Skip to content

Agent Zero AI framework

models temp removed, deepseek reasoner support

frdel7a7d2eebc996
diff --git a/README.md b/README.md
index 4a3c94d76..0a7a9a528 100644
--- a/README.md
+++ b/README.md
@@ -14,6 +14,8 @@
</div>
+[![Browser Agent](/docs/res/081_vid.png)](https://youtu.be/quv145buW74)
+
> [!NOTE]
> **🎉 v0.8.1 Release**: Now featuring a browser agent capable of using Chromium for web interactions! This enables Agent Zero to browse the web, gather information, and interact with web content autonomously.
diff --git a/docs/res/081_vid.png b/docs/res/081_vid.png
new file mode 100644
index 000000000..c4e7349af
Binary files /dev/null and b/docs/res/081_vid.png differ
diff --git a/initialize.py b/initialize.py
index 72a9d5129..ead07d2f2 100644
--- a/initialize.py
+++ b/initialize.py
@@ -16,10 +16,7 @@ def initialize():
limit_requests=current_settings["chat_model_rl_requests"],
limit_input=current_settings["chat_model_rl_input"],
limit_output=current_settings["chat_model_rl_output"],
- kwargs={
- "temperature": current_settings["chat_model_temperature"],
- **current_settings["chat_model_kwargs"],
- },
+ kwargs=current_settings["chat_model_kwargs"],
)
# utility model from user settings
@@ -30,29 +27,21 @@ def initialize():
limit_requests=current_settings["util_model_rl_requests"],
limit_input=current_settings["util_model_rl_input"],
limit_output=current_settings["util_model_rl_output"],
- kwargs={
- "temperature": current_settings["util_model_temperature"],
- **current_settings["util_model_kwargs"],
- },
+ kwargs=current_settings["util_model_kwargs"],
)
# embedding model from user settings
embedding_llm = ModelConfig(
provider=models.ModelProvider[current_settings["embed_model_provider"]],
name=current_settings["embed_model_name"],
limit_requests=current_settings["embed_model_rl_requests"],
- kwargs={
- **current_settings["embed_model_kwargs"],
- },
+ kwargs=current_settings["embed_model_kwargs"],
)
# browser model from user settings
browser_llm = ModelConfig(
provider=models.ModelProvider[current_settings["browser_model_provider"]],
name=current_settings["browser_model_name"],
vision=current_settings["browser_model_vision"],
- kwargs={
- "temperature": current_settings["browser_model_temperature"],
- **current_settings["browser_model_kwargs"],
- },
+ kwargs=current_settings["browser_model_kwargs"],
)
# agent configuration
config = AgentConfig(
diff --git a/models.py b/models.py
index 4887f33cf..a541b3197 100644
--- a/models.py
+++ b/models.py
@@ -26,6 +26,7 @@ from langchain_google_genai import (
embeddings as google_embeddings,
)
from langchain_mistralai import ChatMistralAI
+
# from pydantic.v1.types import SecretStr
from python.helpers import dotenv, runtime
from python.helpers.dotenv import load_dotenv
@@ -34,9 +35,6 @@ from python.helpers.rate_limiter import RateLimiter
# environment variables
load_dotenv()
-# Configuration
-DEFAULT_TEMPERATURE = 0.0
-
class ModelType(Enum):
CHAT = "Chat"
@@ -110,7 +108,6 @@ def get_ollama_base_url():
def get_ollama_chat(
model_name: str,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
num_ctx=8192,
**kwargs,
@@ -119,7 +116,6 @@ def get_ollama_chat(
base_url = get_ollama_base_url()
return ChatOllama(
model=model_name,
- temperature=temperature,
base_url=base_url,
num_ctx=num_ctx,
**kwargs,
@@ -128,14 +124,14 @@ def get_ollama_chat(
def get_ollama_embedding(
model_name: str,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
+ num_ctx=8192,
**kwargs,
):
if not base_url:
base_url = get_ollama_base_url()
return OllamaEmbeddings(
- model=model_name, temperature=temperature, base_url=base_url, **kwargs
+ model=model_name, base_url=base_url, num_ctx=num_ctx, **kwargs
)
@@ -143,7 +139,6 @@ def get_ollama_embedding(
def get_huggingface_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
**kwargs,
):
# different naming convention here
@@ -155,7 +150,6 @@ def get_huggingface_chat(
repo_id=model_name,
task="text-generation",
do_sample=True,
- temperature=temperature,
**kwargs,
)
@@ -177,13 +171,12 @@ def get_lmstudio_base_url():
def get_lmstudio_chat(
model_name: str,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
