File size: 37,585 Bytes
c61c435 f8c73f9 c61c435 f8c73f9 c61c435 f8c73f9 c61c435 f8c73f9 c61c435 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 | from __future__ import annotations
from pathlib import Path
import os
import tempfile
import threading
import time
from typing import Any
from urllib.parse import parse_qs
from adam.assets import Asset, AssetRegistry
from adam.commands import CommandValidationError, TrainingCommand
from adam.dataset_registry import DatasetRecord, DatasetRegistry
from adam.dataset_lab import scan_dataset
from adam.generations import build_generation_plan, generation_tools
from adam.remote_api import (
RemoteApiError,
RemoteResponse,
bounded_float,
bounded_int,
bounded_text,
coerce_json_object,
error_response,
json_response,
media_response,
parse_pagination,
sanitized_arguments,
)
from adam.remote_dispatcher import RemoteCommandDispatcher
from adam.remote_media import OpaqueIdCodec, RemoteMediaStore
from adam.studio import caption_path, image_files, StudioStore
from adam.training_assistant import append_preflight_summary
class RemoteV1Service:
"""Versioned ADAM Remote API built around existing ADAM services."""
def __init__(
self,
*,
root: Path,
config: Any,
jobs: Any,
planner: Any,
dispatcher: RemoteCommandDispatcher,
codec: OpaqueIdCodec,
media: RemoteMediaStore,
auto_approve_training: Any,
) -> None:
self.root = root.resolve()
self.config = config
self.jobs = jobs
self.planner = planner
self.dispatcher = dispatcher
self.codec = codec
self.media = media
self.auto_approve_training = auto_approve_training
self._asset_fallback: AssetRegistry | None = None
self._studio: StudioStore | None = None
# The dashboard requests both schemas at once. They share an asset
# registry, so build one short-lived catalog instead of asking two
# request threads to rediscover and rewrite it concurrently.
self._model_catalog_lock = threading.RLock()
self._model_catalog: list[dict[str, Any]] = []
self._model_catalog_at = 0.0
def route(
self,
method: str,
path: str,
query: str = "",
payload: dict[str, Any] | None = None,
) -> RemoteResponse | None:
if not path.startswith("/api/v1/"):
return None
parts = [part for part in path.removeprefix("/api/v1/").split("/") if part]
parsed_query = parse_qs(query)
try:
if method == "GET":
return self._get(parts, parsed_query)
if method == "POST":
return self._post(parts, coerce_json_object(payload or {}))
except RemoteApiError as exc:
return error_response(str(exc), status=exc.status)
except (CommandValidationError, ValueError) as exc:
return error_response(str(exc), status=400)
except Exception as exc:
return error_response(f"ADAM could not finish that remote request: {exc}", status=500)
return error_response("Unsupported Remote API method.", status=405)
def _get(self, parts: list[str], query: dict[str, list[str]]) -> RemoteResponse:
if parts == ["datasets"]:
return json_response({"ok": True, "datasets": self.datasets()})
if parts == ["datasets", "locations"]:
return json_response({"ok": True, "locations": self.dataset_locations()})
if parts == ["training", "presets"]:
return json_response({"ok": True, "presets": self.training_presets()})
if len(parts) == 2 and parts[0] == "datasets":
return json_response({"ok": True, "dataset": self.dataset_detail(parts[1])})
if len(parts) == 3 and parts[0] == "datasets" and parts[2] == "thumbnail":
return self.dataset_thumbnail(parts[1], query)
if len(parts) == 3 and parts[0] == "datasets" and parts[2] == "items":
return json_response({"ok": True, **self.dataset_items(parts[1], query)})
if len(parts) == 2 and parts[0] == "media":
return self.media_file(parts[1], query)
if parts == ["models"]:
return json_response({"ok": True, "models": self.models()})
if parts == ["training", "schema"]:
return json_response({"ok": True, **self.training_schema()})
if parts == ["generation", "schema"]:
return json_response({"ok": True, **self.generation_schema()})
if len(parts) == 2 and parts[0] == "jobs":
return json_response({"ok": True, "job": self.job_detail(parts[1])})
