Documentation: manage agent's memory (#675)

Co-authored-by: Albert Villanova del Moral <8515462+albertvillanova@users.noreply.github.com>

docs/source/en/_toctree.yml

@@ -14,6 +14,8 @@
     title: 🛠️ Tools - in-depth guide
   - local: tutorials/secure_code_execution
     title: 🛡️ Secure your code execution with E2B
+  - local: tutorials/memory
+    title: 📚 Manage your agent's memory
 - title: Conceptual guides
   sections:
   - local: conceptual_guides/intro_agents

docs/source/en/guided_tour.md

@@ -416,6 +416,17 @@ You can also use this `reset=False` argument to keep the conversation going in a
 
 ## Next steps
 
+Finally, when you've configured your agent to your needs, you can share it to the Hub!
+
+```py
+agent.push_to_hub("m-ric/my_agent")
+```
+
+Similarly, to load an agent that has been pushed to hub, if you trust the code from its tools, use:
+```py
+agent.from_hub("m-ric/my_agent", trust_remote_code=True)
+```
+
 For more in-depth usage, you will then want to check out our tutorials:
 - [the explanation of how our code agents work](./tutorials/secure_code_execution)
 - [this guide on how to build good agents](./tutorials/building_good_agents).

docs/source/en/tutorials/memory.md

@@ -0,0 +1,148 @@
+<!--Copyright 2024 The HuggingFace Team. 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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+# 📚 Manage your agent's memory
+
+[[open-in-colab]]
+
+In the end, an agent can be defined by simple components: it has tools, prompts.
+And most importantly, it has a memory of past steps, drawing a history of planning, execution, and errors.
+
+### Replay your agent's memory
+
+We propose several features to inspect a past agent run.
+
+You can instrument the agent's run to display it in a great UI that lets you zoom in/out on specific steps, as highlighted in the [instrumentation guide](./inspect_runs).
+
+You can also use `agent.replay()`, as follows:
+
+After the agent has run:
+```py
+from smolagents import HfApiModel, CodeAgent
+
+agent = CodeAgent(tools=[], model=HfApiModel(), verbosity_level=0)
+
+result = agent.run("What's the 20th Fibonacci number?")
+```
+
+If you want to replay this last run, just use:
+```py
+agent.replay()
+```
+
+### Dynamically change the agent's memory
+
+Many advanced use cases require dynamic modification of the agent's memory.
+
+You can access the agent's memory using:
+
+```py
+from smolagents import ActionStep
+
+system_prompt_step = agent.memory.system_prompt
+print("The system prompt given to the agent was:")
+print(system_prompt_step.system_prompt)
+
+task_step = agent.memory.steps[0]
+print("\n\nThe first task step was:")
+print(task_step.task)
+
+for step in agent.memory.steps:
+    if isinstance(step, ActionStep):
+        if step.error is not None:
+            print(f"\nStep {step.step_number} got this error:\n{step.error}\n")
+        else:
+            print(f"\nStep {step.step_number} got these observations:\n{step.observations}\n")
+```
+
+Use `agent.memory.get_full_steps()` to get full steps as dictionaries.
+
+You can also use step callbacks to dynamically change the agent's memory.
+
+Step callbacks can access the `agent` itself in their arguments, so they can access any memory step as highlighted above, and change it if needed. For instance, let's say you are observing screenshots of each step performed by a web browser agent. You want to log the newest screenshot, and remove the images from ancient steps to save on token costs.
+
+You culd run something like the following.
+_Note: this code is incomplete, some imports and object definitions have been removed for the sake of concision, visit [the original script](https://github.com/huggingface/smolagents/blob/main/src/smolagents/vision_web_browser.py) to get the full working code._
+
+```py
+import helium
+from PIL import Image
+from io import BytesIO
+from time import sleep
+
+def update_screenshot(memory_step: ActionStep, agent: CodeAgent) -> None:
+    sleep(1.0)  # Let JavaScript animations happen before taking the screenshot
+    driver = helium.get_driver()
+    latest_step = memory_step.step_number
+    for previous_memory_step in agent.memory.steps:  # Remove previous screenshots from logs for lean processing
+        if isinstance(previous_memory_step, ActionStep) and previous_memory_step.step_number <= latest_step - 2:
+            previous_memory_step.observations_images = None
+    png_bytes = driver.get_screenshot_as_png()
+    image = Image.open(BytesIO(png_bytes))
+    memory_step.observations_images = [image.copy()]
+```
+
+Then you should pass this function in the `step_callbacks` argument upon initialization of your agent:
+
+```py
+CodeAgent(
+    tools=[DuckDuckGoSearchTool(), go_back, close_popups, search_item_ctrl_f],
+    model=model,
+    additional_authorized_imports=["helium"],
+    step_callbacks=[update_screenshot],
+    max_steps=20,
+    verbosity_level=2,
+)
+```
+
+Head to our [vision web browser code](https://github.com/huggingface/smolagents/blob/main/src/smolagents/vision_web_browser.py) to see the full working example.
+
+### Run agents one step at a time
+
+This can be useful in case you have tool calls that take days: you can just run your agents step by step.
+This will also let you update the memory on each step.
+
+```py
+from smolagents import HfApiModel, CodeAgent, ActionStep, TaskStep
+
+agent = CodeAgent(tools=[], model=HfApiModel(), verbosity_level=1)
+print(agent.memory.system_prompt)
+
+task = "What is the 20th Fibonacci number?"
+
+# You could modify the memory as needed here by inputting the memory of another agent.
+# agent.memory.steps = previous_agent.memory.steps
+
+# Let's start a new task!
+agent.memory.steps.append(TaskStep(task=task, task_images=[]))
+
+final_answer = None
+step_number = 1
+while final_answer is None and step_number <= 10:
+    memory_step = ActionStep(
+        step_number=step_number,
+        observations_images=[],
+    )
+    # Run one step.
+    final_answer = agent.step(memory_step)
+    agent.memory.steps.append(memory_step)
+    step_number += 1
+
+    # Change the memory as you please!
+    # For instance to update the latest step:
+    # agent.memory.steps[-1] = ...
+
+print("The final answer is:", final_answer)
+```

