Spaces:
Paused
Paused
File size: 7,301 Bytes
99a7ebb | 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 | from __future__ import annotations
import time
import uuid
from typing import Any, Iterable, Iterator
from fastapi import HTTPException
from services.protocol.conversation import (
ConversationRequest,
ImageOutput,
collect_image_outputs,
collect_text,
count_message_tokens,
count_text_tokens,
encode_images,
normalize_messages,
stream_image_outputs_with_pool,
stream_text_deltas,
text_backend,
)
from utils.helper import build_chat_image_markdown_content, extract_chat_image, extract_chat_prompt, is_image_chat_request, parse_image_count
def completion_chunk(model: str, delta: dict[str, Any], finish_reason: str | None = None, completion_id: str = "", created: int | None = None) -> dict[str, Any]:
return {
"id": completion_id or f"chatcmpl-{uuid.uuid4().hex}",
"object": "chat.completion.chunk",
"created": created or int(time.time()),
"model": model,
"choices": [{"index": 0, "delta": delta, "finish_reason": finish_reason}],
}
def completion_response(
model: str,
content: str,
created: int | None = None,
messages: list[dict[str, Any]] | None = None,
) -> dict[str, Any]:
prompt_tokens = count_message_tokens(messages, model) if messages else 0
completion_tokens = count_text_tokens(content, model) if messages else 0
return {
"id": f"chatcmpl-{uuid.uuid4().hex}",
"object": "chat.completion",
"created": created or int(time.time()),
"model": model,
"choices": [{
"index": 0,
"message": {"role": "assistant", "content": content},
"finish_reason": "stop",
}],
"usage": {
"prompt_tokens": prompt_tokens,
"completion_tokens": completion_tokens,
"total_tokens": prompt_tokens + completion_tokens,
},
}
def stream_text_chat_completion(backend, messages: list[dict[str, Any]], model: str) -> Iterator[dict[str, Any]]:
completion_id = f"chatcmpl-{uuid.uuid4().hex}"
created = int(time.time())
sent_role = False
request = ConversationRequest(model=model, messages=messages)
for delta_text in stream_text_deltas(backend, request):
if not sent_role:
sent_role = True
yield completion_chunk(model, {"role": "assistant", "content": delta_text}, None, completion_id, created)
else:
yield completion_chunk(model, {"content": delta_text}, None, completion_id, created)
if not sent_role:
yield completion_chunk(model, {"role": "assistant", "content": ""}, None, completion_id, created)
yield completion_chunk(model, {}, "stop", completion_id, created)
def collect_chat_content(chunks: Iterable[dict[str, Any]]) -> str:
parts: list[str] = []
for chunk in chunks:
choices = chunk.get("choices")
first = choices[0] if isinstance(choices, list) and choices and isinstance(choices[0], dict) else {}
delta = first.get("delta") if isinstance(first.get("delta"), dict) else {}
content = str(delta.get("content") or "")
if content:
parts.append(content)
return "".join(parts)
def chat_messages_from_body(body: dict[str, Any]) -> list[dict[str, Any]]:
messages = body.get("messages")
if isinstance(messages, list) and messages:
return [message for message in messages if isinstance(message, dict)]
prompt = str(body.get("prompt") or "").strip()
if prompt:
return [{"role": "user", "content": prompt}]
raise HTTPException(status_code=400, detail={"error": "messages or prompt is required"})
def chat_image_args(body: dict[str, Any]) -> tuple[str, str, int, list[tuple[bytes, str, str]]]:
model = str(body.get("model") or "gpt-image-2").strip() or "gpt-image-2"
prompt = extract_chat_prompt(body)
if not prompt:
raise HTTPException(status_code=400, detail={"error": "prompt is required"})
images = [
(data, f"image_{idx}.png", mime)
for idx, (data, mime) in enumerate(extract_chat_image(body), start=1)
]
return model, prompt, parse_image_count(body.get("n")), images
def text_chat_parts(body: dict[str, Any]) -> tuple[str, list[dict[str, Any]]]:
model = str(body.get("model") or "auto").strip() or "auto"
messages = normalize_messages(chat_messages_from_body(body))
return model, messages
def image_result_content(result: dict[str, Any]) -> str:
data = result.get("data")
if isinstance(data, list) and data:
return build_chat_image_markdown_content(result)
return str(result.get("message") or "Image generation completed.")
def image_chat_response(body: dict[str, Any]) -> dict[str, Any]:
model, prompt, n, images = chat_image_args(body)
result = collect_image_outputs(stream_image_outputs_with_pool(ConversationRequest(
prompt=prompt,
model=model,
n=n,
response_format="b64_json",
images=encode_images(images) or None,
)))
return completion_response(model, image_result_content(result), int(result.get("created") or 0) or None)
def image_chat_events(body: dict[str, Any]) -> Iterator[dict[str, Any]]:
model, prompt, n, images = chat_image_args(body)
image_outputs = stream_image_outputs_with_pool(ConversationRequest(
prompt=prompt,
model=model,
n=n,
response_format="b64_json",
images=encode_images(images) or None,
))
yield from stream_image_chat_completion(image_outputs, model)
def stream_image_chat_completion(image_outputs: Iterable[ImageOutput], model: str) -> Iterator[dict[str, Any]]:
completion_id = f"chatcmpl-{uuid.uuid4().hex}"
created = int(time.time())
sent_role = False
sent_text = ""
for output in image_outputs:
content = ""
if output.kind == "progress":
content = output.text
sent_text += content
elif output.kind == "result":
content = build_chat_image_markdown_content({"data": output.data})
elif output.kind == "message":
content = output.text[len(sent_text):] if output.text.startswith(sent_text) else output.text
if not content:
continue
if not sent_role:
sent_role = True
yield completion_chunk(model, {"role": "assistant", "content": content}, None, completion_id, created)
else:
yield completion_chunk(model, {"content": content}, None, completion_id, created)
if not sent_role:
yield completion_chunk(model, {"role": "assistant", "content": ""}, None, completion_id, created)
yield completion_chunk(model, {}, "stop", completion_id, created)
def handle(body: dict[str, Any]) -> dict[str, Any] | Iterator[dict[str, Any]]:
if body.get("stream"):
if is_image_chat_request(body):
return image_chat_events(body)
model, messages = text_chat_parts(body)
return stream_text_chat_completion(text_backend(), messages, model)
if is_image_chat_request(body):
return image_chat_response(body)
model, messages = text_chat_parts(body)
request = ConversationRequest(model=model, messages=messages)
return completion_response(model, collect_text(text_backend(), request), messages=messages)
|