Spaces:
Paused
Paused
File size: 13,395 Bytes
9792ea7 | 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 | # -*- coding: utf-8 -*-
"""The model response module."""
import base64
from dataclasses import dataclass, field
from datetime import datetime
from enum import StrEnum
from typing import Any, Literal, Self, List
from ._model_usage import ChatUsage
from .._utils._common import _generate_id
from .._utils._mixin import DictMixin
from ..message import (
TextBlock,
ToolCallBlock,
ThinkingBlock,
DataBlock,
Base64Source,
)
from ..types import JSONSerializableObject
class FinishedReason(StrEnum):
"""The finished reason of the model response."""
INTERRUPTED = "interrupted"
"""The model response is interrupted by the asyncio.CancelledError."""
COMPLETED = "completed"
"""The model response is completed."""
@dataclass
class ChatResponse(DictMixin):
"""The response of chat models."""
content: List[TextBlock | ToolCallBlock | ThinkingBlock | DataBlock]
"""The content of the chat response, which can include text blocks,
tool use blocks, or thinking blocks."""
is_last: bool
"""Whether this response is the last response, if `Ture`, the content will
be the complete response, otherwise the content is a partial response"""
id: str = field(default_factory=_generate_id)
"""The unique identifier."""
created_at: str = field(default_factory=lambda: datetime.now().isoformat())
"""When the response was created"""
type: Literal["chat_response"] = field(
default_factory=lambda: "chat_response",
)
"""The type of the response, which is always 'chat_response'."""
usage: ChatUsage | None = field(default_factory=lambda: None)
"""The usage information of the chat response, if available."""
finished_reason: FinishedReason = field(
default=FinishedReason.COMPLETED,
)
"""The finished reason of the chat response, available when `is_last`
is `True`."""
metadata: dict[str, JSONSerializableObject] = field(
default_factory=lambda: {},
)
"""The metadata of the chat response"""
def append_text(self, text: str, block_id: str | None = None) -> Self:
"""Append text to the current response."""
for block in self.content:
if isinstance(block, TextBlock) and (
block_id is None or block_id == block.id
):
block.text += text
return self
# Append a new block
assert isinstance(self.content, list)
self.content.append(
TextBlock(text=text, id=block_id or _generate_id()),
)
return self
def append_thinking(
self,
thinking: str,
block_id: str | None = None,
**extra_fields: Any,
) -> Self:
"""Append thinking to the current response.
Args:
thinking (`str`):
The thinking content to append.
block_id (`str | None`, defaults to `None`):
The id of the ``ThinkingBlock`` to accumulate into. If no
matching block exists in ``content``, a new one is
appended.
**extra_fields (`Any`):
Additional provider-specific fields to attach to the
``ThinkingBlock`` (e.g. Anthropic's ``signature``, OpenAI
Responses API's ``reasoning_item_id``). Only non-``None``
values are applied.
"""
for block in self.content:
if isinstance(block, ThinkingBlock) and (
block_id is None or block_id == block.id
):
block.thinking += thinking
for key, value in extra_fields.items():
if value is not None:
setattr(block, key, value)
return self
assert isinstance(self.content, list)
block = ThinkingBlock(thinking=thinking, id=block_id or _generate_id())
for key, value in extra_fields.items():
if value is not None:
setattr(block, key, value)
self.content.append(block)
return self
def append_tool_call(
self,
block_id: str,
name: str,
input: str, # pylint: disable=redefined-builtin
**extra_fields: Any,
) -> Self:
"""Append tool call to the current response by tool call block ID.
Args:
block_id (`str`):
The id of the ``ToolCallBlock`` to accumulate into. If no
matching block exists in ``content``, a new one is
appended.
name (`str`):
The name of the tool being called.
input (`str`):
The incremental JSON string arguments to append.
**extra_fields (`Any`):
Additional provider-specific fields to attach to the
``ToolCallBlock`` (e.g. OpenAI Responses API's
``call_id``). Only non-``None`` values are applied.
"""
for block in self.content:
if isinstance(block, ToolCallBlock) and block.id == block_id:
block.input += input
for key, value in extra_fields.items():
if value is not None:
setattr(block, key, value)
return self
block = ToolCallBlock(
id=block_id,
name=name,
input=input,
)
for key, value in extra_fields.items():
if value is not None:
setattr(block, key, value)
assert isinstance(self.content, list)
self.content.append(block)
return self
def append_data_block(
self,
block_id: str,
data: bytes,
media_type: str,
name: str | None = None,
) -> Self:
"""Append raw media bytes to the ``DataBlock`` with the given id.
The accumulated bytes are stored base64-encoded in
``DataBlock.source.data`` (via a :class:`Base64Source`). Callers must
pass the *incremental* raw media bytes only — the method takes care
of base64 (de)coding internally, so consumers never have to worry
about base64 padding (``=``) corrupting concatenation.
