Skip to content

armgit / camel

Public Imported from GitHub
0

🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org

feat: add summarize function in agent using context util (#3149)

Xiaotian Jinaf318018e015
diff --git a/camel/agents/chat_agent.py b/camel/agents/chat_agent.py
index a9ea5d2aa..011cd2df1 100644
--- a/camel/agents/chat_agent.py
+++ b/camel/agents/chat_agent.py
@@ -23,6 +23,7 @@ import textwrap
import threading
import time
import uuid
+from datetime import datetime
from pathlib import Path
from typing import (
TYPE_CHECKING,
@@ -93,6 +94,7 @@ from camel.utils import (
model_from_json_schema,
)
from camel.utils.commons import dependencies_required
+from camel.utils.context_utils import ContextUtility
if TYPE_CHECKING:
from camel.terminators import ResponseTerminator
@@ -535,6 +537,8 @@ class ChatAgent(BaseAgent):
self.retry_attempts = max(1, retry_attempts)
self.retry_delay = max(0.0, retry_delay)
self.step_timeout = step_timeout
+ self._context_utility: Optional[ContextUtility] = None
+ self._context_summary_agent: Optional["ChatAgent"] = None
self.stream_accumulate = stream_accumulate
def reset(self):
@@ -1042,6 +1046,149 @@ class ChatAgent(BaseAgent):
json_store.save(to_save)
logger.info(f"Memory saved to {path}")
+ def summarize(
+ self,
+ filename: Optional[str] = None,
+ summary_prompt: Optional[str] = None,
+ ) -> Dict[str, Any]:
+ r"""Summarize the agent's current conversation context and persist it
+ to a markdown file.
+
+ Args:
+ filename (Optional[str]): The base filename (without extension) to
+ use for the markdown file. Defaults to a timestamped name when
+ not provided.
+ summary_prompt (Optional[str]): Custom prompt for the summarizer.
+ When omitted, a default prompt highlighting key decisions,
+ action items, and open questions is used.
+
+ Returns:
+ Dict[str, Any]: A dictionary containing the summary text, file
+ path, and status message.
+ """
+
+ result: Dict[str, Any] = {
+ "summary": "",
+ "file_path": None,
+ "status": "",
+ }
+
+ try:
+ if self._context_utility is None:
+ self._context_utility = ContextUtility()
+
+ # Get conversation directly from agent's memory
+ messages, _ = self.memory.get_context()
+
+ if not messages:
+ status_message = (
+ "No conversation context available to summarize."
+ )
+ result["status"] = status_message
+ return result
+
+ # Convert messages to conversation text
+ conversation_lines = []
+ for message in messages:
+ role = message.get('role', 'unknown')
+ content = message.get('content', '')
+ if content:
+ conversation_lines.append(f"{role}: {content}")
+
+ conversation_text = "\n".join(conversation_lines).strip()
+
+ if not conversation_text:
+ status_message = (
+ "Conversation context is empty; skipping summary."
+ )
+ result["status"] = status_message
+ return result
+
+ if self._context_summary_agent is None:
+ self._context_summary_agent = ChatAgent(
+ system_message=(
+ "You are a helpful assistant that summarizes "
+ "conversations into concise markdown bullet lists."
+ ),
+ model=self.model_backend,
+ agent_id=f"{self.agent_id}_context_summarizer",
+ )
+ else:
+ self._context_summary_agent.reset()
+
+ prompt_text = summary_prompt or (
+ "Summarize the following conversation in concise markdown "
+ "bullet points highlighting key decisions, action items, and "
+ "open questions.\n\n"
+ f"{conversation_text}"
+ )
+
+ try:
+ response = self._context_summary_agent.step(prompt_text)
+ except Exception as step_exc:
+ error_message = (
+ f"Failed to generate summary using model: {step_exc}"
+ )
+ logger.error(error_message)
+ result["status"] = error_message
+ return result
+
+ if not response.msgs:
+ status_message = (
+ "Failed to generate summary from model response."
+ )
+ result["status"] = status_message
+ return result
+
+ summary_content = response.msgs[-1].content.strip()
+ if not summary_content:
+ status_message = "Generated summary is empty."
+ result["status"] = status_message
+ return result
+
+ base_filename = (
+ filename
+ if filename
+ else f"context_summary_{datetime.now().strftime('%Y%m%d_%H%M%S')}" # noqa: E501
+ )
+ base_filename = Path(base_filename).with_suffix("").name
+
+ metadata = self._context_utility.get_session_metadata()
+ metadata.update(
+ {
+ "agent_id": self.agent_id,
+ "message_count": len(messages),
+ }
+ )
+
+ save_status = self._context_utility.save_markdown_file(
+ base_filename,
+ summary_content,
+ title="Conversation Summary",
+ metadata=metadata,
+ )
+
+ file_path = (
+ self._context_utility.get_working_directory()
+ / f"{base_filename}.md"