raise RemoteApiError("Unknown Remote API endpoint.", status=404)
def _post(self, parts: list[str], payload: dict[str, Any]) -> RemoteResponse:
if len(parts) == 5 and parts[0] == "datasets" and parts[2] == "items" and parts[4] == "caption":
return json_response({"ok": True, **self.update_caption(parts[1], parts[3], payload)})
if len(parts) == 5 and parts[0] == "datasets" and parts[2] == "items" and parts[4] == "decision":
return json_response({"ok": True, **self.update_decision(parts[1], parts[3], payload)})
if len(parts) == 3 and parts[0] == "datasets" and parts[2] == "favorite":
return json_response({"ok": True, **self.favorite_dataset(parts[1], payload)})
if len(parts) == 3 and parts[0] == "datasets" and parts[2] == "use":
return json_response({"ok": True, "dataset": self.use_dataset(parts[1])})
if parts == ["training", "plan"]:
return json_response({"ok": True, "plan": self.training_plan(payload)})
if parts == ["training", "start"]:
return json_response({"ok": True, **self.start_training(payload)})
if parts == ["generation", "start"]:
return json_response({"ok": True, **self.start_generation(payload)})
raise RemoteApiError("Unknown Remote API endpoint.", status=404)
def _assets(self) -> AssetRegistry:
assets = getattr(self.planner, "assets", None)
if assets is None:
if self._asset_fallback is None:
self._asset_fallback = AssetRegistry(self.root)
assets = self._asset_fallback
if hasattr(assets, "discover"):
assets.discover(self.config)
return assets
def _registry(self) -> Any:
registry = getattr(self.planner, "registry", None)
if registry is None:
raise RemoteApiError("The ADAM tool registry is not available.", status=503)
return registry
def _studio_store(self) -> StudioStore:
if self._studio is None:
self._studio = StudioStore(self.root)
return self._studio
def _dataset_registry(self) -> DatasetRegistry:
return DatasetRegistry(self.root, self.config)
def _asset_public_id(self, asset: Asset) -> str:
return self.codec.encode({"kind": asset.kind, "asset_id": asset.id})
def _asset_from_id(self, public_id: str, *, kind: str = "") -> Asset:
payload = self.codec.decode(public_id)
asset_id = str(payload.get("asset_id", ""))
expected_kind = kind or str(payload.get("kind", ""))
for asset in self._assets().assets:
if asset.id == asset_id and (not expected_kind or asset.kind == expected_kind):
return asset
raise RemoteApiError("Unknown resource id.", status=404)
def _item_path(self, item_id: str, *, dataset: Asset | None = None) -> tuple[Asset, Path, int]:
payload = self.codec.decode(item_id)
if payload.get("kind") != "dataset_image":
raise RemoteApiError("Unknown media id.", status=404)
asset_id = str(payload.get("asset_id", ""))
index = bounded_int(payload.get("index"), minimum=0, maximum=100_000, default=0, label="Image index")
if dataset is not None and dataset.id != asset_id:
raise RemoteApiError("Dataset image does not belong to that dataset.", status=403)
dataset_asset = dataset or next((item for item in self._assets().assets if item.id == asset_id and item.kind == "dataset"), None)
if dataset_asset is None:
raise RemoteApiError("Unknown dataset image.", status=404)
paths = image_files(dataset_asset.path, limit=100_000)
if index >= len(paths):
raise RemoteApiError("Dataset image is no longer available.", status=404)
path = paths[index].resolve()
try:
path.relative_to(Path(dataset_asset.path).expanduser().resolve())
except ValueError as exc:
raise RemoteApiError("Dataset image is outside its dataset.", status=403) from exc
return dataset_asset, path, index
@staticmethod
def _dataset_path(asset: Asset, path: Path) -> Path:
resolved = path.expanduser().resolve()
if not resolved.is_relative_to(Path(asset.path).expanduser().resolve()):
raise RemoteApiError("File is outside its dataset.", status=403)
return resolved
def _dataset_counts(self, asset: Asset) -> dict[str, int]:
review = self._studio_store().review(asset.path)
images = image_files(asset.path, limit=100_000)
