Spaces:
Runtime error
Runtime error
File size: 8,834 Bytes
e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 55dc3dd e4f9cbe c14732f e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 55dc3dd 168aee7 55dc3dd 168aee7 55dc3dd 168aee7 55dc3dd e4f9cbe 55dc3dd e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 2226ee3 e4f9cbe cc5eabb e4f9cbe 55dc3dd e4f9cbe 55dc3dd 168aee7 55dc3dd 168aee7 55dc3dd 168aee7 3600417 168aee7 3600417 55dc3dd 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 c14732f 55dc3dd e4f9cbe 55dc3dd e4f9cbe 55dc3dd 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 168aee7 e4f9cbe 55dc3dd e4f9cbe 815971e c14732f e4f9cbe cc5eabb e4f9cbe cc5eabb e4f9cbe |
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 |
"""Router for the concept database."""
from typing import Annotated, Optional
from fastapi import APIRouter, HTTPException
from fastapi.params import Depends
from openai_function_call import OpenAISchema
from pydantic import BaseModel, Field
from .auth import UserInfo, get_session_user
from .concepts.concept import (
DRAFT_MAIN,
Concept,
ConceptColumnInfo,
ConceptMetrics,
DraftId,
draft_examples,
)
from .concepts.db_concept import DISK_CONCEPT_DB, DISK_CONCEPT_MODEL_DB, ConceptInfo, ConceptUpdate
from .config import env
from .router_utils import RouteErrorHandler
from .schema import SignalInputType
router = APIRouter(route_class=RouteErrorHandler)
@router.get('/', response_model_exclude_none=True)
def get_concepts(
user: Annotated[Optional[UserInfo], Depends(get_session_user)]) -> list[ConceptInfo]:
"""List the concepts."""
return DISK_CONCEPT_DB.list(user)
@router.get('/{namespace}/{concept_name}', response_model_exclude_none=True)
def get_concept(namespace: str,
concept_name: str,
draft: Optional[DraftId] = DRAFT_MAIN,
user: Annotated[Optional[UserInfo], Depends(get_session_user)] = None) -> Concept:
"""Get a concept from a database."""
concept = DISK_CONCEPT_DB.get(namespace, concept_name, user)
if not concept:
raise HTTPException(
status_code=404,
detail=f'Concept "{namespace}/{concept_name}" was not found or user does not have access.')
# Only return the examples from the draft.
concept.data = draft_examples(concept, draft or DRAFT_MAIN)
return concept
@router.get('/{namespace}/{concept_name}/column_infos')
def get_concept_column_infos(
namespace: str, concept_name: str,
user: Annotated[Optional[UserInfo], Depends(get_session_user)]) -> list[ConceptColumnInfo]:
"""Return a list of dataset columns where this concept was applied to."""
concept = DISK_CONCEPT_DB.get(namespace, concept_name, user)
if not concept:
raise HTTPException(
status_code=404, detail=f'Concept "{namespace}/{concept_name}" was not found.')
return DISK_CONCEPT_MODEL_DB.get_column_infos(namespace, concept_name)
class CreateConceptOptions(BaseModel):
"""Options for creating a concept."""
# Namespace of the concept.
namespace: str
# Name of the concept.
name: str
# Input type (modality) of the concept.
type: SignalInputType
description: Optional[str] = None
@router.post('/create', response_model_exclude_none=True)
def create_concept(options: CreateConceptOptions,
user: Annotated[Optional[UserInfo],
Depends(get_session_user)]) -> Concept:
"""Edit a concept in the database."""
return DISK_CONCEPT_DB.create(options.namespace, options.name, options.type, options.description,
user)
@router.post('/{namespace}/{concept_name}', response_model_exclude_none=True)
def edit_concept(namespace: str, concept_name: str, change: ConceptUpdate,
user: Annotated[Optional[UserInfo], Depends(get_session_user)]) -> Concept:
"""Edit a concept in the database."""
return DISK_CONCEPT_DB.edit(namespace, concept_name, change, user)
@router.delete('/{namespace}/{concept_name}')
def delete_concept(namespace: str, concept_name: str,
user: Annotated[Optional[UserInfo],
Depends(get_session_user)]) -> None:
"""Deletes the concept from the database."""
