Spaces:
Running
Running
File size: 8,592 Bytes
23680f2 |
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 |
"""LanceDB schema definitions for HyperView.
Storage architecture:
- samples: Core sample metadata (no embeddings)
- metadata: Key-value pairs for dataset config
- spaces: Registry of embedding spaces
- embeddings__<space_key>: One table per embedding space (id + vector)
- layouts__<layout_key>: One table per layout (id + x + y)
"""
import json
import re
from dataclasses import dataclass
from typing import Any
import pyarrow as pa
from hyperview.core.sample import Sample
def create_sample_schema() -> pa.Schema:
"""Create the PyArrow schema for samples.
Samples are pure metadata - embeddings and layouts are stored separately.
"""
return pa.schema(
[
pa.field("id", pa.utf8(), nullable=False),
pa.field("filepath", pa.utf8(), nullable=False),
pa.field("label", pa.utf8(), nullable=True),
pa.field("metadata_json", pa.utf8(), nullable=True),
pa.field("thumbnail_base64", pa.utf8(), nullable=True),
]
)
def create_metadata_schema() -> pa.Schema:
"""Create the PyArrow schema for dataset metadata (key-value store)."""
return pa.schema(
[
pa.field("key", pa.utf8(), nullable=False),
pa.field("value", pa.utf8(), nullable=True),
]
)
def create_spaces_schema() -> pa.Schema:
"""Create the PyArrow schema for the spaces registry.
Each row represents an embedding space (one per model).
"""
return pa.schema(
[
pa.field("space_key", pa.utf8(), nullable=False),
pa.field("model_id", pa.utf8(), nullable=False),
pa.field("dim", pa.int32(), nullable=False),
pa.field("count", pa.int64(), nullable=False),
pa.field("created_at", pa.int64(), nullable=False),
pa.field("updated_at", pa.int64(), nullable=False),
pa.field("config_json", pa.utf8(), nullable=True),
]
)
def create_embeddings_schema(dim: int) -> pa.Schema:
"""Create the PyArrow schema for an embeddings table.
Args:
dim: Vector dimension for this embedding space.
"""
return pa.schema(
[
pa.field("id", pa.utf8(), nullable=False),
pa.field("vector", pa.list_(pa.float32(), dim), nullable=False),
]
)
def create_layouts_schema() -> pa.Schema:
"""Create the PyArrow schema for a layouts table.
Layouts store 2D coordinates for visualization.
"""
return pa.schema(
[
pa.field("id", pa.utf8(), nullable=False),
pa.field("x", pa.float32(), nullable=False),
pa.field("y", pa.float32(), nullable=False),
]
)
@dataclass
class SpaceInfo:
"""Metadata for an embedding space."""
space_key: str
model_id: str
dim: int
count: int
created_at: int
updated_at: int
config: dict[str, Any] | None = None
@property
def provider(self) -> str:
return (self.config or {}).get("provider", "unknown")
@property
def geometry(self) -> str:
return (self.config or {}).get("geometry", "euclidean")
def to_dict(self) -> dict[str, Any]:
return {
"space_key": self.space_key,
"model_id": self.model_id,
"dim": self.dim,
"count": self.count,
"created_at": self.created_at,
"updated_at": self.updated_at,
"config_json": json.dumps(self.config) if self.config else None,
}
def to_api_dict(self) -> dict[str, Any]:
return {
"space_key": self.space_key,
"model_id": self.model_id,
"dim": self.dim,
"count": self.count,
"provider": self.provider,
"geometry": self.geometry,
"config": self.config,
}
@classmethod
def from_dict(cls, row: dict[str, Any]) -> "SpaceInfo":
config_json = row.get("config_json")
config = json.loads(config_json) if config_json else None
return cls(
space_key=row["space_key"],
model_id=row["model_id"],
dim=row["dim"],
count=row["count"],
created_at=row["created_at"],
updated_at=row["updated_at"],
config=config,
)
def create_layouts_registry_schema() -> pa.Schema:
"""Create the PyArrow schema for the layouts registry.
