raise error when msg is invalid; fix docstr; improve ResponsiveAgent;…

… update doc and packaging; capture ipython output; find code blocks with llm when regex fails. (microsoft#1154) * autogen.agent -> autogen.agentchat * bug fix in portfolio * notebook * timeout * timeout * infer lang; close microsoft#1150 * timeout * message context * context handling * add sender to generate_reply * clean up the receive function * move mathchat to contrib * contrib * last_message * Add OptiGuide: agent and notebook * Optiguide notebook: add figures and URL 1. figures and code points to remote URL 2. simplify the prompt for the interpreter, because all information is already in the chat history. * Update name: Agent -> GenericAgent * Update notebook * Rename: GenericAgent -> ResponsiveAgent * Rebase to autogen.agentchat * OptiGuide: Comment, sytle, and notebook updates * simplify optiguide * raise error when msg is invalid; fix docstr * allow return None for generate_reply() * update_system_message * test update_system_message * simplify optiguide * simplify optiguide * simplify optiguide * simplify optiguide * move test * add test and fix bug * doc update * doc update * doc update * color * optiguide * prompt * test danger case * packaging * docker * remove path in traceback * capture ipython output * simplify * find code blocks with llm * find code with llm * order * order * fix bug in context handling * print executing msg * print executing msg * test find code * test find code * disable find_code * default_auto_reply * default auto reply * remove optiguide * remove -e --------- Co-authored-by: Beibin Li <beibin79@gmail.com>
whiskyboy · Aug 1, 2023 · 60f4da7 · 60f4da7
1 parent dba6ce5
commit 60f4da7
Show file tree

Hide file tree

Showing 39 changed files with 1,215 additions and 856 deletions.
diff --git a/.github/workflows/openai.yml b/.github/workflows/openai.yml
@@ -34,6 +34,10 @@ jobs:
           pip install -e .[autogen,blendsearch]
           python -c "import flaml"
           pip install coverage pytest datasets
+      - name: Install packages for test when needed
+        if: matrix.python-version == '3.9'
+        run: |
+          pip install docker
       - name: Install packages for MathChat when needed
         if: matrix.python-version != '3.11'
         run: |

diff --git a/README.md b/README.md
@@ -66,10 +66,10 @@ Use the following guides to get started with FLAML in .NET:
 * (New) The [autogen](https://microsoft.github.io/FLAML/docs/Use-Cases/Auto-Generation) package can help you maximize the utility out of the expensive LLMs such as ChatGPT and GPT-4, including:
     - A drop-in replacement of `openai.Completion` or `openai.ChatCompletion` with powerful functionalites like tuning, caching, templating, filtering. For example, you can optimize generations by LLM with your own tuning data, success metrics and budgets.
     ```python
-    from flaml import oai
+    from flaml import autogen
 
     # perform tuning
-    config, analysis = oai.Completion.tune(
+    config, analysis = autogen.Completion.tune(
         data=tune_data,
         metric="success",
         mode="max",
@@ -80,13 +80,13 @@ Use the following guides to get started with FLAML in .NET:
     )
 
     # perform inference for a test instance
-    response = oai.Completion.create(context=test_instance, **config)
+    response = autogen.Completion.create(context=test_instance, **config)
     ```
     - LLM-driven intelligent agents which can collaborately perform tasks autonomously or with human feedback, including tasks that require using tools via code.
     ```python
-    assistant = AssistantAgent("assistant")
-    user_proxy = UserProxyAgent("user_proxy")
-    user_proxy.initiate_chat("Show me the YTD gain of 10 largest technology companies as of today.")
+    assistant = autogen.AssistantAgent("assistant")
+    user_proxy = autogen.UserProxyAgent("user_proxy")
+    user_proxy.initiate_chat(assistant, message="Show me the YTD gain of 10 largest technology companies as of today.")
     ```
 * With three lines of code, you can start using this economical and fast
 AutoML engine as a [scikit-learn style estimator](https://microsoft.github.io/FLAML/docs/Use-Cases/Task-Oriented-AutoML).

diff --git a/flaml/__init__.py b/flaml/__init__.py
@@ -2,7 +2,6 @@
 from flaml.automl import AutoML, logger_formatter
 from flaml.tune.searcher import CFO, BlendSearch, FLOW2, BlendSearchTuner, RandomSearch
 from flaml.onlineml.autovw import AutoVW
-from flaml.autogen import oai
 from flaml.version import __version__
 
 

diff --git a/flaml/autogen/__init__.py b/flaml/autogen/__init__.py
@@ -1,2 +1,3 @@
-DEFAULT_MODEL = "gpt-4"
-FAST_MODEL = "gpt-3.5-turbo"
+from .oai import *
+from .agentchat import *
+from .code_utils import DEFAULT_MODEL, FAST_MODEL
diff --git a/flaml/autogen/agentchat/__init__.py b/flaml/autogen/agentchat/__init__.py
@@ -1,7 +1,6 @@
 from .agent import Agent
-from .responsive_agent import ResponsiveAgent
 from .assistant_agent import AssistantAgent
+from .responsive_agent import ResponsiveAgent
 from .user_proxy_agent import UserProxyAgent
 
-
 __all__ = ["Agent", "ResponsiveAgent", "AssistantAgent", "UserProxyAgent"]
diff --git a/flaml/autogen/agentchat/agent.py b/flaml/autogen/agentchat/agent.py
@@ -38,13 +38,13 @@ def generate_reply(
         messages: Optional[List[Dict]] = None,
         default_reply: Optional[Union[str, Dict]] = "",
         sender: Optional["Agent"] = None,
-    ) -> Union[str, Dict]:
+    ) -> Union[str, Dict, None]:
         """(Abstract method) Generate a reply based on the received messages.
 