+ )
+
+ result.update(
+ {
+ "summary": summary_content,
+ "file_path": str(file_path),
+ "status": save_status,
+ }
+ )
+ logger.info("Conversation summary saved to %s", file_path)
+ return result
+
+ except Exception as exc:
+ error_message = f"Failed to summarize conversation context: {exc}"
+ logger.error(error_message)
+ result["status"] = error_message
+ return result
+
def clear_memory(self) -> None:
r"""Clear the agent's memory and reset to initial state.
diff --git a/examples/agents/agent_summarize.py b/examples/agents/agent_summarize.py
new file mode 100644
index 000000000..92d0352eb
--- /dev/null
+++ b/examples/agents/agent_summarize.py
@@ -0,0 +1,117 @@
+# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+# 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.
+# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+"""Example showing how to summarize an agent conversation to markdown."""
+
+from __future__ import annotations
+
+from camel.agents import ChatAgent
+from camel.models import ModelFactory
+from camel.toolkits import FunctionTool
+from camel.types import ModelPlatformType, ModelType
+
+
+def calculate_add(a: int, b: int) -> int:
+ """Add two integers and return the result."""
+
+ return a + b
+
+
+def calculate_multiply(a: int, b: int) -> int:
+ """Multiply two integers and return the product."""
+
+ return a * b
+
+
+def main() -> None:
+ """Run a short interaction and persist the summary to markdown."""
+
+ system_message = (
+ "You are a helpful assistant that can call math tools when needed."
+ )
+
+ model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ )
+
+ add_tool = FunctionTool(calculate_add)
+ multiply_tool = FunctionTool(calculate_multiply)
+
+ agent = ChatAgent(
+ system_message=system_message,
+ model=model,
+ tools=[add_tool, multiply_tool],
+ )
+
+ # Interact with the agent so it has context to summarize.
+ agent.step("We are preparing a math practice worksheet for students.")
+
+ agent.step(
+ "Compute 27 + 15 using the add tool to double-check the answer."
+ )
+
+ agent.step("Now multiply the result by 3.")
+
+ # Summarize the conversation to a markdown file in the context directory.
+ summary_result = agent.summarize(filename="math_planning_summary")
+
+ print(summary_result.get("status", ""))
+ if summary_result.get("file_path"):
+ print(f"Summary saved to: {summary_result['file_path']}")
+
+ if summary_result.get("summary"):
+ print("Generated summary:\n")
+ print(summary_result["summary"])
+
+
+'''
+Markdown file 'math_planning_summary.md' saved successfully
+Summary saved to: context_files/session_20250918_192642_141905/math_planning_summary.md
+Generated summary:
+
+- Context
+ - User is preparing a math practice worksheet for students.
+ - Assistant offered to tailor the worksheet and asked clarifying questions (grade/age, topics, number of problems, answer key vs step-by-step, difficulty, time/layout).
+
+- Key decisions made in conversation
+ - Assistant provided a ready-to-use sample worksheet for Grade 6 (mixed skills) with 12 problems and a brief answer key.
+ - User requested numerical checks using tools for arithmetic steps.
+
+- Actions performed (completed)
+ - Assistant used tools to verify calculations:
+ - calculate_add(27, 15) → 42 (reported: 27 + 15 = 42)
+ - calculate_multiply(42, 3) → 126 (reported: 42 * 3 = 126)
+
+- Pending action items (who can/should act)
+ - Assistant: prepare a customized worksheet based on user preferences (grade/topic/number of problems/difficulty).
+ - Assistant: format the worksheet as a printable PDF or Google Doc if requested.
+ - Assistant: provide full step-by-step solutions if the user wants them.
+ - User: specify preferences (see open questions) so assistant can generate the customized worksheet.
+
+- Open questions / user choices to confirm
+ - What grade or age range should the worksheet target?
+ - Which topics to include (arithmetic, fractions, decimals, ratios, pre-algebra, algebra, geometry, word problems, etc.)?
+ - How many problems total?
+ - Answer key only or full step-by-step solutions?
+ - Preferred difficulty level (review, mixed, challenge)?
+ - Any time limit, point values, or layout/format requirements?
+ - Do you want the worksheet as a PDF or Google Doc?
+
+- Next suggested step
+ - User provides the preferences above and/or confirms whether to use the provided Grade 6 sample; assistant will generate the customized worksheet (with chosen format and solution detail).
+''' # noqa: E501
+
+
+if __name__ == "__main__":
+ main()