spacedrive/core/prisma/schema.prisma
Ericson "Fogo" Soares 7c90bcb95b
[ENG-1479] AI Prototype (#1845)
* First draft on image labeling

* Fixing execution providers for other OSs

* Better error handling and shutdown

* Working with shallow media processor

* bruh

* Fix warnings

* Now hooked to media processor job

* Link desktop app with libonnxruntime to avoid TLS error during startup

* Be able to change models on runtime
Revert to use labels table instead of tags

* A bug on a model-less inference

* Show AI labels on Inspector
 - Change yolo inference to use half precision
 - Add labels api to core

* Remove LD_PRELOAD

* Fix race condition on model executor shutdown

* Don't load all images in memory moron

* Embeed yolo model in prod build
 - Change yolo model path to new one relative to executable

* Disable volume watcher on linux, it was crashing the app
 - Invalidate labels when they are updated

* Rust fmt

* Minor changes

* Gate onnxruntime linking to the ai-models feature

* Add build script to sd-server to handle onnxruntime linking workaround

* Move AI stuff to its own crate and normalize deps

* Rust fmt

* Don't regenerate labels unless asked to

* Now blazingly fast

* Bad merge

* Fix

* Fix

* Add backend logic to download extra yolo models

* Add models api route
 - Add api call to get available model version
 - Add api call to change the model version

* Improve new model download logic
 - Add frontend to change image labeler model

* Fix new model downloader

* Fix model select width

* invalidate labels count after media_processor generates a new output

* Rename AI crate and first draft on download notifications

* fix types

---------

Co-authored-by: Vítor Vasconcellos <vasconcellos.dev@gmail.com>
Co-authored-by: Brendan Allan <brendonovich@outlook.com>
2023-12-19 09:28:57 +00:00