         Args:
             messages (list[dict]): a list of messages received.
             default_reply (str or dict): the default reply if no other reply is generated.
             sender: sender of an Agent instance.
         Returns:
-            str or dict: the generated reply.
+            str or dict or None: the generated reply. If None, no reply is generated.
         """
diff --git a/flaml/autogen/agentchat/assistant_agent.py b/flaml/autogen/agentchat/assistant_agent.py
@@ -27,7 +27,7 @@ def __init__(
         self,
         name: str,
         system_message: Optional[str] = DEFAULT_SYSTEM_MESSAGE,
-        oai_config: Optional[Union[Dict, bool]] = None,
+        llm_config: Optional[Union[Dict, bool]] = None,
         is_termination_msg: Optional[Callable[[Dict], bool]] = None,
         max_consecutive_auto_reply: Optional[int] = None,
         human_input_mode: Optional[str] = "NEVER",
@@ -37,10 +37,10 @@ def __init__(
         """
         Args:
             name (str): agent name.
-            system_message (str): system message for the oai inference.
+            system_message (str): system message for the ChatCompletion inference.
                 Please override this attribute if you want to reprogram the agent.
-            oai_config (dict): oai inference configuration.
-                Please refer to [oai.Completion.create](/docs/reference/autogen/oai/completion#create)
+            llm_config (dict): llm inference configuration.
+                Please refer to [autogen.Completion.create](/docs/reference/autogen/oai/completion#create)
                 for available options.
             is_termination_msg (function): a function that takes a message in the form of a dictionary
                 and returns a boolean value indicating if this received message is a termination message.
@@ -58,6 +58,6 @@ def __init__(
             max_consecutive_auto_reply,
             human_input_mode,
             code_execution_config=code_execution_config,
-            oai_config=oai_config,
+            llm_config=llm_config,
             **kwargs,
         )
diff --git a/flaml/autogen/agentchat/contrib/math_user_proxy_agent.py b/flaml/autogen/agentchat/contrib/math_user_proxy_agent.py
@@ -136,6 +136,7 @@ def __init__(
             Callable[[Dict], bool]
         ] = _is_termination_msg_mathchat,  # terminate if \boxed{} in message
         human_input_mode: Optional[str] = "NEVER",  # Fully automated
+        default_auto_reply: Optional[Union[str, Dict, None]] = DEFAULT_REPLY,
         max_invalid_q_per_step=3,  # a parameter needed in MathChat
         **kwargs,
     ):
@@ -153,13 +154,15 @@ def __init__(
                     the number of auto reply reaches the max_consecutive_auto_reply.
                 (3) (Default) When "NEVER", the agent will never prompt for human input. Under this mode, the conversation stops
                     when the number of auto reply reaches the max_consecutive_auto_reply or when is_termination_msg is True.
+            default_auto_reply (str or dict or None): the default auto reply message when no code execution or llm based reply is generated.
             max_invalid_q_per_step (int): (ADDED) the maximum number of invalid queries per step.
             **kwargs (dict): other kwargs in [UserProxyAgent](user_proxy_agent#__init__).
         """
         super().__init__(
             name=name,
             is_termination_msg=is_termination_msg,
             human_input_mode=human_input_mode,
+            default_auto_reply=default_auto_reply,
             **kwargs,
         )
 
@@ -220,14 +223,6 @@ def execute_one_python_code(self, pycode):
         return_code, output, _ = execute_code(pycode, **self._code_execution_config, timeout=5)
         is_success = return_code == 0
 
-        # Decode the output
-        if isinstance(output, bytes):
-            try:
-                output = output.decode("utf-8")
-            except UnicodeDecodeError:
-                is_success = False
-                output = "The return cannot be decoded."
-
         if not is_success:
             # Remove the file information from the error string
             pattern = r'File "/[^"]+\.py", line \d+, in .+\n'
@@ -285,11 +280,11 @@ def generate_reply(
         self,
         messages: Optional[List[Dict]] = None,
         default_reply: Optional[Union[str, Dict]] = DEFAULT_REPLY,
-        sender: Optional["Agent"] = None,
-    ) -> Union[str, Dict]:
+        sender: Optional[Agent] = None,
+    ) -> Union[str, Dict, None]:
         """Generate an auto reply."""
         if messages is None:
-            messages = self._oai_conversations[sender.name]
+            messages = self._oai_messages[sender.name]
         message = messages[-1]
         message = message.get("content", "")
         code_blocks = extract_code(message)