**kwargs,
):
if not base_url:
base_url = get_lmstudio_base_url()
- return ChatOpenAI(model_name=model_name, base_url=base_url, temperature=temperature, api_key="none", **kwargs) # type: ignore
+ return ChatOpenAI(model_name=model_name, base_url=base_url, api_key="none", **kwargs) # type: ignore
def get_lmstudio_embedding(
@@ -200,12 +193,16 @@ def get_lmstudio_embedding(
def get_anthropic_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
+ base_url=None,
**kwargs,
):
if not api_key:
api_key = get_api_key("anthropic")
- return ChatAnthropic(model_name=model_name, temperature=temperature, api_key=api_key, **kwargs) # type: ignore
+ if not base_url:
+ base_url = (
+ dotenv.get_dotenv_value("ANTHROPIC_BASE_URL") or "https://api.anthropic.com"
+ )
+ return ChatAnthropic(model_name=model_name, api_key=api_key, base_url=base_url, **kwargs) # type: ignore
# right now anthropic does not have embedding models, but that might change
@@ -223,12 +220,11 @@ def get_anthropic_embedding(
def get_openai_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
**kwargs,
):
if not api_key:
api_key = get_api_key("openai")
- return ChatOpenAI(model_name=model_name, temperature=temperature, api_key=api_key, **kwargs) # type: ignore
+ return ChatOpenAI(model_name=model_name, api_key=api_key, **kwargs) # type: ignore
def get_openai_embedding(model_name: str, api_key=None, **kwargs):
@@ -240,7 +236,6 @@ def get_openai_embedding(model_name: str, api_key=None, **kwargs):
def get_openai_azure_chat(
deployment_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
azure_endpoint=None,
**kwargs,
):
@@ -248,7 +243,7 @@ def get_openai_azure_chat(
api_key = get_api_key("openai_azure")
if not azure_endpoint:
azure_endpoint = dotenv.get_dotenv_value("OPENAI_AZURE_ENDPOINT")
- return AzureChatOpenAI(deployment_name=deployment_name, temperature=temperature, api_key=api_key, azure_endpoint=azure_endpoint, **kwargs) # type: ignore
+ return AzureChatOpenAI(deployment_name=deployment_name, api_key=api_key, azure_endpoint=azure_endpoint, **kwargs) # type: ignore
def get_openai_azure_embedding(
@@ -268,12 +263,11 @@ def get_openai_azure_embedding(
def get_google_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
**kwargs,
):
if not api_key:
api_key = get_api_key("google")
- return GoogleGenerativeAI(model=model_name, temperature=temperature, google_api_key=api_key, safety_settings={HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE}, **kwargs) # type: ignore
+ return GoogleGenerativeAI(model=model_name, google_api_key=api_key, safety_settings={HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE}, **kwargs) # type: ignore
def get_google_embedding(
@@ -290,31 +284,28 @@ def get_google_embedding(
def get_mistralai_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
**kwargs,
):
if not api_key:
api_key = get_api_key("mistral")
- return ChatMistralAI(model=model_name, temperature=temperature, api_key=api_key, **kwargs) # type: ignore
+ return ChatMistralAI(model=model_name, api_key=api_key, **kwargs) # type: ignore
# Groq models
def get_groq_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
**kwargs,
):
if not api_key:
api_key = get_api_key("groq")
- return ChatGroq(model_name=model_name, temperature=temperature, api_key=api_key, **kwargs) # type: ignore
+ return ChatGroq(model_name=model_name, api_key=api_key, **kwargs) # type: ignore
# DeepSeek models
def get_deepseek_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
**kwargs,
):
@@ -322,17 +313,19 @@ def get_deepseek_chat(
api_key = get_api_key("deepseek")
if not base_url:
base_url = (
- dotenv.get_dotenv_value("DEEPSEEK_BASE_URL")
- or "https://api.deepseek.com"
+ dotenv.get_dotenv_value("DEEPSEEK_BASE_URL") or "https://api.deepseek.com"
)
- return ChatOpenAI(api_key=api_key, model=model_name, temperature=temperature, base_url=base_url, **kwargs) # type: ignore
-
+
+ model = ChatOpenAI(api_key=api_key, model=model_name, base_url=base_url, **kwargs) # type: ignore
+ # little hack for reasoning model's problem with temperature
+ if not "temperature" in kwargs:
+ model.temperature = None # type: ignore
+ return model
