返回提交历史
Modified
g4f/Provider/Blackbox.py
+43
-49
Modified
g4f/Provider/Cloudflare.py
+5
-6
Modified
g4f/Provider/Copilot.py
+2
-2
Modified
g4f/Provider/openai/har_file.py
+1
-0
Modified
g4f/api/__init__.py
+2
-2
Modified
g4f/client/__init__.py
+2
-2
Modified
g4f/client/stubs.py
+28
-3
Modified
g4f/cookies.py
+2
-2
Modified
g4f/gui/client/background.html
+8
-5
Modified
g4f/gui/server/backend_api.py
+22
-15
Modified
g4f/providers/helper.py
+10
-0
XFEstudio/gpt4free
feat: introduce render_messages and enhance HAR/conversation handling
- **g4f/providers/helper.py** - Add `render_messages()` to normalise message contents that are lists of blocks. - **g4f/Provider/Blackbox.py** - Import `get_har_files` and `render_messages`. - Replace manual walk of `get_cookies_dir()` with `get_har_files()` in `_find_session_in_har`. - Simplify session‑parsing loop and exception logging; drop permissions check. - Build `current_messages` with `render_messages(messages)` instead of raw list. - **g4f/Provider/Cloudflare.py** - Swap `to_string` import for `render_messages`. - Add `"impersonate": "chrome"` to default `_args`. - Construct `data["messages"]` with `render_messages(messages)` and inline `"parts"`; remove `to_string()` calls. - Move `cache_file` write outside inner `try` to always save arguments. - **g4f/Provider/Copilot.py** - Defer `yield conversation` until after `conversation` is created when `return_conversation` is requested. - **g4f/Provider/openai/har_file.py** - Break out of `os.walk` after first directory in `get_har_files()` to avoid deep traversal. - **g4f/api/__init__.py** - Use `config.conversation` directly and set `return_conversation` when present. - **g4f/client/__init__.py** - Pass `conversation` to both `ChatCompletionChunk.model_construct()` and `ChatCompletion.model_construct()`. - **g4f/client/stubs.py** - Import `field_serializer` (with stub fallback). - Add serializers for `conversation` (objects and dicts) and for `content` fields. - Extend model constructors to accept/propagate `conversation`. - **g4f/cookies.py** - Insert ".huggingface.co" into `DOMAINS` list. - Stop recursive directory walk in `read_cookie_files()` with early `break`. - **g4f/gui/client/background.html** - Reorder error‑handling branches; reset `errorImage` in `onload`. - Revise `skipRefresh` logic and random image URL building. - **g4f/gui/server/backend_api.py** - Add `self.match_files` cache for repeated image searches. - Use `safe_search` for sanitised term matching and `min` comparison. - Limit walk to one directory level; support deterministic random selection via `random` query param. - **Miscellaneous** - Update imports where `render_messages` replaces `to_string`. - Ensure all modified providers iterate messages through `render_messages` for consistent formatting.
