.
`group`required	`string` one of `state` · `action` · `other`	Which feature family this channel came from.
`feature`required	`string` length ≤ `128`	Source feature key, e.g. `observation.state`.
`index`required	`integer` ≥ `0`	Position of this channel inside the feature vector.

EpisodeSeriesPoint

object

One sample of one channel.

EpisodeSeriesPoint fields
Field	Type	Description
`t`required	`number`	Seconds from the start of the episode.
`name`required	`string` length ≤ `128`	Channel name (matches an `EpisodeChannel.name`).
`value`required	`number`

EpisodeVideo

object

One camera stream for this episode, ready to play.

EpisodeVideo fields
Field	Type	Description
`camera_key`required	`string` length ≤ `128`	Feature key of the camera, e.g. `observation.images.front`.
`path`required	`FilePath`
`url`required	`string (uri)`	Short-lived presigned GET for the video file.
`expires_at`required	`string (date-time)`
`width`	`integer \| null` ≥ `1`
`height`	`integer \| null` ≥ `1`

datasets

GET /v1/datasets/{owner}/{repo}/episodes

Parameters

Responses

GET /v1/datasets/{owner}/{repo}/episodes/{episode_index}/frames-index

Parameters

Responses

GET /v1/datasets/{owner}/{repo}/episodes/{episode_index}/series

Parameters

Responses

PUT /v1/datasets/{owner}/{repo}/meta

Parameters

Request body

Responses

Schemas (9)

EpisodePage

FramesIndex

EpisodeSeries

DatasetMetaPut

Episode

FrameFile

EpisodeChannel

EpisodeSeriesPoint

EpisodeVideo

GET `/v1/datasets/{owner}/{repo}/episodes`

GET `/v1/datasets/{owner}/{repo}/episodes/{episode_index}/frames-index`

GET `/v1/datasets/{owner}/{repo}/episodes/{episode_index}/series`

PUT `/v1/datasets/{owner}/{repo}/meta`