src/smolagents/memory.py

@@ -3,7 +3,7 @@ from logging import getLogger
 from typing import TYPE_CHECKING, Any, Dict, List, TypedDict, Union
 
 from smolagents.models import ChatMessage, MessageRole
-from smolagents.monitoring import AgentLogger
+from smolagents.monitoring import AgentLogger, LogLevel
 from smolagents.utils import AgentError, make_json_serializable
 
 
@@ -216,19 +216,19 @@ class AgentMemory:
         logger.console.log("Replaying the agent's steps:")
         for step in self.steps:
             if isinstance(step, SystemPromptStep) and detailed:
-                logger.log_markdown(title="System prompt", content=step.system_prompt)
+                logger.log_markdown(title="System prompt", content=step.system_prompt, level=LogLevel.ERROR)
             elif isinstance(step, TaskStep):
-                logger.log_task(step.task, "", 2)
+                logger.log_task(step.task, "", level=LogLevel.ERROR)
             elif isinstance(step, ActionStep):
-                logger.log_rule(f"Step {step.step_number}")
+                logger.log_rule(f"Step {step.step_number}", level=LogLevel.ERROR)
                 if detailed:
                     logger.log_messages(step.model_input_messages)
-                logger.log_markdown(title="Agent output:", content=step.model_output)
+                logger.log_markdown(title="Agent output:", content=step.model_output, level=LogLevel.ERROR)
             elif isinstance(step, PlanningStep):
-                logger.log_rule("Planning step")
+                logger.log_rule("Planning step", level=LogLevel.ERROR)
                 if detailed:
-                    logger.log_messages(step.model_input_messages)
+                    logger.log_messages(step.model_input_messages, level=LogLevel.ERROR)
-                logger.log_markdown(title="Agent output:", content=step.facts + "\n" + step.plan)
+                logger.log_markdown(title="Agent output:", content=step.facts + "\n" + step.plan, level=LogLevel.ERROR)
 
 
 __all__ = ["AgentMemory"]

tests/test_agents.py

@@ -490,6 +490,20 @@ class AgentTests(unittest.TestCase):
         str_output = capture.get()
         assert "`additional_authorized_imports`" in str_output.replace("\n", "")
 
+    def test_replay_shows_logs(self):
+        agent = CodeAgent(
+            tools=[], model=fake_code_model_import, verbosity_level=0, additional_authorized_imports=["numpy"]
+        )
+        agent.run("Count to 3")
+
+        with agent.logger.console.capture() as capture:
+            agent.replay()
+        str_output = capture.get().replace("\n", "")
+        assert "New run" in str_output
+        assert "Agent output:" in str_output
+        assert 'final_answer("got' in str_output
+        assert "```<end_code>" in str_output
+
     def test_code_nontrivial_final_answer_works(self):
         def fake_code_model_final_answer(messages, stop_sequences=None, grammar=None):
             return ChatMessage(