3ab36ebc
代码差异
11 个文件
+125
-86
@@ -14,9 +14,10 @@ from datetime import datetime, timedelta
14
14
from ..typing import AsyncResult, Messages, MediaListType
15
15
from ..requests.raise_for_status import raise_for_status
16
16
from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
17
from .openai.har_file import get_har_files
17
18
from ..image import to_data_uri
18
19
from ..cookies import get_cookies_dir
19
from .helper import format_image_prompt
20
from .helper import format_image_prompt, render_messages
20
21
from ..providers.response import JsonConversation, ImageResponse
21
22
from ..tools.media import merge_media
22
23
from ..errors import RateLimitError
@@ -428,53 +429,46 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
428
429
Optional[dict]: Session data if found, None otherwise
429
430
"""
430
431
try:
431
har_dir = get_cookies_dir()
432
if not os.access(har_dir, os.R_OK):
433
return None
434
435
for root, _, files in os.walk(har_dir):
436
for file in files:
437
if file.endswith(".har"):
438
try:
439
with open(os.path.join(root, file), 'rb') as f:
440
har_data = json.load(f)
441
442
for entry in har_data['log']['entries']:
443
# Only look at blackbox API responses
444
if 'blackbox.ai/api' in entry['request']['url']:
445
# Look for a response that has the right structure
446
if 'response' in entry and 'content' in entry['response']:
447
content = entry['response']['content']
448
# Look for both regular and Google auth session formats
449
if ('text' in content and
450
isinstance(content['text'], str) and
451
'"user"' in content['text'] and
452
'"email"' in content['text'] and
453
'"expires"' in content['text']):
454
455
try:
456
# Remove any HTML or other non-JSON content
457
text = content['text'].strip()
458
if text.startswith('{') and text.endswith('}'):
459
# Replace escaped quotes
460
text = text.replace('\\"', '"')
461
har_session = json.loads(text)
462
463
# Check if this is a valid session object
464
if (isinstance(har_session, dict) and
465
'user' in har_session and
466
'email' in har_session['user'] and
467
'expires' in har_session):
468
469
debug.log(f"Blackbox: Found session in HAR file: {file}")
470
return har_session
471
except json.JSONDecodeError as e:
472
# Only print error for entries that truly look like session data
473
if ('"user"' in content['text'] and
474
'"email"' in content['text']):
475
debug.log(f"Blackbox: Error parsing likely session data: {e}")
476
except Exception as e:
477
debug.log(f"Blackbox: Error reading HAR file {file}: {e}")
432
for file in get_har_files():
433
try:
434
with open(file, 'rb') as f:
435
har_data = json.load(f)
436
437
for entry in har_data['log']['entries']:
438
# Only look at blackbox API responses
439
if 'blackbox.ai/api' in entry['request']['url']:
440
# Look for a response that has the right structure
441
if 'response' in entry and 'content' in entry['response']:
442
content = entry['response']['content']
443
# Look for both regular and Google auth session formats
444
if ('text' in content and
445
isinstance(content['text'], str) and
446
'"user"' in content['text'] and
447
'"email"' in content['text'] and
448
'"expires"' in content['text']):
449
try:
450
# Remove any HTML or other non-JSON content
451
text = content['text'].strip()
452
if text.startswith('{') and text.endswith('}'):
453
# Replace escaped quotes
454
text = text.replace('\\"', '"')
455
har_session = json.loads(text)
456
457
# Check if this is a valid session object
458
if (isinstance(har_session, dict) and
459
'user' in har_session and
460
'email' in har_session['user'] and
461
'expires' in har_session):
462
463
debug.log(f"Blackbox: Found session in HAR file: {file}")
464
return har_session
465
except json.JSONDecodeError as e:
466
# Only print error for entries that truly look like session data
467
if ('"user"' in content['text'] and
468
'"email"' in content['text']):
469
debug.log(f"Blackbox: Error parsing likely session data: {e}")
470
except Exception as e:
471
debug.log(f"Blackbox: Error reading HAR file {file}: {e}")
