@samitouri / QOSami_Agents_Zero-2 / commits / 3b18b864

api_base fix + litellm update

frdel committed Jul 9, 2025 at 10:30 UTC 3b18b86454715bae7cdb33dd029938f5ecc5b2aa

7 files changed +69 -69

agent.py

+15 -26

@@ -202,25 +202,13 @@ class AgentContext:

202 agent.handle_critical_exception(e)

203

204

205 -@dataclass

206 -class ModelConfig:

207 - provider: models.ModelProvider

208 - name: str

209 - api_base: str = ""

210 - ctx_length: int = 0

211 - limit_requests: int = 0

212 - limit_input: int = 0

213 - limit_output: int = 0

214 - vision: bool = False

215 - kwargs: dict = field(default_factory=dict)

216 -

205

206 @dataclass

207 class AgentConfig:

220 - chat_model: ModelConfig

221 - utility_model: ModelConfig

222 - embeddings_model: ModelConfig

223 - browser_model: ModelConfig

208 + chat_model: models.ModelConfig

209 + utility_model: models.ModelConfig

210 + embeddings_model: models.ModelConfig

211 + browser_model: models.ModelConfig

212 mcp_servers: str

213 prompts_subdir: str = ""

214 memory_subdir: str = ""

@@ -582,29 +570,30 @@ class Agent:

570 return models.get_chat_model(

571 self.config.chat_model.provider,

572 self.config.chat_model.name,

585 - **self._get_model_kwargs(self.config.chat_model),

573 + **self.config.chat_model.build_kwargs(),

574 )

575

576 def get_utility_model(self):

577 return models.get_chat_model(

578 self.config.utility_model.provider,

579 self.config.utility_model.name,

592 - **self._get_model_kwargs(self.config.utility_model),

580 + **self.config.utility_model.build_kwargs(),

581 + )

582 +

583 + def get_browser_model(self):

584 + return models.get_browser_model(

585 + self.config.browser_model.provider,

586 + self.config.browser_model.name,

587 + **self.config.browser_model.build_kwargs(),

588 )

589

590 def get_embedding_model(self):

591 return models.get_embedding_model(

592 self.config.embeddings_model.provider,

593 self.config.embeddings_model.name,

599 - **self._get_model_kwargs(self.config.embeddings_model),

594 + **self.config.embeddings_model.build_kwargs(),

595 )

596

602 - def _get_model_kwargs(self, model_config: ModelConfig):

603 - kwargs = model_config.kwargs.copy() or {}

604 - if model_config.api_base and "api_base" not in kwargs:

605 - kwargs["api_base"] = model_config.api_base

606 - return kwargs

607 -

597 async def call_utility_model(

598 self,

599 system: str,

@@ -670,7 +659,7 @@ class Agent:

659 return response, reasoning

660

661 async def rate_limiter(

673 - self, model_config: ModelConfig, input: str, background: bool = False

662 + self, model_config: models.ModelConfig, input: str, background: bool = False

663 ):

664 # rate limiter log

665 wait_log = None

initialize.py

+9 -5

@@ -1,5 +1,5 @@

1 import models

2 -from agent import AgentConfig, ModelConfig

2 +from agent import AgentConfig

3 from python.helpers import runtime, settings, defer

4 from python.helpers.print_style import PrintStyle

@@ -26,7 +26,8 @@ def initialize_agent():

26 return result

28 # chat model from user settings

29 - chat_llm = ModelConfig(

29 + chat_llm = models.ModelConfig(

30 + type=models.ModelType.CHAT,

31 provider=models.ModelProvider[current_settings["chat_model_provider"]],

32 name=current_settings["chat_model_name"],

33 api_base=current_settings["chat_model_api_base"],

@@ -39,7 +40,8 @@ def initialize_agent():

40 )

42 # utility model from user settings

42 - utility_llm = ModelConfig(

43 + utility_llm = models.ModelConfig(

44 + type=models.ModelType.CHAT,

45 provider=models.ModelProvider[current_settings["util_model_provider"]],

46 name=current_settings["util_model_name"],

47 api_base=current_settings["util_model_api_base"],

@@ -50,7 +52,8 @@ def initialize_agent():