# OpenRouter models
def get_openrouter_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
**kwargs,
):
@@ -343,7 +336,7 @@ def get_openrouter_chat(
dotenv.get_dotenv_value("OPEN_ROUTER_BASE_URL")
or "https://openrouter.ai/api/v1"
)
- return ChatOpenAI(api_key=api_key, model=model_name, temperature=temperature, base_url=base_url, **kwargs) # type: ignore
+ return ChatOpenAI(api_key=api_key, model=model_name, base_url=base_url, **kwargs) # type: ignore
def get_openrouter_embedding(
@@ -366,7 +359,6 @@ def get_openrouter_embedding(
def get_sambanova_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
max_tokens=1024,
**kwargs,
@@ -378,7 +370,7 @@ def get_sambanova_chat(
dotenv.get_dotenv_value("SAMBANOVA_BASE_URL")
or "https://fast-api.snova.ai/v1"
)
- return ChatOpenAI(api_key=api_key, model=model_name, temperature=temperature, base_url=base_url, max_tokens=max_tokens, **kwargs) # type: ignore
+ return ChatOpenAI(api_key=api_key, model=model_name, base_url=base_url, max_tokens=max_tokens, **kwargs) # type: ignore
# right now sambanova does not have embedding models, but that might change
@@ -402,11 +394,10 @@ def get_sambanova_embedding(
def get_other_chat(
model_name: str,
api_key=None,
- temperature=DEFAULT_TEMPERATURE,
base_url=None,
**kwargs,
):
- return ChatOpenAI(api_key=api_key, model=model_name, temperature=temperature, base_url=base_url, **kwargs) # type: ignore
+ return ChatOpenAI(api_key=api_key, model=model_name, base_url=base_url, **kwargs) # type: ignore
def get_other_embedding(model_name: str, api_key=None, base_url=None, **kwargs):
diff --git a/prompts/default/agent.system.main.role.md b/prompts/default/agent.system.main.role.md
index 4dab77600..f8a3ab932 100644
--- a/prompts/default/agent.system.main.role.md
+++ b/prompts/default/agent.system.main.role.md
@@ -1,5 +1,5 @@
## Your role
-{{agent_name}} autonomous json ai agent
+agent zero autonomous json ai agent
solve superior tasks using tools and subordinates
follow behavioral rules instructions
execute code actions yourself not instruct superior
diff --git a/python/helpers/settings.py b/python/helpers/settings.py
index ea5a67b55..2e49858d3 100644
--- a/python/helpers/settings.py
+++ b/python/helpers/settings.py
@@ -13,7 +13,6 @@ from . import files, dotenv
class Settings(TypedDict):
chat_model_provider: str
chat_model_name: str
- chat_model_temperature: float
chat_model_kwargs: dict[str, str]
chat_model_ctx_length: int
chat_model_ctx_history: float
@@ -23,7 +22,6 @@ class Settings(TypedDict):
util_model_provider: str
util_model_name: str
- util_model_temperature: float
util_model_kwargs: dict[str, str]
util_model_ctx_length: int
util_model_ctx_input: float
@@ -40,7 +38,6 @@ class Settings(TypedDict):
browser_model_provider: str
browser_model_name: str
browser_model_vision: bool
- browser_model_temperature: float
browser_model_kwargs: dict[str, str]
agent_prompts_subdir: str
@@ -129,19 +126,6 @@ def convert_out(settings: Settings) -> SettingsOutput:
}
)
- chat_model_fields.append(
- {
- "id": "chat_model_temperature",
- "title": "Chat model temperature",
- "description": "Determines the randomness of generated responses. 0 is deterministic, 1 is random",
- "type": "range",
- "min": 0,
- "max": 1,
- "step": 0.01,
- "value": settings["chat_model_temperature"],
- }
- )
-
chat_model_fields.append(
{
"id": "chat_model_ctx_length",
@@ -234,41 +218,6 @@ def convert_out(settings: Settings) -> SettingsOutput:
}
)
- util_model_fields.append(
- {
- "id": "util_model_temperature",
- "title": "Utility model temperature",
- "description": "Determines the randomness of generated responses. 0 is deterministic, 1 is random",
- "type": "range",
- "min": 0,
- "max": 1,
- "step": 0.01,
- "value": settings["util_model_temperature"],
- }
- )
-
- # util_model_fields.append(
- # {
- # "id": "util_model_ctx_length",
- # "title": "Utility model context length",
- # "description": "Maximum number of tokens in the context window for LLM. System prompt, message and response all count towards this limit.",
- # "type": "number",
- # "value": settings["util_model_ctx_length"],
- # }
- # )
- # util_model_fields.append(