478
472
return None
479
473
except Exception as e:
480
474
debug.log(f"Blackbox: Error searching HAR files: {e}")
@@ -573,7 +567,7 @@ class Blackbox(AsyncGeneratorProvider, ProviderModelMixin):
573
567
conversation.message_history = []
574
568
575
569
current_messages = []
576
for i, msg in enumerate(messages):
570
for i, msg in enumerate(render_messages(messages)):
577
571
msg_id = conversation.chat_id if i == 0 and msg["role"] == "user" else cls.generate_id()
578
572
current_msg = {
579
573
"id": msg_id,
@@ -9,7 +9,7 @@ from ..requests import Session, StreamSession, get_args_from_nodriver, raise_for
9
9
from ..requests import DEFAULT_HEADERS, has_nodriver, has_curl_cffi
10
10
from ..providers.response import FinishReason, Usage
11
11
from ..errors import ResponseStatusError, ModelNotFoundError
12
from .helper import to_string
12
from .helper import render_messages
13
13
14
14
class Cloudflare(AsyncGeneratorProvider, ProviderModelMixin, AuthFileMixin):
15
15
label = "Cloudflare AI"
@@ -82,7 +82,7 @@ class Cloudflare(AsyncGeneratorProvider, ProviderModelMixin, AuthFileMixin):
82
82
elif has_nodriver:
83
83
cls._args = await get_args_from_nodriver(cls.url, proxy, timeout, cookies)
84
84
else:
85
cls._args = {"headers": DEFAULT_HEADERS, "cookies": {}}
85
cls._args = {"headers": DEFAULT_HEADERS, "cookies": {}, "impersonate": "chrome"}
86
86
try:
87
87
model = cls.get_model(model)
88
88
except ModelNotFoundError:
@@ -90,8 +90,7 @@ class Cloudflare(AsyncGeneratorProvider, ProviderModelMixin, AuthFileMixin):
90
90
data = {
91
91
"messages": [{
92
92
**message,
93
"content": to_string(message["content"]),
94
"parts": [{"type":"text", "text": to_string(message["content"])}]} for message in messages],
93
"parts": [{"type":"text", "text": message["content"]}]} for message in render_messages(messages)],
95
94
"lora": None,
96
95
"model": model,
97
96
"max_tokens": max_tokens,
@@ -120,5 +119,5 @@ class Cloudflare(AsyncGeneratorProvider, ProviderModelMixin, AuthFileMixin):
120
119
yield Usage(**finish.get("usage"))
121
120
yield FinishReason(finish.get("finishReason"))
122
121
123
with cache_file.open("w") as f:
124
json.dump(cls._args, f)
122
with cache_file.open("w") as f:
123
json.dump(cls._args, f)
@@ -116,8 +116,6 @@ class Copilot(AsyncGeneratorProvider, ProviderModelMixin):
116
116
response.raise_for_status()
117
117
conversation_id = response.json().get("id")
118
118
conversation = Conversation(conversation_id)
119
if return_conversation:
120
yield conversation
121
119
if prompt is None:
122
120
prompt = format_prompt_max_length(messages, 10000)
123
121
debug.log(f"Copilot: Created conversation: {conversation_id}")
@@ -126,6 +124,8 @@ class Copilot(AsyncGeneratorProvider, ProviderModelMixin):
126
124
if prompt is None:
127
125
prompt = get_last_user_message(messages)
128
126
debug.log(f"Copilot: Use conversation: {conversation_id}")
127
if return_conversation:
128
yield conversation
129
129
130
130
uploaded_images = []
131
131
for media, _ in merge_media(media, messages):
@@ -49,6 +49,7 @@ def get_har_files():
49
49
for file in files:
50
50
if file.endswith(".har"):
51
51
harPath.append(os.path.join(root, file))
52
break
52
53
if not harPath:
53
54
raise NoValidHarFileError("No .har file found")
54
55
harPath.sort(key=lambda x: os.path.getmtime(x))
@@ -309,9 +309,9 @@ class Api:
309
309
if credentials is not None and credentials.credentials != "secret":
310
310
config.api_key = credentials.credentials
311
311
312
conversation = None
312
conversation = config.conversation
313
313
return_conversation = config.return_conversation
314
if conversation is not None:
314
if conversation:
315
315
conversation = JsonConversation(**conversation)
316
316
return_conversation = True
317
317
elif config.conversation_id is not None and config.provider is not None:
@@ -217,7 +217,7 @@ async def async_iter_response(
217
217
218
218
if stream:
219
219
chat_completion = ChatCompletionChunk.model_construct(
220
None, finish_reason, completion_id, int(time.time()), usage=usage