588 lines
15 KiB
Plaintext

datasource db {
provider = "sqlite"
url = "file:dev.db"
}
generator client {
provider = "cargo prisma"
output = "../../crates/prisma/src/prisma"
module_path = "prisma"
client_format = "folder"
}
generator sync {
provider = "cargo prisma-sync"
output = "../../crates/prisma/src/prisma_sync"
client_format = "folder"
}
//// Sync ////
model SharedOperation {
id Bytes @id
timestamp BigInt
model String
record_id Bytes
// Enum: ??
kind String
data Bytes
instance_id Int
instance Instance @relation(fields: [instance_id], references: [id])
// attestation Bytes
@@map("shared_operation")
}
model RelationOperation {
id Bytes @id
timestamp BigInt
relation String
item_id Bytes
group_id Bytes
kind String
data Bytes
instance_id Int
instance Instance @relation(fields: [instance_id], references: [id])
@@map("relation_operation")
}
/// @deprecated: This model has to exist solely for backwards compatibility.
model Node {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String
// Enum: sd_core::node::Platform
platform Int
date_created DateTime
identity Bytes? // TODO: Change to required field in future
node_peer_id String? // TODO: Remove as part of - https://linear.app/spacedriveapp/issue/ENG-757/p2p-library-portability
@@map("node")
}
/// @local(id: pub_id)
// represents a single `.db` file (SQLite DB) that is paired to the current library.
// A `LibraryInstance` is always owned by a single `Node` but it's possible for that node to change (or two to be owned by a single node).
model Instance {
id Int @id @default(autoincrement()) // This is is NOT globally unique
pub_id Bytes @unique // This UUID is meaningless and exists soley cause the `uhlc::ID` must be 16-bit. Really this should be derived from the `identity` field.
// Enum: sd_core::p2p::IdentityOrRemoteIdentity
identity Bytes
node_id Bytes
node_name String
// Enum: sd_core::node::Platform
node_platform Int
last_seen DateTime // Time core started for owner, last P2P message for P2P node
date_created DateTime
// clock timestamp for sync
timestamp BigInt?
// attestation Bytes
locations Location[]
SharedOperation SharedOperation[]
RelationOperation RelationOperation[]
CloudSharedOperation CloudSharedOperation[]
CloudRelationOperation CloudRelationOperation[]
@@map("instance")
}
model Statistics {
id Int @id @default(autoincrement())
date_captured DateTime @default(now())
total_object_count Int @default(0)
library_db_size String @default("0")
total_bytes_used String @default("0")
total_bytes_capacity String @default("0")
total_unique_bytes String @default("0")
total_bytes_free String @default("0")
preview_media_bytes String @default("0")
@@map("statistics")
}
/// @local
model Volume {
id Int @id @default(autoincrement())
name String
mount_point String
total_bytes_capacity String @default("0")
total_bytes_available String @default("0")
disk_type String?
filesystem String?
is_system Boolean @default(false)
date_modified DateTime @default(now())
@@unique([mount_point, name])
@@map("volume")
}
/// @shared(id: pub_id)
model Location {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String?
path String?
total_capacity Int?
available_capacity Int?
size_in_bytes Bytes?
is_archived Boolean?
generate_preview_media Boolean?
sync_preview_media Boolean?
hidden Boolean?
date_created DateTime?
instance_id Int?
instance Instance? @relation(fields: [instance_id], references: [id], onDelete: SetNull)
file_paths FilePath[]
indexer_rules IndexerRulesInLocation[]
@@map("location")
}
/// @shared(id: pub_id)
model FilePath {
id Int @id @default(autoincrement())
pub_id Bytes @unique
is_dir Boolean?
// content addressable storage id - blake3 sampled checksum
cas_id String?
// full byte contents digested into blake3 checksum
integrity_checksum String?
// location that owns this path
location_id Int?
location Location? @relation(fields: [location_id], references: [id], onDelete: SetNull)
// the path of the file relative to its location
materialized_path String?
// the name and extension, MUST have 'COLLATE NOCASE' in migration
name String?
extension String?
hidden Boolean?
size_in_bytes String? // deprecated
size_in_bytes_bytes Bytes?
inode Bytes? // This is actually an unsigned 64 bit integer, but we don't have this type in SQLite
// the unique Object for this file path
object_id Int?
object Object? @relation(fields: [object_id], references: [id], onDelete: SetNull)
key_id Int? // replacement for encryption
// permissions String?
date_created DateTime?
date_modified DateTime?
date_indexed DateTime?
// key Key? @relation(fields: [key_id], references: [id])
@@unique([location_id, materialized_path, name, extension])
@@unique([location_id, inode])
@@index([location_id])
@@index([location_id, materialized_path])
@@map("file_path")
}
/// @shared(id: pub_id)
model Object {
id Int @id @default(autoincrement())
pub_id Bytes @unique
// Enum: sd_file_ext::kind::ObjectKind
kind Int?
key_id Int?
// handy ways to mark an object
hidden Boolean?
favorite Boolean?
important Boolean?
// if we have generated preview media for this object on at least one Node
// commented out for now by @brendonovich since they they're irrelevant to the sync system
// has_thumbnail Boolean?
// has_thumbstrip Boolean?
// has_video_preview Boolean?
// TODO: change above to:
// has_generated_thumbnail Boolean @default(false)
// has_generated_thumbstrip Boolean @default(false)
// has_generated_video_preview Boolean @default(false)
// integration with ipfs
// ipfs_id String?
// plain text note
note String?
// the original known creation date of this object
date_created DateTime?
date_accessed DateTime?
tags TagOnObject[]
labels LabelOnObject[]
albums ObjectInAlbum[]
spaces ObjectInSpace[]
file_paths FilePath[]
// comments Comment[]
media_data MediaData?
// key Key? @relation(fields: [key_id], references: [id])
@@map("object")
}
// if there is a conflicting cas_id, the conficting file should be updated to have a larger cas_id as
//the field is unique, however this record is kept to tell the indexer (upon discovering this CAS) that
//there is alternate versions of the file and to check by a full integrity hash to define for which to associate with.
// @brendan: nah this probably won't fly
// model FileConflict {
// original_object_id Int @unique
// detactched_object_id Int @unique
// @@map("file_conflict")
// }
// keys allow us to know exactly which files can be decrypted with a given key
// they can be "mounted" to a client, and then used to decrypt files automatically
/// @shared(id: uuid)
// model Key {
// id Int @id @default(autoincrement())
// // uuid to identify the key
// uuid String @unique
// version String
// key_type String
// // the name that the user sets
