File size: 11,521 Bytes
a4b70d9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import requests

from ..helper import filter_none, format_media_prompt
from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin, RaiseErrorMixin
from ...typing import Union, AsyncResult, Messages, MediaListType
from ...requests import StreamSession, StreamResponse, raise_for_status, sse_stream
from ...image import use_aspect_ratio
from ...image.copy_images import save_response_media
from ...providers.response import FinishReason, ToolCalls, Usage, ImageResponse, ProviderInfo, AudioResponse, Reasoning, JsonConversation
from ...tools.media import render_messages
from ...tools.run_tools import AuthManager
from ...errors import MissingAuthError
from ... import debug

class OpenaiTemplate(AsyncGeneratorProvider, ProviderModelMixin, RaiseErrorMixin):
    api_base = ""
    api_key = None
    api_endpoint = None
    supports_message_history = True
    supports_system_message = True
    default_model = ""
    fallback_models = []
    sort_models = True
    models_needs_auth = False
    use_model_names = False
    ssl = None
    add_user = True
    use_image_size = False
    max_tokens: int = None

    @classmethod
    def get_models(cls, api_key: str = None, api_base: str = None) -> list[str]:
        if not cls.models:
            try:
                headers = {}
                if api_base is None:
                    api_base = cls.api_base
                if api_key is None and cls.api_key is not None:
                    api_key = cls.api_key
                if not api_key:
                    api_key = AuthManager.load_api_key(cls)
                if cls.models_needs_auth and not api_key:
                    raise MissingAuthError('Add a "api_key"')
                if api_key is not None:
                    headers["authorization"] = f"Bearer {api_key}"
                response = requests.get(f"{api_base}/models", headers=headers, verify=cls.ssl)
                raise_for_status(response)
                data = response.json()
                data = data.get("data") if isinstance(data, dict) else data
                if (not cls.needs_auth or cls.models_needs_auth or api_key) and data:
                    cls.live += 1
                cls.image_models = [model.get("name") if cls.use_model_names else model.get("id", model.get("name")) for model in data if model.get("image") or model.get("type") == "image" or model.get("supports_images")]
                cls.vision_models = cls.vision_models.copy()
                cls.vision_models += [model.get("name") if cls.use_model_names else model.get("id", model.get("name")) for model in data if model.get("vision")]
                cls.models = [model.get("name") if cls.use_model_names else model.get("id", model.get("name")) for model in data]
                cls.models_count = {model.get("name") if cls.use_model_names else model.get("id", model.get("name")): len(model.get("providers", [])) for model in data if len(model.get("providers", [])) > 1}
                if cls.sort_models:
                    cls.models.sort()
            except MissingAuthError:
                raise
            except Exception as e:
                if cls.fallback_models:
                    debug.error(e)
                    return cls.fallback_models
                raise e
        return cls.models

    @classmethod
    async def create_async_generator(
        cls,
        model: str,
        messages: Messages,
        proxy: str = None,
        timeout: int = 120,
        conversation: JsonConversation = None,
        media: MediaListType = None,
        api_key: str = None,
        api_endpoint: str = None,
        api_base: str = None,
        temperature: float = None,
        max_tokens: int = None,
        top_p: float = None,
        stop: Union[str, list[str]] = None,
        stream: bool = None,
        prompt: str = None,
        user: str = None,
        headers: dict = None,
        impersonate: str = None,
        download_media: bool = True,
        extra_parameters: list[str] = ["tools", "parallel_tool_calls", "tool_choice", "reasoning_effort", "logit_bias", "modalities", "audio", "stream_options"],
        extra_body: dict = None,
        **kwargs
    ) -> AsyncResult:
        if api_key is None and cls.api_key is not None:
            api_key = cls.api_key
        if cls.needs_auth and api_key is None:
            raise MissingAuthError('Add a "api_key"')
        async with StreamSession(
            proxy=proxy,
            headers=cls.get_headers(stream, api_key, headers),
            timeout=timeout,
            impersonate=impersonate,
        ) as session:
            model = cls.get_model(model, api_key=api_key, api_base=api_base)
            if api_base is None:
                api_base = cls.api_base

            # Proxy for image generation feature
            if model and model in cls.image_models:
                prompt = format_media_prompt(messages, prompt)
                size = use_aspect_ratio({"width": kwargs.get("width"), "height": kwargs.get("height")}, kwargs.get("aspect_ratio", None))
                size = {"size": f"{size['width']}x{size['height']}", **size} if cls.use_image_size and "width" in size and "height" in size else size
                data = {"prompt": prompt, "model": model, **size}