.. note::
Byte-level delta accumulation only has a well-defined semantics
for streaming media where chunks can be safely concatenated
(currently ``audio/*``). For non-streamable media types
(e.g. ``image/*``, ``video/*``) each ``DataBlock`` should be
treated as a complete file and passed atomically rather than
accumulated through this method.
Args:
block_id (`str`):
The id of the ``DataBlock`` to accumulate into. If no
matching block exists in ``content``, a new one is
appended.
data (`bytes`):
The incremental raw media bytes to append.
media_type (`str`):
The IANA media type of the raw bytes (e.g. ``audio/pcm``,
``audio/wav``). Used both to tag a newly-created block and
to guard against accidentally mixing bytes from different
media types under the same id.
name (`str | None`, defaults to `None`):
The optional ``name`` field used when a new ``DataBlock``
needs to be created. Ignored when an existing block with
``block_id`` is found.
Returns:
`Self`:
The current ``ChatResponse`` instance for chaining.
"""
for block in self.content:
if (
isinstance(block, DataBlock)
and block.id == block_id
and isinstance(block.source, Base64Source)
and block.source.media_type == media_type
):
old_bytes = (
base64.b64decode(block.source.data)
if block.source.data
else b""
)
block.source.data = base64.b64encode(
old_bytes + data,
).decode("ascii")
return self
self.content.append(
DataBlock(
id=block_id,
source=Base64Source(
data=base64.b64encode(data).decode("ascii"),
media_type=media_type,
),
name=name,
),
)
return self
def append_chat_response(self, chat_response: Self) -> Self:
"""Append chat response to the current response."""
# Append content
new_block_dict = {_.id: _ for _ in chat_response.content}
for block in self.content:
if block.id in new_block_dict:
delta_block = new_block_dict.pop(block.id)
# Append data according to the block type
if isinstance(block, ThinkingBlock):
block.thinking += delta_block.thinking
# Provider-specific extra fields (e.g. Anthropic's
# ``signature``, OpenAI Responses API's
# ``reasoning_item_id``) are carried on the delta
# ``ThinkingBlock`` via pydantic's ``extra="allow"``.
# Copy any non-``None`` extras onto the accumulator.
for key, value in (delta_block.model_extra or {}).items():
if value is not None:
setattr(block, key, value)
elif isinstance(block, TextBlock):
block.text += delta_block.text
elif isinstance(block, ToolCallBlock):
block.input += delta_block.input
# Provider-specific extras (e.g. OpenAI Responses
# API's ``call_id``) may be attached on the delta
# block via pydantic's ``extra="allow"``.
for key, value in (delta_block.model_extra or {}).items():
if value is not None:
setattr(block, key, value)
elif isinstance(block, DataBlock):
# Only ``audio/*`` is treated as a streamable delta:
# callers accumulate raw media bytes across chunks and
# the concatenated result is a well-defined stream.
# For non-audio media types (``image/*``, ``video/*``,
# ...) each ``DataBlock`` is a complete standalone
# asset — byte concatenation is meaningless — so we
# overwrite in place with the latest delta instead.
if not (
isinstance(block.source, Base64Source)
and isinstance(delta_block.source, Base64Source)
and block.source.media_type
== delta_block.source.media_type
):
# Source shape / media type mismatch: replace the
# whole block to avoid mixing incompatible data.
block.source = delta_block.source
elif block.source.media_type.startswith("audio/"):
old_bytes = (
base64.b64decode(block.source.data)
if block.source.data
else b""
)
delta_bytes = (
base64.b64decode(delta_block.source.data)
if delta_block.source.data
else b""
)
block.source.data = base64.b64encode(
old_bytes + delta_bytes,
).decode("ascii")
else:
# Non-streamable media: latest delta wins.
block.source.data = delta_block.source.data
if new_block_dict:
# Attach new blocks to the content.
self.content.extend(
block.model_copy(deep=True)
for block in new_block_dict.values()
)
# Override the chat usage
if chat_response.usage:
self.usage = chat_response.usage
return self
@dataclass
class StructuredResponse:
"""The structured response of chat models."""
content: dict
"""The structured output of the model."""
id: str = field(default_factory=_generate_id)
"""The unique identifier."""
created_at: str = field(default_factory=lambda: datetime.now().isoformat())
"""When the response was created"""
type: Literal["structured_response"] = field(
default_factory=lambda: "structured_response",
)
"""The type of the response, which is always 'structured_response'."""
usage: ChatUsage | None = field(default_factory=lambda: None)
"""The usage information of the chat response, if available."""
metadata: dict[str, JSONSerializableObject] = field(
default_factory=lambda: {},
)
"""The metadata of the chat response"""
finished_reason: FinishedReason = field(
default=FinishedReason.COMPLETED,
)
"""The finished reason of the structured response."""
|