resolved = {str(path.resolve()) for path in images}
keep = sum(1 for path, decision in review.decisions.items() if decision == "keep" and path in resolved)
reject = sum(1 for path, decision in review.decisions.items() if decision == "reject" and path in resolved)
return {
"accepted": keep,
"rejected": reject,
"unreviewed": max(0, len(images) - keep - reject),
}
def _dimensions(self, path: Path) -> dict[str, int]:
try:
from PIL import Image
with Image.open(path) as image:
return {"width": int(image.width), "height": int(image.height)}
except Exception:
return {"width": 0, "height": 0}
def datasets(self) -> list[dict[str, Any]]:
assets = self._assets()
records = self._dataset_registry().discover(asset_registry=assets)
by_path = {str(Path(asset.path).expanduser().resolve()): asset for asset in assets.assets if asset.kind == "dataset"}
rows = []
for record in records[:300]:
asset = by_path.get(str(Path(record.path).expanduser().resolve()))
if asset is None and record.exists:
asset = assets.register(
kind="dataset",
name=record.name,
path=record.path,
metadata={
"dataset_registry_source": record.source,
"dataset_location_id": record.location_id,
},
persist=False,
)
if asset is None:
continue
rows.append(self._dataset_summary(asset, record))
if rows:
assets.save()
return rows
def _dataset_summary(self, asset: Asset, record: DatasetRecord | None = None) -> dict[str, Any]:
if record is None:
record = self._dataset_registry().record_for_path(asset.path)
exists = Path(asset.path).is_dir()
public_id = self._asset_public_id(asset)
if exists:
review_state = self._studio_store().review(asset.path)
accepted = sum(1 for decision in review_state.decisions.values() if decision == "keep")
rejected = sum(1 for decision in review_state.decisions.values() if decision == "reject")
review = {
"accepted": accepted,
"rejected": rejected,
"unreviewed": max(0, record.image_count - accepted - rejected),
}
else:
review = {"accepted": 0, "rejected": 0, "unreviewed": 0}
warnings = list(record.warnings)
if not exists and "Dataset folder is unavailable." not in warnings:
warnings.append("Dataset folder is unavailable.")
return {
"id": public_id,
"name": asset.name or record.name,
"exists": exists,
"available": exists,
"source": record.source,
"location_id": record.location_id,
"favorite": record.favorite,
"last_used_at": record.last_used_at,
"item_count": record.item_count,
"image_count": record.image_count,
"video_count": record.video_count,
"caption_count": record.caption_count,
"missing_caption_count": record.missing_caption_count,
"dataset_format": record.dataset_format,
"warnings": warnings[:6],
"thumbnail_url": f"/api/v1/datasets/{public_id}/thumbnail" if exists else "",
"review": review,
}
def dataset_locations(self) -> list[dict[str, Any]]:
registry = self._dataset_registry()
return [
{
"id": item.id,
"name": item.name,
"source": item.source,
"exists": Path(item.path).is_dir(),
"available": Path(item.path).is_dir(),
"last_seen_at": item.last_seen_at,
}
for item in registry.known_locations()
][:200]
def _first_image(self, folder: str | Path) -> Path | None:
root = Path(folder).expanduser()
if not root.is_dir():
return None
try:
for path in root.rglob("*"):
if path.is_file() and path.suffix.casefold() in {".png", ".jpg", ".jpeg", ".webp", ".bmp", ".gif"}:
return path
except OSError:
return None
return None
def dataset_detail(self, dataset_id: str) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
registry = self._dataset_registry()
record = registry.record_for_path(asset.path)
if not Path(asset.path).is_dir():
return self._dataset_summary(asset, record)
report = scan_dataset(asset.path, limit=300)
registry.refresh_async(asset.path, source=record.source, location_id=record.location_id)
return {
"id": self._asset_public_id(asset),
"name": asset.name,
"exists": True,
"available": True,
"source": record.source,
"location_id": record.location_id,
"favorite": record.favorite,
"last_used_at": record.last_used_at,
"image_count": report.image_count,
"video_count": report.video_count,
