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
|
import logging
import platform
import time
import distro
from autogpt.core.configuration import (
Configurable,
SystemConfiguration,
SystemSettings,
UserConfigurable,
)
from autogpt.core.planning import strategies
from autogpt.core.planning.base import PromptStrategy
from autogpt.core.planning.schema import (
LanguageModelClassification,
LanguageModelResponse,
Task,
)
from autogpt.core.resource.model_providers import (
LanguageModelProvider,
ModelProviderName,
OpenAIModelName,
)
from autogpt.core.workspace import Workspace
class LanguageModelConfiguration(SystemConfiguration):
"""Struct for model configuration."""
model_name: str = UserConfigurable()
provider_name: ModelProviderName = UserConfigurable()
temperature: float = UserConfigurable()
class PromptStrategiesConfiguration(SystemConfiguration):
name_and_goals: strategies.NameAndGoalsConfiguration
initial_plan: strategies.InitialPlanConfiguration
next_ability: strategies.NextAbilityConfiguration
class PlannerConfiguration(SystemConfiguration):
"""Configuration for the Planner subsystem."""
models: dict[LanguageModelClassification, LanguageModelConfiguration]
prompt_strategies: PromptStrategiesConfiguration
class PlannerSettings(SystemSettings):
"""Settings for the Planner subsystem."""
configuration: PlannerConfiguration
class SimplePlanner(Configurable):
"""Manages the agent's planning and goal-setting by constructing language model prompts."""
default_settings = PlannerSettings(
name="planner",
description="Manages the agent's planning and goal-setting by constructing language model prompts.",
configuration=PlannerConfiguration(
models={
LanguageModelClassification.FAST_MODEL: LanguageModelConfiguration(
model_name=OpenAIModelName.GPT3,
provider_name=ModelProviderName.OPENAI,
temperature=0.9,
),
LanguageModelClassification.SMART_MODEL: LanguageModelConfiguration(
model_name=OpenAIModelName.GPT4,
provider_name=ModelProviderName.OPENAI,
temperature=0.9,
),
},
prompt_strategies=PromptStrategiesConfiguration(
name_and_goals=strategies.NameAndGoals.default_configuration,
initial_plan=strategies.InitialPlan.default_configuration,
next_ability=strategies.NextAbility.default_configuration,
),
),
)
def __init__(
self,
settings: PlannerSettings,
logger: logging.Logger,
model_providers: dict[ModelProviderName, LanguageModelProvider],
workspace: Workspace = None, # Workspace is not available during bootstrapping.
) -> None:
self._configuration = settings.configuration
self._logger = logger
self._workspace = workspace
self._providers: dict[LanguageModelClassification, LanguageModelProvider] = {}
for model, model_config in self._configuration.models.items():
self._providers[model] = model_providers[model_config.provider_name]
self._prompt_strategies = {
"name_and_goals": strategies.NameAndGoals(
**self._configuration.prompt_strategies.name_and_goals.dict()
),
"initial_plan": strategies.InitialPlan(
**self._configuration.prompt_strategies.initial_plan.dict()
),
"next_ability": strategies.NextAbility(
**self._configuration.prompt_strategies.next_ability.dict()
),
}
async def decide_name_and_goals(self, user_objective: str) -> LanguageModelResponse:
return await self.chat_with_model(
self._prompt_strategies["name_and_goals"],
user_objective=user_objective,
)
async def make_initial_plan(
self,
agent_name: str,
agent_role: str,
agent_goals: list[str],
abilities: list[str],
) -> LanguageModelResponse:
return await self.chat_with_model(
self._prompt_strategies["initial_plan"],
agent_name=agent_name,
agent_role=agent_role,
agent_goals=agent_goals,
abilities=abilities,
)
async def determine_next_ability(
self,
task: Task,
ability_schema: list[dict],
):
return await self.chat_with_model(
self._prompt_strategies["next_ability"],
task=task,
ability_schema=ability_schema,
)
async def chat_with_model(
self,
prompt_strategy: PromptStrategy,
**kwargs,
) -> LanguageModelResponse:
model_classification = prompt_strategy.model_classification
model_configuration = self._configuration.models[model_classification].dict()
self._logger.debug(f"Using model configuration: {model_configuration}")
del model_configuration["provider_name"]
provider = self._providers[model_classification]
template_kwargs = self._make_template_kwargs_for_strategy(prompt_strategy)
template_kwargs.update(kwargs)
prompt = prompt_strategy.build_prompt(**template_kwargs)
self._logger.debug(f"Using prompt:\n{prompt}\n\n")
response = await provider.create_language_completion(
model_prompt=prompt.messages,
functions=prompt.functions,
**model_configuration,
completion_parser=prompt_strategy.parse_response_content,
)
return LanguageModelResponse.parse_obj(response.dict())
def _make_template_kwargs_for_strategy(self, strategy: PromptStrategy):
provider = self._providers[strategy.model_classification]
template_kwargs = {
"os_info": get_os_info(),
"api_budget": provider.get_remaining_budget(),
"current_time": time.strftime("%c"),
}
return template_kwargs
def get_os_info() -> str:
os_name = platform.system()
os_info = (
platform.platform(terse=True)
if os_name != "Linux"
else distro.name(pretty=True)
)
return os_info
|