                # Handle media if provided
                if media is not None:
                    data["image_url"] = next(iter([data for data, _ in media if data and isinstance(data, str) and data.startswith("http://") or data.startswith("https://")]), None)
                async with session.post(f"{api_base.rstrip('/')}/images/generations", json=data, ssl=cls.ssl) as response:
                    data = await response.json()
                    cls.raise_error(data, response.status)
                    model = data.get("model")
                    if model:
                        yield ProviderInfo(**cls.get_dict(), model=model)
                    await raise_for_status(response)
                    yield ImageResponse([image["url"] for image in data["data"]], prompt)
                return

            extra_parameters = {key: kwargs[key] for key in extra_parameters if key in kwargs}
            if extra_body is None:
                extra_body = {}
            data = filter_none(
                messages=list(render_messages(messages, media)),
                model=model,
                temperature=temperature,
                max_tokens=max_tokens if max_tokens is not None else cls.max_tokens,
                top_p=top_p,
                stop=stop,
                stream="audio" not in extra_parameters if stream is None else stream,
                user=user if cls.add_user else None,
                conversation=conversation.get_dict() if conversation else None,
                **extra_parameters,
                **extra_body
            )
            if api_endpoint is None:
                if api_base:
                    api_endpoint = f"{api_base.rstrip('/')}/chat/completions"
                if api_endpoint is None:
                    api_endpoint = cls.api_endpoint
            async with session.post(api_endpoint, json=data, ssl=cls.ssl) as response:
                async for chunk in read_response(response, stream, prompt, cls.get_dict(), download_media):
                    yield chunk

    @classmethod
    def get_headers(cls, stream: bool, api_key: str = None, headers: dict = None) -> dict:
        return {
            "Accept": "text/event-stream" if stream else "application/json",
            "Content-Type": "application/json",
            **(
                {"Authorization": f"Bearer {api_key}"}
                if api_key else {}
            ),
            **({} if headers is None else headers)
        }
    
async def read_response(response: StreamResponse, stream: bool, prompt: str, provider_info: dict, download_media: bool) -> AsyncResult:
    content_type = response.headers.get("content-type", "text/event-stream" if stream else "application/json")
    if content_type.startswith("application/json"):
        data = await response.json()
        OpenaiTemplate.raise_error(data, response.status)
        await raise_for_status(response)
        model = data.get("model")
        if model:
            yield ProviderInfo(**provider_info, model=model)
        if "usage" in data:
            yield Usage(**data["usage"])
        if "conversation" in data:
            yield JsonConversation.from_dict(data["conversation"])
        if "choices" in data:
            choice = next(iter(data["choices"]), None)
            message = choice.get("message", {})
            if choice and "content" in message and message["content"]:
                yield message["content"].strip()
            if "tool_calls" in message:
                yield ToolCalls(message["tool_calls"])
            if choice:
                reasoning_content = choice.get("delta", {}).get("reasoning_content", choice.get("delta", {}).get("reasoning"))
                if reasoning_content:
                    yield Reasoning(reasoning_content, status="")
            audio = message.get("audio", {})
            if "data" in audio:
                if download_media:
                    async for chunk in save_response_media(audio, prompt, [model]):
                        yield chunk
                else:
                    yield AudioResponse(f"data:audio/mpeg;base64,{audio['data']}", transcript=audio.get("transcript"))
            if choice and "finish_reason" in choice and choice["finish_reason"] is not None:
                yield FinishReason(choice["finish_reason"])
                return
    elif content_type.startswith("text/event-stream"):
        await raise_for_status(response)
        reasoning = False
        first = True
        model_returned = False
        async for data in sse_stream(response):
            OpenaiTemplate.raise_error(data)
            model = data.get("model")
            if not model_returned and model:
                yield ProviderInfo(**provider_info, model=model)
                model_returned = True
            choice = next(iter(data["choices"]), None)
            if choice:
                content = choice.get("delta", {}).get("content")
                if content:
                    if first:
                        content = content.lstrip()
                    if content:
                        first = False
                        if reasoning:
                            yield Reasoning(status="")
                            reasoning = False
                        yield content
                tool_calls = choice.get("delta", {}).get("tool_calls")
                if tool_calls:
                    yield ToolCalls(tool_calls)
                reasoning_content = choice.get("delta", {}).get("reasoning_content", choice.get("delta", {}).get("reasoning"))
                if reasoning_content:
                    reasoning = True
                    yield Reasoning(reasoning_content)
            if "usage" in data and data["usage"]:
                yield Usage(**data["usage"])
            if "conversation" in data and data["conversation"]:
                yield JsonConversation.from_dict(data["conversation"])
            if choice and choice.get("finish_reason") is not None:
                yield FinishReason(choice["finish_reason"])
    else:
        await raise_for_status(response)
        async for chunk in save_response_media(response, prompt, [model]):
            yield chunk