- # {
- # "id": "util_model_ctx_input",
- # "title": "Context window space for input tokens",
- # "description": "Portion of context window dedicated to input tokens. The remaining space can be filled with response.",
- # "type": "range",
- # "min": 0.01,
- # "max": 1,
- # "step": 0.01,
- # "value": settings["util_model_ctx_input"],
- # }
- # )
-
util_model_fields.append(
{
"id": "util_model_rl_requests",
@@ -407,19 +356,6 @@ def convert_out(settings: Settings) -> SettingsOutput:
}
)
- browser_model_fields.append(
- {
- "id": "browser_model_temperature",
- "title": "Web Browser model temperature",
- "description": "Determines the randomness of generated responses. 0 is deterministic, 1 is random",
- "type": "range",
- "min": 0,
- "max": 1,
- "step": 0.01,
- "value": settings["browser_model_temperature"],
- }
- )
-
browser_model_fields.append(
{
"id": "browser_model_kwargs",
@@ -509,9 +445,7 @@ def convert_out(settings: Settings) -> SettingsOutput:
)
api_keys_fields.append(_get_api_key_field(settings, "groq", "Groq API Key"))
api_keys_fields.append(_get_api_key_field(settings, "google", "Google API Key"))
- api_keys_fields.append(
- _get_api_key_field(settings, "deepseek", "DeepSeek API Key")
- )
+ api_keys_fields.append(_get_api_key_field(settings, "deepseek", "DeepSeek API Key"))
api_keys_fields.append(
_get_api_key_field(settings, "openrouter", "OpenRouter API Key")
)
@@ -794,41 +728,6 @@ def normalize_settings(settings: Settings) -> Settings:
return copy
-# def get_chat_model(settings: Settings | None = None) -> BaseChatModel:
-# if not settings:
-# settings = get_settings()
-# return get_model(
-# type=ModelType.CHAT,
-# provider=ModelProvider[settings["chat_model_provider"]],
-# name=settings["chat_model_name"],
-# temperature=settings["chat_model_temperature"],
-# **settings["chat_model_kwargs"],
-# )
-
-
-# def get_utility_model(settings: Settings | None = None) -> BaseChatModel:
-# if not settings:
-# settings = get_settings()
-# return get_model(
-# type=ModelType.CHAT,
-# provider=ModelProvider[settings["util_model_provider"]],
-# name=settings["util_model_name"],
-# temperature=settings["util_model_temperature"],
-# **settings["util_model_kwargs"],
-# )
-
-
-# def get_embedding_model(settings: Settings | None = None) -> Embeddings:
-# if not settings:
-# settings = get_settings()
-# return get_model(
-# type=ModelType.EMBEDDING,
-# provider=ModelProvider[settings["embed_model_provider"]],
-# name=settings["embed_model_name"],
-# **settings["embed_model_kwargs"],
-# )
-
-
def _read_settings_file() -> Settings | None:
if os.path.exists(SETTINGS_FILE):
content = files.read_file(SETTINGS_FILE)
@@ -875,8 +774,7 @@ def get_default_settings() -> Settings:
return Settings(
chat_model_provider=ModelProvider.OPENAI.name,
chat_model_name="gpt-4o",
- chat_model_temperature=0.0,
- chat_model_kwargs={},
+ chat_model_kwargs={ "temperature": "0" },
chat_model_ctx_length=120000,
chat_model_ctx_history=0.7,
chat_model_rl_requests=0,
@@ -884,10 +782,9 @@ def get_default_settings() -> Settings:
chat_model_rl_output=0,
util_model_provider=ModelProvider.OPENAI.name,
util_model_name="gpt-4o-mini",
- util_model_temperature=0.0,
util_model_ctx_length=120000,
util_model_ctx_input=0.7,
- util_model_kwargs={},
+ util_model_kwargs={ "temperature": "0" },
util_model_rl_requests=60,
util_model_rl_input=0,
util_model_rl_output=0,
@@ -899,8 +796,7 @@ def get_default_settings() -> Settings:
browser_model_provider=ModelProvider.OPENAI.name,
browser_model_name="gpt-4o",
browser_model_vision=False,
- browser_model_temperature=0.0,
- browser_model_kwargs={},
+ browser_model_kwargs={ "temperature": "0" },
api_keys={},
auth_login="",
auth_password="",
@@ -936,7 +832,9 @@ def _apply_settings():
agent = agent.get_data(agent.DATA_NAME_SUBORDINATE)
# reload whisper model if necessary
- task = defer.DeferredTask().start_task(whisper.preload, _settings["stt_model_size"]) #TODO overkill, replace with background task
+ task = defer.DeferredTask().start_task(
+ whisper.preload, _settings["stt_model_size"]
+ ) # TODO overkill, replace with background task
def _env_to_dict(data: str):