-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathbase_agent.py
More file actions
196 lines (153 loc) · 5.65 KB
/
Copy pathbase_agent.py
File metadata and controls
196 lines (153 loc) · 5.65 KB
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
from typing import Optional
"""
BaseAgent: Abstract base class for all Sentience agents
Defines the interface that all agent implementations must follow
"""
from abc import ABC, abstractmethod
from .models import ActionHistory, AgentActionResult, Element, Snapshot, TokenStats
class BaseAgent(ABC):
"""
Abstract base class for all Sentience agents.
Provides a standard interface for:
- Executing natural language goals (act)
- Tracking execution history
- Monitoring token usage
- Filtering elements based on goals
Subclasses must implement:
- act(): Execute a natural language goal
- get_history(): Return execution history
- get_token_stats(): Return token usage statistics
- clear_history(): Reset history and token counters
Subclasses can override:
- filter_elements(): Customize element filtering logic
"""
@abstractmethod
def act(self, goal: str, **kwargs) -> AgentActionResult:
"""
Execute a natural language goal using the agent.
Args:
goal: Natural language instruction (e.g., "Click the login button")
**kwargs: Additional parameters (implementation-specific)
Returns:
AgentActionResult with execution details
Raises:
RuntimeError: If execution fails after retries
"""
pass
@abstractmethod
def get_history(self) -> list[ActionHistory]:
"""
Get the execution history of all actions taken.
Returns:
List of ActionHistory entries
"""
pass
@abstractmethod
def get_token_stats(self) -> TokenStats:
"""
Get token usage statistics for the agent session.
Returns:
TokenStats with cumulative token counts
"""
pass
@abstractmethod
def clear_history(self) -> None:
"""
Clear execution history and reset token counters.
This resets the agent to a clean state.
"""
pass
def filter_elements(self, snapshot: Snapshot, goal: str | None = None) -> list[Element]:
"""
Filter elements from a snapshot based on goal context.
Default implementation returns all elements unchanged.
Subclasses can override to implement custom filtering logic
such as:
- Removing irrelevant elements based on goal keywords
- Boosting importance of matching elements
- Filtering by role, size, or visual properties
Args:
snapshot: Current page snapshot
goal: User's goal (can inform filtering strategy)
Returns:
Filtered list of elements (default: all elements)
Example:
>>> agent = SentienceAgent(browser, llm)
>>> snap = snapshot(browser)
>>> filtered = agent.filter_elements(snap, goal="Click login")
>>> # filtered now contains only relevant elements
"""
return snapshot.elements
class BaseAgentAsync(ABC):
"""
Abstract base class for all async Sentience agents.
Provides a standard interface for:
- Executing natural language goals (act)
- Tracking execution history
- Monitoring token usage
- Filtering elements based on goals
Subclasses must implement:
- act(): Execute a natural language goal (async)
- get_history(): Return execution history
- get_token_stats(): Return token usage statistics
- clear_history(): Reset history and token counters
Subclasses can override:
- filter_elements(): Customize element filtering logic
"""
@abstractmethod
async def act(self, goal: str, **kwargs) -> AgentActionResult:
"""
Execute a natural language goal using the agent (async).
Args:
goal: Natural language instruction (e.g., "Click the login button")
**kwargs: Additional parameters (implementation-specific)
Returns:
AgentActionResult with execution details
Raises:
RuntimeError: If execution fails after retries
"""
pass
@abstractmethod
def get_history(self) -> list[ActionHistory]:
"""
Get the execution history of all actions taken.
Returns:
List of ActionHistory entries
"""
pass
@abstractmethod
def get_token_stats(self) -> TokenStats:
"""
Get token usage statistics for the agent session.
Returns:
TokenStats with cumulative token counts
"""
pass
@abstractmethod
def clear_history(self) -> None:
"""
Clear execution history and reset token counters.
This resets the agent to a clean state.
"""
pass
def filter_elements(self, snapshot: Snapshot, goal: str | None = None) -> list[Element]:
"""
Filter elements from a snapshot based on goal context.
Default implementation returns all elements unchanged.
Subclasses can override to implement custom filtering logic
such as:
- Removing irrelevant elements based on goal keywords
- Boosting importance of matching elements
- Filtering by role, size, or visual properties
Args:
snapshot: Current page snapshot
goal: User's goal (can inform filtering strategy)
Returns:
Filtered list of elements (default: all elements)
Example:
>>> agent = SentienceAgentAsync(browser, llm)
>>> snap = await snapshot_async(browser)
>>> filtered = agent.filter_elements(snap, goal="Click login")
>>> # filtered now contains only relevant elements
"""
return snapshot.elements