220
None, finish_reason, completion_id, int(time.time()), usage=usage, conversation=conversation
221
221
)
222
222
else:
223
223
if response_format is not None and "type" in response_format:
@@ -228,7 +228,7 @@ async def async_iter_response(
228
228
**filter_none(
229
229
tool_calls=[ToolCallModel.model_construct(**tool_call) for tool_call in tool_calls]
230
230
) if tool_calls is not None else {},
231
conversation=None if conversation is None else conversation.get_dict()
231
conversation=conversation
232
232
)
233
233
if provider is not None:
234
234
chat_completion.provider = provider.name
@@ -10,7 +10,7 @@ from ..client.helper import filter_markdown
10
10
from .helper import filter_none
11
11
12
12
try:
13
from pydantic import BaseModel
13
from pydantic import BaseModel, field_serializer
14
14
except ImportError:
15
15
class BaseModel():
16
16
@classmethod
@@ -19,6 +19,9 @@ except ImportError:
19
19
for key, value in data.items():
20
20
setattr(new, key, value)
21
21
return new
22
class field_serializer():
23
def __init__(self, field_name):
24
self.field_name = field_name
22
25
23
26
class BaseModel(BaseModel):
24
27
@classmethod
@@ -72,6 +75,7 @@ class ChatCompletionChunk(BaseModel):
72
75
provider: Optional[str]
73
76
choices: List[ChatCompletionDeltaChoice]
74
77
usage: UsageModel
78
conversation: dict
75
79
76
80
@classmethod
77
81
def model_construct(
@@ -80,7 +84,8 @@ class ChatCompletionChunk(BaseModel):
80
84
finish_reason: str,
81
85
completion_id: str = None,
82
86
created: int = None,
83
usage: UsageModel = None
87
usage: UsageModel = None,
88
conversation: dict = None
84
89
):
85
90
return super().model_construct(
86
91
id=f"chatcmpl-{completion_id}" if completion_id else None,
@@ -92,9 +97,15 @@ class ChatCompletionChunk(BaseModel):
92
97
ChatCompletionDelta.model_construct(content),
93
98
finish_reason
94
99
)],
95
**filter_none(usage=usage)
100
**filter_none(usage=usage, conversation=conversation)
96
101
)
97
102
103
@field_serializer('conversation')
104
def serialize_conversation(self, conversation: dict):
105
if hasattr(conversation, "get_dict"):
106
return conversation.get_dict()
107
return conversation
108
98
109
class ChatCompletionMessage(BaseModel):
99
110
role: str
100
111
content: str
@@ -104,6 +115,10 @@ class ChatCompletionMessage(BaseModel):
104
115
def model_construct(cls, content: str, tool_calls: list = None):
105
116
return super().model_construct(role="assistant", content=content, **filter_none(tool_calls=tool_calls))
106
117
118
@field_serializer('content')
119
def serialize_content(self, content: str):
120
return str(content)
121
107
122
def save(self, filepath: str, allowd_types = None):
108
123
if hasattr(self.content, "data"):
109
124
os.rename(self.content.data.replace("/media", images_dir), filepath)
@@ -160,6 +175,12 @@ class ChatCompletion(BaseModel):
160
175
**filter_none(usage=usage, conversation=conversation)
161
176
)
162
177
178
@field_serializer('conversation')
179
def serialize_conversation(self, conversation: dict):
180
if hasattr(conversation, "get_dict"):
181
return conversation.get_dict()
182
return conversation
183
163
184
class ChatCompletionDelta(BaseModel):
164
185
role: str
165
186
content: str
@@ -168,6 +189,10 @@ class ChatCompletionDelta(BaseModel):
168
189
def model_construct(cls, content: Optional[str]):
169
190
return super().model_construct(role="assistant", content=content)
170
191
192
@field_serializer('content')
193
def serialize_content(self, content: str):
194
return str(content)
195
171
196
class ChatCompletionDeltaChoice(BaseModel):
172
197
index: int
173
198
delta: ChatCompletionDelta
@@ -56,12 +56,11 @@ DOMAINS = [
56
56
".google.com",
57
57
"www.whiterabbitneo.com",
58
58
"huggingface.co",
59
".huggingface.co"
59
60
"chat.reka.ai",
60
61
"chatgpt.com",
61
62
".cerebras.ai",
62
63
"github.com",
63
"huggingface.co",
64
".huggingface.co"
65
64
]
66
65
67
66
if has_browser_cookie3 and os.environ.get('DBUS_SESSION_BUS_ADDRESS') == "/dev/null":
