feat: Introduce LLM context compaction interface

Provide a more efficient way to compact LLM context for better agentic performance.

* `app`: the top level abstraction for an ADK application. It contains an root agent, and plugins.
* `content_strategy`: the abstraction for selecting the contents for LLM request.
* `compaction_strategy`: the abstraction for compacting the events.
* Added `sequence_id` and `summary_range` in event class.

PiperOrigin-RevId: 808634224
This commit is contained in:
Hangfei Lin
2025-09-18 10:13:34 -07:00
committed by Copybara-Service
parent e86647d446
commit c37bd2742c
3 changed files with 75 additions and 2 deletions
+4 -1
View File
@@ -13,7 +13,6 @@
# limitations under the License.
from __future__ import annotations
from abc import ABC
from typing import Optional
from pydantic import BaseModel
@@ -21,6 +20,7 @@ from pydantic import ConfigDict
from pydantic import Field
from ..agents.base_agent import BaseAgent
from ..apps.base_events_compactor import BaseEventsCompactor
from ..plugins.base_plugin import BasePlugin
from ..utils.feature_decorator import experimental
@@ -50,3 +50,6 @@ class App(BaseModel):
plugins: list[BasePlugin] = Field(default_factory=list)
"""The plugins in the application."""
event_compactor: Optional[BaseEventsCompactor] = None
"""The event compactor strategy for the application."""
@@ -0,0 +1,50 @@
# Copyright 2025 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
import abc
from typing import Optional
from google.genai.types import Content
from ..events.event import Event
from ..utils.feature_decorator import experimental
@experimental
class BaseEventsCompactor(abc.ABC):
"""Base interface for compacting events."""
async def maybe_compact_events(
self, *, events: list[Event]
) -> Optional[Content]:
"""A list of uncompacted events, decide whether to compact.
If no need to compact, return None. Otherwise, compact into a content and
return it.
This method will summarize the events and return a new summray event
indicating the range of events it summarized.
When sending events to the LLM, if a summary event is present, the events it
replaces (those identified in itssummary_range) should not be included.
Args:
events: Events to compact.
agent_name: The name of the agent.
Returns:
The new compacted content, or None if no compaction is needed.
"""
raise NotImplementedError()
+21 -1
View File
@@ -14,9 +14,9 @@
from __future__ import annotations
from typing import Any
from typing import Optional
from google.genai.types import Content
from pydantic import alias_generators
from pydantic import BaseModel
from pydantic import ConfigDict
@@ -26,6 +26,23 @@ from ..auth.auth_tool import AuthConfig
from ..tools.tool_confirmation import ToolConfirmation
class EventCompaction(BaseModel):
"""The compaction of the events."""
model_config = ConfigDict(
extra='forbid',
alias_generator=alias_generators.to_camel,
populate_by_name=True,
)
"""The pydantic model config."""
compaction_range: Optional[tuple[float, float]] = None
"""The sequence ID range of the events that are summarized, in the form(start_sequence_id, end_sequence_id)`"""
compacted_content: Content
"""The summarized content of the events."""
class EventActions(BaseModel):
"""Represents the actions attached to an event."""
@@ -72,3 +89,6 @@ class EventActions(BaseModel):
)
"""A dict of tool confirmation requested by this event, keyed by
function call id."""
compaction: Optional[EventCompaction] = None
"""The compaction of the events."""