DISK_CONCEPT_DB.remove(namespace, concept_name, user)
# Delete concept models from all datasets that are using this concept.
DISK_CONCEPT_MODEL_DB.remove_all(namespace, concept_name)
class MergeConceptDraftOptions(BaseModel):
"""Merge a draft into main."""
draft: DraftId
@router.post('/{namespace}/{concept_name}/merge_draft', response_model_exclude_none=True)
def merge_concept_draft(namespace: str, concept_name: str, options: MergeConceptDraftOptions,
user: Annotated[Optional[UserInfo],
Depends(get_session_user)]) -> Concept:
"""Merge a draft in the concept into main."""
return DISK_CONCEPT_DB.merge_draft(namespace, concept_name, options.draft, user)
class ScoreExample(BaseModel):
"""Example to score along a specific concept."""
text: Optional[str] = None
img: Optional[bytes] = None
class ScoreBody(BaseModel):
"""Request body for the score endpoint."""
examples: list[ScoreExample]
draft: str = DRAFT_MAIN
class ScoreResponse(BaseModel):
"""Response body for the score endpoint."""
scores: list[dict]
model_synced: bool
class ConceptModelInfo(BaseModel):
"""Information about a concept model."""
namespace: str
concept_name: str
embedding_name: str
version: int
column_info: Optional[ConceptColumnInfo] = None
metrics: Optional[ConceptMetrics] = None
@router.get('/{namespace}/{concept_name}/model')
def get_concept_models(
namespace: str,
concept_name: str,
user: Annotated[Optional[UserInfo],
Depends(get_session_user)] = None) -> list[ConceptModelInfo]:
"""Get a concept model from a database."""
concept = DISK_CONCEPT_DB.get(namespace, concept_name, user)
if not concept:
raise HTTPException(
status_code=404, detail=f'Concept "{namespace}/{concept_name}" was not found')
models = DISK_CONCEPT_MODEL_DB.get_models(namespace, concept_name, user)
for m in models:
DISK_CONCEPT_MODEL_DB.sync(m, user)
return [
ConceptModelInfo(
namespace=m.namespace,
concept_name=m.concept_name,
embedding_name=m.embedding_name,
version=m.version,
column_info=m.column_info,
metrics=m.get_metrics(concept)) for m in models
]
@router.get('/{namespace}/{concept_name}/model/{embedding_name}')
def get_concept_model(
namespace: str,
concept_name: str,
embedding_name: str,
user: Annotated[Optional[UserInfo], Depends(get_session_user)] = None) -> ConceptModelInfo:
"""Get a concept model from a database."""
concept = DISK_CONCEPT_DB.get(namespace, concept_name, user)
if not concept:
raise HTTPException(
status_code=404, detail=f'Concept "{namespace}/{concept_name}" was not found')
model = DISK_CONCEPT_MODEL_DB.get(namespace, concept_name, embedding_name, user=user)
if not model:
model = DISK_CONCEPT_MODEL_DB.create(namespace, concept_name, embedding_name, user=user)
DISK_CONCEPT_MODEL_DB.sync(model)
model_info = ConceptModelInfo(
namespace=model.namespace,
concept_name=model.concept_name,
embedding_name=model.embedding_name,
version=model.version,
column_info=model.column_info,
metrics=model.get_metrics(concept))
return model_info
class MetricsBody(BaseModel):
"""Request body for the compute_metrics endpoint."""
column_info: Optional[ConceptColumnInfo] = None
@router.post(
'/{namespace}/{concept_name}/model/{embedding_name}/score', response_model_exclude_none=True)
def score(namespace: str, concept_name: str, embedding_name: str, body: ScoreBody,
user: Annotated[Optional[UserInfo], Depends(get_session_user)]) -> ScoreResponse:
"""Score examples along the specified concept."""
concept = DISK_CONCEPT_DB.get(namespace, concept_name, user)
if not concept:
raise HTTPException(
status_code=404, detail=f'Concept "{namespace}/{concept_name}" was not found')
model = DISK_CONCEPT_MODEL_DB.get(namespace, concept_name, embedding_name, user=user)
if model is None:
model = DISK_CONCEPT_MODEL_DB.create(namespace, concept_name, embedding_name, user=user)
model_updated = DISK_CONCEPT_MODEL_DB.sync(model, user)
# TODO(smilkov): Support images.
texts = [example.text or '' for example in body.examples]
return ScoreResponse(scores=model.score(body.draft, texts), model_synced=model_updated)
class Examples(OpenAISchema):
"""Generated text examples."""
examples: list[str] = Field(..., description='List of generated examples')
@router.get('/generate_examples')
def generate_examples(description: str) -> list[str]:
"""Generate positive examples for a given concept using an LLM model."""
try:
import openai
except ImportError:
raise ImportError('Could not import the "openai" python package. '
'Please install it with `pip install openai`.')
openai.api_key = env('OPENAI_API_KEY')
completion = openai.ChatCompletion.create(
model='gpt-3.5-turbo-0613',
functions=[Examples.openai_schema],
messages=[
{
'role': 'system',
'content': 'You must call the `Examples` function with the generated examples',
},
{
'role': 'user',
'content': f'Write 5 diverse, unnumbered, and concise examples of "{description}"',
},
],
)
result = Examples.from_response(completion)
return result.examples
|