> For the complete documentation index, see [llms.txt](https://docs.quilt.bio/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.quilt.bio/version-5.0.x/mentalmodel.md).

# Mental model

Quilt represents datasets as *packages*. A package is an immutable collection of related files with a handle of the form `AUTHOR/DESCRIPTION`, a cryptographic *top-hash* (or hash of hashes) that uniquely identifies package contents, and a backing *manifest*.

The manifest is serialized as file that contains *entries*. Manifest entries are tuples of the following form:

`(LOGICAL_KEY, PHYSICAL_KEYS, HASH, METADATA)`

*Logical keys* are user-facing friendly names, like `"README.md"`. *Physical keys* are fully qualified paths to bytes on disk, or bytes in S3. A *hash* is a digest of the physical key's contents, usually SHA-256. *Metadata* are a dictionary that may contain user-defined keys for metadata like bounding boxes, labels, or provenance information (e.g. {"algorithm\_version": "4.4.1"} to indicate how a given file was created).

Package manifests are stored in *registries*. Quilt supports both local disk and Amazon S3 buckets as registry. A registry may store manifests as well as the primary data. S3 was chosen for its widespread adoption, first-class versioning support, and cost/performance profile. The Quilt roadmap includes plans to support more storage formats in the future (e.g. GCP, Azure, NAS, etc.).

By way of illustration first entry of a package manifest for the COCO machine learning dataset are shown below.

```json
{
    "logical_key": "annotations/captions_train2017.json",
    "physical_keys":
    ["s3://quilt-ml-data/data/raw/annotations/captions_train2017.json?versionId=UtzkAN8FP4irtroeN9bfYP1yKzX7ko3G"],
    "size": 91865115,
    "hash": {
    "type": "SHA256",
    "value":
    "4b62086319480e0739ef390d04084515defb9c213ff13605a036061e33314317"},
    "meta": {}
}
```

## Buckets are branches

In Quilt, S3 buckets are analogous to branches in git. Each bucket is a self-contained registry for one or more packages. As package data and schemas are refined, you can promote a package to a new bucket to signify its increased data quality.

We generally recommend a minimum of three buckets for the data lifecycle:

* Raw
* Stage
* Production

![](/files/UOwbkCcnxxdlBsDXUUVd)

See [Quilt workflows](/version-5.0.x/advanced/workflows.md) for more on how you can control data quality with schemas.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.quilt.bio/version-5.0.x/mentalmodel.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.