"item_count": report.image_count + report.video_count,
"caption_count": report.caption_count,
"missing_caption_count": report.missing_caption_count,
"duplicate_groups": report.duplicate_groups,
"dimensions": dict(list(report.dimensions.items())[:20]),
"extensions": report.extensions,
"warnings": report.warnings,
"dataset_format": record.dataset_format,
"thumbnail_url": f"/api/v1/datasets/{dataset_id}/thumbnail",
"review": self._dataset_counts(asset),
}
def dataset_thumbnail(self, dataset_id: str, query: dict[str, list[str]]) -> RemoteResponse:
asset = self._asset_from_id(dataset_id, kind="dataset")
source = self._first_image(asset.path)
if source is None:
raise RemoteApiError("Dataset thumbnail is not available.", status=404)
source = self._dataset_path(asset, source)
size = bounded_int((query.get("size") or ["320"])[0], minimum=64, maximum=640, default=320, label="Media size")
media = self.media.thumbnail(source, size=size)
return media_response(media.path.read_bytes(), media.content_type, cache_seconds=86400)
def dataset_items(self, dataset_id: str, query: dict[str, list[str]]) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
page = parse_pagination(query, default_size=24, max_size=60)
paths = image_files(asset.path, limit=100_000)
total = len(paths)
review = self._studio_store().review(asset.path)
rows = []
for index, path in enumerate(paths[page["offset"]: page["offset"] + page["limit"]], page["offset"]):
path = self._dataset_path(asset, path)
item_id = self.media.media_id(kind="dataset_image", asset_id=asset.id, index=index)
caption_file = self._dataset_path(asset, caption_path(path))
try:
caption = ""
if caption_file.is_file():
with caption_file.open(encoding="utf-8") as handle:
caption = handle.read(4000)
except (OSError, UnicodeError):
caption = ""
rows.append({
"id": item_id,
"display_name": path.name,
"dimensions": self._dimensions(path),
"caption": caption,
"has_caption": bool(caption.strip()),
"decision": review.decisions.get(str(path.resolve()), "unreviewed"),
"thumbnail_url": f"/api/v1/media/{item_id}?size=320",
"preview_url": f"/api/v1/media/{item_id}?size=960",
})
return {
"dataset": self.dataset_detail(dataset_id),
"items": rows,
"pagination": {
"page": page["page"],
"page_size": page["page_size"],
"total": total,
"has_next": page["offset"] + page["limit"] < total,
},
}
def media_file(self, media_id: str, query: dict[str, list[str]]) -> RemoteResponse:
_dataset, source, _index = self._item_path(media_id)
size = bounded_int((query.get("size") or ["320"])[0], minimum=64, maximum=1200, default=320, label="Media size")
media = self.media.thumbnail(source, size=size)
return media_response(media.path.read_bytes(), media.content_type, cache_seconds=86400)
def update_caption(self, dataset_id: str, item_id: str, payload: dict[str, Any]) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
_asset, image, _index = self._item_path(item_id, dataset=asset)
caption = bounded_text(payload.get("caption", ""), max_length=4000, label="Caption")
target = self._dataset_path(asset, caption_path(image))
# Replace the file instead of writing through a possible hard link.
temporary = None
try:
with tempfile.NamedTemporaryFile(mode="w", encoding="utf-8", dir=target.parent, delete=False) as handle:
temporary = Path(handle.name)
handle.write(caption.rstrip() + ("\n" if caption else ""))
os.replace(temporary, target)
finally:
if temporary is not None:
temporary.unlink(missing_ok=True)
return {"message": "Caption saved.", "item": {"id": item_id, "caption": caption, "has_caption": bool(caption)}}
def update_decision(self, dataset_id: str, item_id: str, payload: dict[str, Any]) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
_asset, image, _index = self._item_path(item_id, dataset=asset)
decision = bounded_text(payload.get("decision", "unreviewed"), max_length=32, label="Decision")
if decision not in {"keep", "reject", "unreviewed"}:
raise RemoteApiError("Decision must be keep, reject, or unreviewed.")