52 kwargs=_normalize_model_kwargs(current_settings["util_model_kwargs"]),

53 )

54 # embedding model from user settings

53 - embedding_llm = ModelConfig(

55 + embedding_llm = models.ModelConfig(

56 + type=models.ModelType.EMBEDDING,

57 provider=models.ModelProvider[current_settings["embed_model_provider"]],

58 name=current_settings["embed_model_name"],

59 api_base=current_settings["embed_model_api_base"],

@@ -58,7 +61,8 @@ def initialize_agent():

61 kwargs=_normalize_model_kwargs(current_settings["embed_model_kwargs"]),

62 )

63 # browser model from user settings

61 - browser_llm = ModelConfig(

64 + browser_llm = models.ModelConfig(

65 + type=models.ModelType.CHAT,

66 provider=models.ModelProvider[current_settings["browser_model_provider"]],

67 name=current_settings["browser_model_name"],

68 api_base=current_settings["browser_model_api_base"],

models.py

+27 -1

@@ -1,3 +1,4 @@

1 +from dataclasses import dataclass, field

2 from enum import Enum

3 import logging

4 import os

@@ -73,6 +74,26 @@ class ModelProvider(Enum):

74 OTHER = "Other OpenAI compatible"

77 +@dataclass

78 +class ModelConfig:

79 + type: ModelType

80 + provider: ModelProvider

81 + name: str

82 + api_base: str = ""

83 + ctx_length: int = 0

84 + limit_requests: int = 0

85 + limit_input: int = 0

86 + limit_output: int = 0

87 + vision: bool = False

88 + kwargs: dict = field(default_factory=dict)

89 +

90 + def build_kwargs(self):

91 + kwargs = self.kwargs.copy() or {}

92 + if self.api_base and "api_base" not in kwargs:

93 + kwargs["api_base"] = self.api_base

94 + return kwargs

95 +

96 +

97 class ChatChunk(TypedDict):

98 """Simplified response chunk for chat models."""

@@ -233,6 +254,9 @@ class LiteLLMChatWrapper(SimpleChatModel):

254 tokens_callback: Callable[[str, int], Awaitable[None]] | None = None,

255 **kwargs: Any,

256 ) -> Tuple[str, str]:

257 +

258 + turn_off_logging()

259 +

260 if not messages:

261 messages = []

262 # construct messages

@@ -408,7 +432,9 @@ def _get_litellm_embedding(model_name: str, provider_name: str, **kwargs: Any):

432

433 def _parse_chunk(chunk: Any) -> ChatChunk:

434 delta = chunk["choices"][0].get("delta", {})

411 - message = chunk["choices"][0].get("model_extra", {}).get("message", {})

435 + message = chunk["choices"][0].get("message", {}) or chunk["choices"][0].get(

436 + "model_extra", {}

437 + ).get("message", {})

438 response_delta = (

439 delta.get("content", "")

440 if isinstance(delta, dict)

python/helpers/memory.py

+3 -10

@@ -28,7 +28,7 @@ import uuid

28 from python.helpers import knowledge_import

29 from python.helpers.log import Log, LogItem

30 from enum import Enum

31 -from agent import Agent, ModelConfig

31 +from agent import Agent

32 import models

33 import logging

@@ -96,7 +96,7 @@ class Memory:

96 @staticmethod

97 def initialize(

98 log_item: LogItem | None,

99 - model_config: ModelConfig,

99 + model_config: models.ModelConfig,

100 memory_subdir: str,

101 in_memory=False,

102 ) -> tuple[MyFaiss, bool]:

@@ -120,17 +120,10 @@ class Memory:

120 os.makedirs(em_dir, exist_ok=True)

121 store = LocalFileStore(em_dir)

122

123 - # HOTFIX TODO: unify with agent.py

124 - def _get_model_kwargs(model_config: ModelConfig):