@@ -152,6 +151,7 @@ def read_cookie_files(dirPath: str = None):
152
151
harFiles.append(os.path.join(root, file))
153
152
elif file.endswith(".json"):
154
153
cookieFiles.append(os.path.join(root, file))
154
break
155
155
156
156
CookiesConfig.cookies = {}
157
157
for path in harFiles:
@@ -169,15 +169,15 @@
169
169
if (errorVideo < 3 || !refreshOnHide) {
170
170
return;
171
171
}
172
if (skipRefresh > 0) {
173
skipRefresh -= 1;
174
return;
175
}
172
176
if (errorImage < 3) {
173
177
imageFeed.src = "/search/image+g4f?skip=" + skipImage;
174
178
skipImage++;
175
179
return;
176
180
}
177
if (skipRefresh > 0) {
178
skipRefresh -= 1;
179
return;
180
}
181
181
if (images.length > 0) {
182
182
imageFeed.classList.remove("hidden");
183
183
imageFeed.src = images.shift();
@@ -194,10 +194,13 @@
194
194
imageFeed.onload = () => {
195
195
imageFeed.classList.remove("hidden");
196
196
gradient.classList.add("hidden");
197
errorImage = 0;
197
198
};
198
199
imageFeed.onclick = () => {
199
200
imageFeed.src = "/search/image?random=" + Math.random();
200
skipRefresh = 2;
201
if (skipRefresh < 4) {
202
skipRefresh += 1;
203
}
201
204
};
202
205
})();
203
206
</script>
@@ -341,28 +341,35 @@ class Backend_Api(Api):
341
341
return redirect(source_url)
342
342
raise
343
343
344
self.match_files = {}
345
344
346
@app.route('/search/<search>', methods=['GET'])
345
347
def find_media(search: str):
346
search = [secure_filename(chunk.lower()) for chunk in search.split("+")]
348
safe_search = [secure_filename(chunk.lower()) for chunk in search.split("+")]
347
349
if not os.access(images_dir, os.R_OK):
348
350
return jsonify({"error": {"message": "Not found"}}), 404
349
match_files = {}
350
for root, _, files in os.walk(images_dir):
351
for file in files:
352
mime_type = is_allowed_extension(file)
353
if mime_type is not None:
354
mime_type = secure_filename(mime_type)
355
for tag in search:
356
if tag in mime_type:
357
match_files[file] = match_files.get(file, 0) + 1
358
break
359
for tag in search:
360
if tag in file.lower():
361
match_files[file] = match_files.get(file, 0) + 1
362
match_files = [file for file, count in match_files.items() if count >= request.args.get("min", len(search))]
351
if search not in self.match_files:
352
self.match_files[search] = {}
353
for root, _, files in os.walk(images_dir):
354
for file in files:
355
mime_type = is_allowed_extension(file)
356
if mime_type is not None:
357
mime_type = secure_filename(mime_type)
358
for tag in safe_search:
359
if tag in mime_type:
360
self.match_files[search][file] = self.match_files[search].get(file, 0) + 1
361
break
362
for tag in safe_search:
363
if tag in file.lower():
364
self.match_files[search][file] = self.match_files[search].get(file, 0) + 1
365
break
366
match_files = [file for file, count in self.match_files[search].items() if count >= request.args.get("min", len(safe_search))]
363
367
if int(request.args.get("skip", 0)) >= len(match_files):
364
368
return jsonify({"error": {"message": "Not found"}}), 404
365
369
if (request.args.get("random", False)):
370
seed = request.args.get("random")
371
if seed not in ["true", "True", "1"]:
372
random.seed(seed)
366
373
return redirect(f"/media/{random.choice(match_files)}"), 302
367
374
return redirect(f"/media/{match_files[int(request.args.get('skip', 0))]}", 302)
368
375
@@ -24,6 +24,16 @@ def to_string(value) -> str:
24
24
return "".join([to_string(v) for v in value if v.get("type", "text") == "text"])
25
25
return str(value)
26
26
27
def render_messages(messages: Messages) -> Iterator:
28
for idx, message in enumerate(messages):
29
if isinstance(message, dict) and isinstance(message.get("content"), list):
30
yield {
31
**message,
32
"content": to_string(message["content"]),
33
}
34
else:
35
yield message
36
27
37
def format_prompt(messages: Messages, add_special_tokens: bool = False, do_continue: bool = False, include_system: bool = True) -> str:
28
38
"""
29
39
Format a series of messages into a single string, optionally adding special tokens.