// name String?
// // is this key the default for encryption?
// // was not tagged as unique as i'm not too sure if PCR will handle it
// // can always be tagged as unique, the keys API will need updating to use `find_unique()`
// default Boolean @default(false)
// // nullable if concealed for security
// date_created DateTime? @default(now())
// // encryption algorithm used to encrypt the key
// algorithm String
// // hashing algorithm used for hashing the key with the content salt
// hashing_algorithm String
// // salt used for encrypting data with this key
// content_salt Bytes
// // the *encrypted* master key (48 bytes)
// master_key Bytes
// // the nonce used for encrypting the master key
// master_key_nonce Bytes
// // the nonce used for encrypting the key
// key_nonce Bytes
// // the *encrypted* key
// key Bytes
// // the salt used for deriving the KEK (used for encrypting the master key) from the root key
// salt Bytes
// automount Boolean @default(false)
// objects Object[]
// file_paths FilePath[]
// @@map("key")
// }
model MediaData {
id Int @id @default(autoincrement())
resolution Bytes?
media_date Bytes?
media_location Bytes?
camera_data Bytes?
artist String?
description String?
copyright String?
exif_version String?
// purely for sorting/ordering, never sent to the frontend as they'd be useless
// these are also usually one-way, and not reversible
// (e.g. we can't get `MediaDate::Utc(2023-09-26T22:04:37+01:00)` from `1695758677` as we don't store the TZ)
epoch_time BigInt? // time since unix epoch
// video-specific
// duration Int?
// fps Int?
// streams Int?
// video_codec String? // eg: "h264, h265, av1"
// audio_codec String? // eg: "opus"
object_id Int @unique
object Object @relation(fields: [object_id], references: [id], onDelete: Cascade)
@@map("media_data")
}
//// Tag ////
/// @shared(id: pub_id)
model Tag {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String?
color String?
is_hidden Boolean? // user hidden entire tag
date_created DateTime?
date_modified DateTime?
tag_objects TagOnObject[]
@@map("tag")
}
/// @relation(item: tag, group: object)
model TagOnObject {
tag_id Int
tag Tag @relation(fields: [tag_id], references: [id], onDelete: Restrict)
object_id Int
object Object @relation(fields: [object_id], references: [id], onDelete: Restrict)
date_created DateTime?
@@id([tag_id, object_id])
@@map("tag_on_object")
}
//// Label ////
model Label {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String @unique
date_created DateTime @default(now())
date_modified DateTime @default(now())
label_objects LabelOnObject[]
@@map("label")
}
model LabelOnObject {
date_created DateTime @default(now())
label_id Int
label Label @relation(fields: [label_id], references: [id], onDelete: Restrict)
object_id Int
object Object @relation(fields: [object_id], references: [id], onDelete: Restrict)
@@id([label_id, object_id])
@@map("label_on_object")
}
//// Space ////
model Space {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String?
description String?
date_created DateTime?
date_modified DateTime?
objects ObjectInSpace[]
@@map("space")
}
model ObjectInSpace {
space_id Int
space Space @relation(fields: [space_id], references: [id], onDelete: Restrict)
object_id Int
object Object @relation(fields: [object_id], references: [id], onDelete: Restrict)
@@id([space_id, object_id])
@@map("object_in_space")
}
//// Job ////
model Job {
id Bytes @id
name String?
action String? // Will be composed of "{action_description}(-{children_order})*"
// Enum: sd_core::job::job_manager:JobStatus
status Int? // 0 = Queued
// List of errors, separated by "\n\n" in case of failed jobs or completed with errors
errors_text String?
data Bytes? // Serialized data to be used on pause/resume
metadata Bytes? // Serialized metadata field with info about the job after completion
parent_id Bytes?
task_count Int?
completed_task_count Int?
date_estimated_completion DateTime? // Estimated timestamp that the job will be complete at
date_created DateTime?
date_started DateTime? // Started execution
date_completed DateTime? // Finished execution
parent Job? @relation("jobs_dependency", fields: [parent_id], references: [id], onDelete: SetNull)
children Job[] @relation("jobs_dependency")
@@map("job")
}
//// Album ////
model Album {
id Int @id
pub_id Bytes @unique
name String?
is_hidden Boolean?
date_created DateTime?
date_modified DateTime?
objects ObjectInAlbum[]
@@map("album")
}
model ObjectInAlbum {
date_created DateTime?
album_id Int
album Album @relation(fields: [album_id], references: [id], onDelete: NoAction)
object_id Int
object Object @relation(fields: [object_id], references: [id], onDelete: NoAction)
@@id([album_id, object_id])
@@map("object_in_album")
}
//// Comment ////
// model Comment {
// id Int @id @default(autoincrement())
// pub_id Bytes @unique
// content String
// date_created DateTime @default(now())
// date_modified DateTime @default(now())
// object_id Int?
// object Object? @relation(fields: [object_id], references: [id])
// @@map("comment")
// }
//// Indexer Rules ////
model IndexerRule {
id Int @id @default(autoincrement())
pub_id Bytes @unique
name String?
default Boolean?
rules_per_kind Bytes?
date_created DateTime?
date_modified DateTime?
locations IndexerRulesInLocation[]
@@map("indexer_rule")
}
model IndexerRulesInLocation {
location_id Int
location Location @relation(fields: [location_id], references: [id], onDelete: Restrict)
indexer_rule_id Int
indexer_rule IndexerRule @relation(fields: [indexer_rule_id], references: [id], onDelete: Restrict)
@@id([location_id, indexer_rule_id])
@@map("indexer_rule_in_location")
}
/// @shared(id: key)
model Preference {
key String @id
value Bytes?
@@map("preference")
}
model Notification {
id Int @id @default(autoincrement())
read Boolean @default(false)
// Enum: crate::api::notifications::NotificationData
data Bytes
expires_at DateTime?
@@map("notification")
}
model SavedSearch {
id Int @id @default(autoincrement())
pub_id Bytes @unique
search String?
filters String?
name String?
icon String?
description String?
// order Int? // Add this line to include ordering
date_created DateTime?
date_modified DateTime?
@@map("saved_search")
}
model CloudSharedOperation {
id Bytes @id
timestamp BigInt
model String
record_id Bytes
// Enum: ??
kind String
data Bytes
instance_id Int
instance Instance @relation(fields: [instance_id], references: [id])
@@map("cloud_shared_operation")
}
model CloudRelationOperation {
id Bytes @id
timestamp BigInt
relation String
item_id Bytes
group_id Bytes
kind String
data Bytes
instance_id Int
instance Instance @relation(fields: [instance_id], references: [id])
@@map("cloud_relation_operation")
}