self._studio_store().set_decision(asset.path, str(image), decision)
return {"message": "Review decision saved.", "item": {"id": item_id, "decision": decision}}
def favorite_dataset(self, dataset_id: str, payload: dict[str, Any]) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
enabled = bool(payload.get("favorite", True))
record = self._dataset_registry().favorite(asset.path, enabled)
return {
"message": "Dataset favorite updated.",
"dataset": self._dataset_summary(asset, record),
}
def use_dataset(self, dataset_id: str) -> dict[str, Any]:
asset = self._asset_from_id(dataset_id, kind="dataset")
record = self._dataset_registry().touch(asset.path)
return self._dataset_summary(asset, record)
def models(self) -> list[dict[str, Any]]:
with self._model_catalog_lock:
if time.monotonic() - self._model_catalog_at < 1.0:
return list(self._model_catalog)
assets = self._assets()
experiment_by_model: dict[str, Any] = {}
try:
for run in getattr(getattr(self.jobs, "experiments", None), "list_runs", lambda limit=100: [])(limit=100):
experiment_by_model.setdefault(run.model_name, run)
except Exception:
experiment_by_model = {}
rows = []
for asset in assets.assets:
if asset.kind not in {"model", "base_model"} or not Path(asset.path).exists():
continue
dataset = next((item for item in assets.assets if item.id == asset.dataset_id), None)
metadata = dict(asset.metadata or {})
trigger_word = str(metadata.get("trigger_word") or (asset.name if asset.trainer == "lora" else ""))
latest = experiment_by_model.get(asset.name)
rows.append({
"id": self._asset_public_id(asset),
"name": asset.name,
"kind": asset.kind,
"architecture": asset.trainer or ("stable_diffusion" if asset.kind == "base_model" else ""),
"trainer": asset.trainer,
"checkpoint_name": Path(asset.checkpoint or asset.path).name,
"dataset": None if dataset is None else {"id": self._asset_public_id(dataset), "name": dataset.name},
"epochs": asset.epochs,
"trigger_word": trigger_word,
"latest_experiment": None if latest is None else {
"id": latest.id,
"status": latest.status,
"resolution": latest.resolution,
"dataset_name": latest.dataset_name,
"trigger_word": getattr(latest, "trigger_word", ""),
},
})
self._model_catalog = rows[:300]
self._model_catalog_at = time.monotonic()
return list(self._model_catalog)
def training_schema(self) -> dict[str, Any]:
registry = self._registry()
trainers = []
for tool in registry.enabled():
if not tool.id.endswith("_trainer"):
continue
trainer = tool.id.removesuffix("_trainer")
schema = registry.model_plugins.training_schema(trainer)
trainers.append({
"id": trainer,
"tool_id": tool.id,
"name": tool.name,
"requires_confirmation": tool.requires_confirmation,
"settings": schema,
"arguments": [arg for arg in tool.arguments if arg not in {"dataset_dir", "output_dir"}],
"required_arguments": [arg for arg in tool.required_arguments if arg not in {"dataset_dir", "output_dir"}],
})
return {
"trainers": trainers,
"datasets": self.datasets(),
"base_models": [item for item in self.models() if item["kind"] == "base_model"],
"presets": self.training_presets(),
}
def training_presets(self) -> list[dict[str, Any]]:
presets = [
{
"id": "standard_ddpm",
"name": "Standard DDPM",
"trainer": "ddpm",
"epochs": 100,
"settings": {
"resolution": 128,
"batch_size": 1,
"learning_rate": 0.0001,
"save_every": 10,
"preview_enabled": True,
"preview_every": 5,
},
},
{
"id": "quick_ddpm_test",
"name": "Quick DDPM Test",
"trainer": "ddpm",
"epochs": 3,
"settings": {
"resolution": 64,
"batch_size": 1,
"learning_rate": 0.0001,
"save_every": 1,
"preview_enabled": True,
"preview_every": 1,
"training_intensity": 25,
},
},
{
"id": "oasis_training",
"name": "Oasis Training",
"trainer": "oasis",
"epochs": 25,
"settings": {
"resolution": "256x144",
"batch_size": 2,
"learning_rate": 0.00002,
"workers": 2,
"preview_enabled": True,
"preview_every": 5,
},
},
]
for recipe in self._studio_store().recipes:
presets.append({
"id": f"recipe_{recipe.id}",
"name": recipe.name,
"trainer": recipe.trainer,
"epochs": recipe.epochs,
"settings": {
"preview_prompt": recipe.preview_prompt,
**({"base_model": recipe.base_model} if recipe.base_model else {}),
},
"user": True,
})
available = {trainer["id"] for trainer in self.training_schema_no_presets()}
return [preset for preset in presets if preset["trainer"] in available]
def training_schema_no_presets(self) -> list[dict[str, Any]]:
registry = self._registry()
trainers = []
for tool in registry.enabled():
if not tool.id.endswith("_trainer"):
continue
trainer = tool.id.removesuffix("_trainer")
trainers.append({"id": trainer})
return trainers
def _build_training_plan(self, payload: dict[str, Any]) -> Any:
if self.planner is None:
raise RemoteApiError("Planning is not available in this ADAM session.", status=503)
trainer = bounded_text(payload.get("trainer", "lora"), max_length=64, label="Trainer", required=True).casefold()
dataset = self._asset_from_id(bounded_text(payload.get("dataset_id"), max_length=4000, label="Dataset", required=True), kind="dataset")
if not Path(dataset.path).is_dir():
raise RemoteApiError("That dataset is unavailable on the PC.", status=404)
model_name = bounded_text(payload.get("model_name") or dataset.name, max_length=96, label="Model name", required=True)
epochs = bounded_int(payload.get("epochs"), minimum=1, maximum=100_000, default=10, label="Epochs")
options = dict(payload.get("settings") or payload.get("training_options") or {})
if not isinstance(options, dict):
raise RemoteApiError("Training settings must be an object.")