Each row represents a layout (2D projection of an embedding space).
"""
return pa.schema(
[
pa.field("layout_key", pa.utf8(), nullable=False),
pa.field("space_key", pa.utf8(), nullable=False),
pa.field("method", pa.utf8(), nullable=False),
pa.field("geometry", pa.utf8(), nullable=False),
pa.field("count", pa.int64(), nullable=False),
pa.field("created_at", pa.int64(), nullable=False),
pa.field("params_json", pa.utf8(), nullable=True),
]
)
@dataclass
class LayoutInfo:
"""Metadata for a layout (2D projection)."""
layout_key: str
space_key: str
method: str
geometry: str
count: int
created_at: int
params: dict[str, Any] | None = None
def to_dict(self) -> dict[str, Any]:
return {
"layout_key": self.layout_key,
"space_key": self.space_key,
"method": self.method,
"geometry": self.geometry,
"count": self.count,
"created_at": self.created_at,
"params_json": json.dumps(self.params) if self.params else None,
}
def to_api_dict(self) -> dict[str, Any]:
return {
"layout_key": self.layout_key,
"space_key": self.space_key,
"method": self.method,
"geometry": self.geometry,
"count": self.count,
"params": self.params,
}
@classmethod
def from_dict(cls, row: dict[str, Any]) -> "LayoutInfo":
params_json = row.get("params_json")
params = json.loads(params_json) if params_json else None
return cls(
layout_key=row["layout_key"],
space_key=row["space_key"],
method=row["method"],
geometry=row["geometry"],
count=row["count"],
created_at=row["created_at"],
params=params,
)
def slugify_model_id(model_id: str) -> str:
"""Convert a model ID to a safe table name component.
Examples:
"openai/clip-vit-base-patch32" -> "openai_clip-vit-base-patch32"
"sentence-transformers/all-MiniLM-L6-v2" -> "sentence-transformers_all-MiniLM-L6-v2"
"""
# Replace / with _
slug = model_id.replace("/", "_")
# Replace any other unsafe characters with _
slug = re.sub(r"[^a-zA-Z0-9_\-]", "_", slug)
# Collapse multiple underscores
slug = re.sub(r"_+", "_", slug)
return slug.strip("_")
def make_space_key(model_id: str) -> str:
"""Generate a space_key from a model_id.
For simplicity, this is just the slugified model_id.
For provider-aware keys, use `make_provider_aware_space_key` from providers.py.
"""
return slugify_model_id(model_id)
def make_layout_key(
space_key: str,
method: str = "umap",
geometry: str = "euclidean",
params: dict | None = None,
) -> str:
"""Generate a layout_key from space, method, geometry, and params.
The params are hashed to ensure different parameter sets get different keys.
"""
base = f"{space_key}__{geometry}_{method}"
if params:
# Create a stable hash of params
import hashlib
params_str = "_".join(f"{k}={v}" for k, v in sorted(params.items()))
params_hash = hashlib.md5(params_str.encode()).hexdigest()[:8]
return f"{base}_{params_hash}"
return base
def sample_to_dict(sample: Sample) -> dict[str, Any]:
"""Convert a Sample to a dictionary for LanceDB insertion."""
return {
"id": sample.id,
"filepath": sample.filepath,
"label": sample.label,
"metadata_json": json.dumps(sample.metadata) if sample.metadata else None,
"thumbnail_base64": sample.thumbnail_base64,
}
def dict_to_sample(row: dict[str, Any]) -> Sample:
"""Convert a LanceDB row to a Sample object."""
metadata_json = row.get("metadata_json")
metadata = json.loads(metadata_json) if metadata_json else {}
return Sample(
id=row["id"],
filepath=row["filepath"],
label=row.get("label"),
metadata=metadata,
thumbnail_base64=row.get("thumbnail_base64"),
)
|