125 - kwargs = model_config.kwargs.copy() or {}

126 - if model_config.api_base and "api_base" not in kwargs:

127 - kwargs["api_base"] = model_config.api_base

128 - return kwargs

129 -

123 embeddings_model = models.get_embedding_model(

124 model_config.provider,

125 model_config.name,

133 - **_get_model_kwargs(model_config),

126 + **model_config.build_kwargs(),

127 )

128 embeddings_model_id = files.safe_file_name(

129 model_config.provider.name + "_" + model_config.name

python/helpers/settings.py

+13 -21

@@ -122,6 +122,7 @@ _settings: Settings | None = None

122

123 def convert_out(settings: Settings) -> SettingsOutput:

124 from models import ModelProvider

125 + default_settings = get_default_settings()

126

127 # main model section

128 chat_model_fields: list[SettingsField] = []

@@ -381,7 +382,7 @@ def convert_out(settings: Settings) -> SettingsOutput:

382 embed_model_section: SettingsSection = {

383 "id": "embed_model",

384 "title": "Embedding Model",

384 - "description": "Settings for the embedding model used by Agent Zero.",

385 + "description": f"Settings for the embedding model used by Agent Zero.<br><h4>⚠️ No need to change</h4>The default HuggingFace model {default_settings['embed_model_name']} is preloaded and runs locally within the docker container and there's no need to change it unless you have a specific requirements for embedding.",

386 "fields": embed_model_fields,

387 "tab": "agent",

388 }

@@ -408,6 +409,16 @@ def convert_out(settings: Settings) -> SettingsOutput:

409 }

410 )

411

412 + browser_model_fields.append(

413 + {

414 + "id": "browser_model_api_base",

415 + "title": "Web Browser model API base URL",

416 + "description": "API base URL for web browser model. Leave empty for default. Only relevant for Azure, local and custom (other) providers.",

417 + "type": "text",

418 + "value": settings["browser_model_api_base"],

419 + }

420 + )

421 +

422 browser_model_fields.append(

423 {

424 "id": "browser_model_vision",

@@ -436,24 +447,6 @@ def convert_out(settings: Settings) -> SettingsOutput:

447 "tab": "agent",

448 }

449

439 - # # Memory settings section

440 - # memory_fields: list[SettingsField] = []

441 - # memory_fields.append(

442 - # {

443 - # "id": "memory_settings",

444 - # "title": "Memory Settings",

445 - # "description": "<settings for memory>",

446 - # "type": "text",

447 - # "value": "",

448 - # }

449 - # )

450 -

451 - # memory_section: SettingsSection = {

452 - # "id": "memory",

453 - # "title": "Memory Settings",

454 - # "description": "<settings for memory management here>",

455 - # "fields": memory_fields,

456 - # }

450

451 # basic auth section

452 auth_fields: list[SettingsField] = []

@@ -829,9 +822,8 @@ def convert_out(settings: Settings) -> SettingsOutput:

822 agent_section,

823 chat_model_section,

824 util_model_section,

832 - embed_model_section,

825 browser_model_section,

834 - # memory_section,

826 + embed_model_section,

827 stt_section,

828 api_keys_section,

829 auth_section,

python/tools/browser_agent.py

+1 -5

@@ -125,11 +125,7 @@ class State:

125 )

126 return result

127

128 - model = models.get_browser_model(

129 - provider=self.agent.config.browser_model.provider,

130 - name=self.agent.config.browser_model.name,

131 - **self.agent._get_model_kwargs(self.agent.config.browser_model),

132 - )

128 + model = self.agent.get_browser_model()

129

130 try:

131 self.use_agent = browser_use.Agent(

requirements.txt

+1 -1

@@ -30,7 +30,7 @@ unstructured-client==0.31.0

30 webcolors==24.6.0

31 nest-asyncio==1.6.0

32 crontab==1.0.1

33 -litellm==1.72.4

33 +litellm==1.74

34 markdownify==1.1.0

35 pymupdf==1.25.3

36 pytesseract==0.3.13