trigger_word = bounded_text(payload.get("trigger_word") or options.get("trigger_word") or "", max_length=128, label="Trigger word")
if trigger_word:
options["trigger_word"] = trigger_word
base_model = ""
if trainer == "lora":
base_id = bounded_text(payload.get("base_model_id", ""), max_length=4000, label="Base model")
if base_id:
base_model = self._asset_from_id(base_id, kind="base_model").path
else:
base_model = str(options.get("base_model") or "")
if not base_model and hasattr(self.planner, "_lora_base_model"):
base_model = str(self.planner._lora_base_model())
if base_model:
options["base_model"] = base_model
output = ""
if hasattr(self.planner, "_training_output"):
output_path = self.planner._training_output(trainer, model_name)
output = str(output_path) if output_path else ""
command = TrainingCommand.from_dict({
"action": "train",
"trainer": trainer,
"dataset": dataset.path,
"model_name": model_name,
"epochs": epochs,
"output": output,
"base_model": base_model,
"trigger_word": trigger_word,
"training_options": options,
})
if not hasattr(self.planner, "_plan_training_command"):
raise RemoteApiError("Structured training validation is not available.", status=503)
plan = self.planner._plan_training_command("Remote structured training", command)
append_preflight_summary(plan, self.config)
return plan
def training_plan(self, payload: dict[str, Any]) -> dict[str, Any]:
plan = self._build_training_plan(payload)
return self._plan_payload(plan)
def start_training(self, payload: dict[str, Any]) -> dict[str, Any]:
if self.jobs is None:
raise RemoteApiError("Jobs are not available in this ADAM session.", status=503)
plan = self._build_training_plan(payload)
if not plan.steps:
raise RemoteApiError(plan.summary or "ADAM could not build a runnable training plan.")
job = self.dispatcher.submit_job(self.jobs, plan)
auto_approved = False
if self.auto_approve_training(plan):
self.dispatcher.confirm_job(self.jobs, job.id)
auto_approved = True
try:
dataset = self._asset_from_id(bounded_text(payload.get("dataset_id"), max_length=4000, label="Dataset"), kind="dataset")
self._dataset_registry().touch(dataset.path)
except Exception:
pass
return {
"message": f"Queued {job.plan.project_name}.",
"job_id": job.id,
"requires_approval": bool(plan.requires_confirmation and not auto_approved),
"auto_approved": auto_approved,
"plan": self._plan_payload(plan),
}
def generation_schema(self) -> dict[str, Any]:
registry = self._registry()
providers = []
for tool in generation_tools(registry):
providers.append({
"id": tool.id,
"name": tool.name,
"model_trainers": tool.model_trainers,
"requires_confirmation": tool.requires_confirmation,
"options": tool.generation_options,
"settings": registry.model_plugins.generation_schema_for_tool(tool.id),
})
models = self.models()
return {
"providers": providers,
"models": [item for item in models if item["kind"] == "model"],
"base_models": [item for item in models if item["kind"] == "base_model"],
}
def start_generation(self, payload: dict[str, Any]) -> dict[str, Any]:
if self.jobs is None:
raise RemoteApiError("Jobs are not available in this ADAM session.", status=503)
registry = self._registry()
provider_id = bounded_text(payload.get("provider_id", ""), max_length=96, label="Provider", required=True)
tool = next((item for item in generation_tools(registry) if item.id == provider_id), None)
if tool is None:
raise RemoteApiError("Unknown generation provider.", status=404)
model = self._asset_from_id(bounded_text(payload.get("model_id", ""), max_length=4000, label="Model", required=True), kind="model")
if model.trainer not in tool.model_trainers:
raise RemoteApiError("That model is not compatible with the selected provider.")
