File size: 9,729 Bytes
b3d711f | 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 | """Utility functions for managing collection statistics.
This module provides standalone functions for enabling, disabling, and retrieving
statistics for ChromaDB collections. These functions work with the attached function
system to automatically compute metadata value frequencies.
Example:
>>> from chromadb.utils.statistics import attach_statistics_function, get_statistics
>>> import chromadb
>>>
>>> client = chromadb.Client()
>>> collection = client.get_or_create_collection("my_collection")
>>>
>>> # Attach statistics function with output collection name
>>> attach_statistics_function(collection, "my_collection_statistics")
>>>
>>> # Add some data
>>> collection.add(
... ids=["id1", "id2"],
... documents=["doc1", "doc2"],
... metadatas=[{"category": "A"}, {"category": "B"}]
... )
>>>
>>> # Get statistics from the named output collection
>>> stats = get_statistics(collection, "my_collection_statistics")
>>> print(stats)
"""
from typing import TYPE_CHECKING, Optional, Dict, Any, cast, Tuple
from collections import defaultdict
from chromadb.api.types import OneOrMany, Where, maybe_cast_one_to_many
from chromadb.api.functions import STATISTICS_FUNCTION
if TYPE_CHECKING:
from chromadb.api.models.Collection import Collection
from chromadb.api.models.AttachedFunction import AttachedFunction
def get_statistics_fn_name(collection: "Collection") -> str:
"""Generate the default name for the statistics attached function.
Args:
collection: The collection to generate the name for
Returns:
str: The statistics function name
"""
return f"{collection.name}_stats"
def attach_statistics_function(
collection: "Collection", stats_collection_name: str
) -> Tuple["AttachedFunction", bool]:
"""Attach statistics collection function to a collection.
This attaches the statistics function which will automatically compute
and update metadata value frequencies whenever records are added, updated,
or deleted.
Args:
collection: The collection to enable statistics for
stats_collection_name: Name of the collection where statistics will be stored.
Returns:
Tuple of (AttachedFunction, created) where created is True if newly created,
False if already existed (idempotent request)
Example:
>>> attached_fn, created = attach_statistics_function(collection, "my_collection_statistics")
>>> if created:
... print("Statistics function newly attached")
>>> collection.add(ids=["id1"], documents=["doc1"], metadatas=[{"key": "value"}])
>>> # Statistics are automatically computed
>>> stats = get_statistics(collection, "my_collection_statistics")
"""
return collection.attach_function(
function=STATISTICS_FUNCTION,
name=get_statistics_fn_name(collection),
output_collection=stats_collection_name,
params=None,
)
def get_statistics_fn(collection: "Collection") -> "AttachedFunction":
"""Get the statistics attached function for a collection.
Args:
collection: The collection to get the statistics function for
Returns:
AttachedFunction: The statistics function
Raises:
NotFoundError: If statistics are not enabled
AssertionError: If the attached function is not a statistics function
"""
af = collection.get_attached_function(get_statistics_fn_name(collection))
assert (
af.function_name == "statistics"
), "Attached function is not a statistics function"
return af
def detach_statistics_function(
collection: "Collection", delete_stats_collection: bool = False
) -> bool:
"""Detach statistics collection function from a collection.
Args:
collection: The collection to disable statistics for
delete_stats_collection: If True, also delete the statistics output collection.
Defaults to False.
Returns:
bool: True if successful
Example:
>>> detach_statistics_function(collection, delete_stats_collection=True)
"""
attached_fn = get_statistics_fn(collection)
return collection.detach_function(
attached_fn.name, delete_output_collection=delete_stats_collection
)
def get_statistics(
collection: "Collection",
stats_collection_name: str,
keys: Optional[OneOrMany[str]] = None,
) -> Dict[str, Any]:
"""Get the current statistics for a collection.
Statistics include frequency counts for all metadata key-value pairs,
as well as a summary with the total record count.
Args:
collection: The collection to get statistics for
stats_collection_name: Name of the statistics collection to read from.
keys: Optional metadata key(s) to filter statistics for. Can be a single key
string or a list of keys. If provided, only returns statistics for
those specific keys.
Returns:
Dict[str, Any]: A dictionary with the structure:
{
"statistics": {
"key1": {
"value1": {"count": count, ...},
"value2": {"count": count, ...}
},
"key2": {...},
...
},
"summary": {
"total_count": count
}
}
Example:
>>> attach_statistics_function(collection, "my_collection_statistics")
>>> collection.add(
... ids=["id1", "id2"],
... documents=["doc1", "doc2"],
... metadatas=[{"category": "A", "score": 10}, {"category": "B", "score": 10}]
... )
>>> # Wait for statistics to be computed
>>> stats = get_statistics(collection, "my_collection_statistics")
>>> print(stats)
{
"statistics": {
"category": {
"A": {"count": 1},
"B": {"count": 1}
},
"score": {
"10": {"count": 2}
}
},
"summary": {
"total_count": 2
}
}
Raises:
ValueError: If more than 30 keys are provided in the keys filter.
"""
# Normalize keys to list
keys_list = maybe_cast_one_to_many(keys)
# Validate keys count to avoid issues with large $in queries
MAX_KEYS = 30
if keys_list is not None and len(keys_list) > MAX_KEYS:
raise ValueError(
f"Too many keys provided: {len(keys_list)}. "
f"Maximum allowed is {MAX_KEYS} keys per request. "
"Consider calling get_statistics multiple times with smaller key batches."
)
# Import here to avoid circular dependency
from chromadb.api.models.Collection import Collection
# Get the statistics output collection model from the server
stats_collection_model = collection._client.get_collection(
name=stats_collection_name,
tenant=collection.tenant,
database=collection.database,
)
# Wrap it in a Collection object to access get/query methods
stats_collection = Collection(
client=collection._client,
model=stats_collection_model,
embedding_function=None, # Statistics collections don't need embedding functions
data_loader=None,
)
# Get all statistics records by paginating through the stats collection
stats: Dict[str, Dict[str, Dict[str, int]]] = defaultdict(lambda: defaultdict(dict))
summary: Dict[str, Any] = {}
offset = 0
# When filtering by keys, also include "summary" entries to get total_count
where_filter: Optional[Where] = (
cast(Where, {"key": {"$in": keys_list + ["summary"]}})
if keys_list is not None
else None
)
while True:
page = stats_collection.get(
include=["metadatas"], offset=offset, where=where_filter
)
metadatas = page.get("metadatas") or []
if not metadatas:
break
for metadata in metadatas:
if metadata is None:
continue
meta_key = metadata.get("key")
value = metadata.get("value")
value_label = metadata.get("value_label")
value_type = metadata.get("type")
count = metadata.get("count")
if (
meta_key is not None
and value is not None
and value_type is not None
and count is not None
):
if meta_key == "summary":
if value == "total_count":
summary["total_count"] = count
else:
# Prioritize value_label if present, otherwise use value
stats_key = value_label if value_label is not None else value
assert isinstance(meta_key, str)
assert isinstance(stats_key, str)
assert isinstance(count, int)
stats[meta_key][stats_key]["count"] = count
# Advance to next page using the actual number of items returned
offset += len(metadatas)
result = {"statistics": dict(stats)}
if summary:
result["summary"] = summary
return result
|