prompt = bounded_text(payload.get("prompt", ""), max_length=2000, label="Prompt")
count = bounded_int(payload.get("image_count"), minimum=1, maximum=32, default=1, label="Image count")
if provider_id == "lora_generator":
count = min(count, 8)
options = tool.generation_options
step_max = int(options.get("step_max", 500) or 500)
steps = bounded_int(payload.get("steps"), minimum=1, maximum=step_max, default=int(options.get("step_default", 30) or 30), label="Steps")
seed = bounded_int(payload.get("seed"), minimum=0, maximum=2_147_483_647, default=0, label="Seed")
sampler_options = [str(item) for item in options.get("samplers", [])]
sampler = bounded_text(payload.get("sampler") or (sampler_options[0] if sampler_options else "DDIM"), max_length=64, label="Sampler")
if sampler_options and sampler not in sampler_options:
raise RemoteApiError("Sampler is not supported by that provider.")
aspect_options = [str(item) for item in options.get("aspect_ratios", [])]
aspect = bounded_text(payload.get("aspect_ratio") or (aspect_options[0] if aspect_options else "1:1 (Square)"), max_length=64, label="Aspect ratio")
if aspect_options and aspect not in aspect_options:
raise RemoteApiError("Aspect ratio is not supported by that provider.")
extra: dict[str, Any] = {}
if provider_id == "lora_generator":
base_id = bounded_text(payload.get("base_model_id", ""), max_length=4000, label="Base model")
base_model_path = self._asset_from_id(base_id, kind="base_model").path if base_id else ""
extra.update({
"negative_prompt": bounded_text(payload.get("negative_prompt", ""), max_length=3000, label="Negative prompt"),
"base_model_path": base_model_path,
"width": bounded_int(payload.get("width"), minimum=0, maximum=2048, default=0, label="Width"),
"height": bounded_int(payload.get("height"), minimum=0, maximum=2048, default=0, label="Height"),
"cfg_scale": bounded_float(payload.get("cfg_scale"), minimum=0.0, maximum=30.0, default=0.0, label="CFG scale"),
"lora_strength": bounded_float(payload.get("lora_strength"), minimum=0.0, maximum=3.0, default=1.0, label="LoRA strength"),
"denoise_strength": bounded_float(payload.get("denoise_strength"), minimum=0.0, maximum=1.0, default=0.0, label="Denoise strength"),
"prompt_weighting": bool(payload.get("prompt_weighting", True)),
})
elif provider_id == "ddpm_generator":
extra.update({
"reference_strength": bounded_int(payload.get("reference_strength"), minimum=0, maximum=100, default=65, label="Reference strength"),
"width": bounded_int(payload.get("width"), minimum=0, maximum=2048, default=0, label="Width"),
"height": bounded_int(payload.get("height"), minimum=0, maximum=2048, default=0, label="Height"),
})
plugin_settings = payload.get("settings") or {}
if isinstance(plugin_settings, dict):
extra.update(plugin_settings)
plan = build_generation_plan(
tool,
model_name=model.name,
model_path=model.path,
prompt=prompt,
image_count=count,
steps=steps,
seed=seed,
sampler=sampler,
aspect_ratio=aspect,
extra_arguments=extra,
)
job = self.dispatcher.submit_job(self.jobs, plan)
return {
"message": f"Queued {job.plan.project_name}.",
"job_id": job.id,
"requires_approval": bool(plan.requires_confirmation),
"plan": self._plan_payload(plan),
}
def job_detail(self, job_id: str) -> dict[str, Any]:
if self.jobs is None or not hasattr(self.jobs, "get"):
raise RemoteApiError("Jobs are not available in this ADAM session.", status=503)
job = self.jobs.get(job_id)
if job is None:
raise RemoteApiError("ADAM could not find that job.", status=404)
return {
"id": job.id,
"project": job.plan.project_name,
"status": job.status.value,
"progress": job.progress,
"current_step": job.current_step,
"logs": list(job.logs)[-40:],
"steps": [
{
"tool_id": step.tool_id,
"title": step.title,
"description": step.description,
"status": step.status.value,
"arguments": sanitized_arguments(dict(step.arguments)),
}
for step in job.plan.steps
],
}
@staticmethod
def _plan_payload(plan: Any) -> dict[str, Any]:
return {
"id": getattr(plan, "id", ""),
"summary": getattr(plan, "summary", ""),
"project_name": getattr(plan, "project_name", ""),
"requires_confirmation": bool(getattr(plan, "requires_confirmation", False)),
"confirmation_reason": getattr(plan, "confirmation_reason", ""),
"steps": [
{
"tool_id": step.tool_id,
"title": step.title,
"description": step.description,
"arguments": sanitized_arguments(dict(step.arguments)),
}
for step in getattr(plan, "steps", [])
],
}
|