Merge branch 'main' into main

.gitignore

@@ -8,7 +8,7 @@ node_modules/
 *.log
 
 # Python virtualenv
-.venv
+.venv*
 
 # Byte-compiled / optimized / DLL files
 __pycache__/

README.md

@@ -58,7 +58,15 @@ The easiest way to start playing is
  2. Copy OAI_CONFIG_LIST_sample to ./notebook folder, name to OAI_CONFIG_LIST, and set the correct configuration.
  3. Start playing with the notebooks!
 
+## Using existing docker image
+Install docker, save your oai key into an environment variable name OPENAI_API_KEY, and then run the following.
 
+```
+docker pull yuandongtian/autogen:latest
+docker run -it -e OPENAI_API_KEY=$OPENAI_API_KEY -p 8081:8081 docker.io/yuandongtian/autogen:latest
+```
+
+Then open `http://localhost:8081/` in your browser to use AutoGen. The UI is from `./samples/apps/autogen-assistant`. See docker hub [link](https://hub.docker.com/r/yuandongtian/autogen) for more details.
 
 ## Installation
 

autogen/_pydantic.py

@@ -0,0 +1,110 @@
+from typing import Any, Dict, Optional, Tuple, Type, Union, get_args
+
+from pydantic import BaseModel
+from pydantic.version import VERSION as PYDANTIC_VERSION
+from typing_extensions import get_origin
+
+__all__ = ("JsonSchemaValue", "model_dump", "model_dump_json", "type2schema")
+
+PYDANTIC_V1 = PYDANTIC_VERSION.startswith("1.")
+
+if not PYDANTIC_V1:
+    from pydantic import TypeAdapter
+    from pydantic._internal._typing_extra import eval_type_lenient as evaluate_forwardref
+    from pydantic.json_schema import JsonSchemaValue
+
+    def type2schema(t: Optional[Type]) -> JsonSchemaValue:
+        """Convert a type to a JSON schema
+
+        Args:
+            t (Type): The type to convert
+
+        Returns:
+            JsonSchemaValue: The JSON schema
+        """
+        return TypeAdapter(t).json_schema()
+
+    def model_dump(model: BaseModel) -> Dict[str, Any]:
+        """Convert a pydantic model to a dict
+
+        Args:
+            model (BaseModel): The model to convert
+
+        Returns:
+            Dict[str, Any]: The dict representation of the model
+
+        """
+        return model.model_dump()
+
+    def model_dump_json(model: BaseModel) -> str:
+        """Convert a pydantic model to a JSON string
+
+        Args:
+            model (BaseModel): The model to convert
+
+        Returns:
+            str: The JSON string representation of the model
+        """
+        return model.model_dump_json()
+
+
+# Remove this once we drop support for pydantic 1.x
+else:  # pragma: no cover
+    from pydantic import schema_of
+    from pydantic.typing import evaluate_forwardref as evaluate_forwardref
+
+    JsonSchemaValue = Dict[str, Any]
+
+    def type2schema(t: Optional[Type]) -> JsonSchemaValue:
+        """Convert a type to a JSON schema
+
+        Args:
+            t (Type): The type to convert
+
+        Returns:
+            JsonSchemaValue: The JSON schema
+        """
+        if PYDANTIC_V1:
+            if t is None:
+                return {"type": "null"}
+            elif get_origin(t) is Union:
+                return {"anyOf": [type2schema(tt) for tt in get_args(t)]}
+            elif get_origin(t) in [Tuple, tuple]:
+                prefixItems = [type2schema(tt) for tt in get_args(t)]
+                return {
+                    "maxItems": len(prefixItems),
+                    "minItems": len(prefixItems),
+                    "prefixItems": prefixItems,
+                    "type": "array",
+                }
+
+        d = schema_of(t)
+        if "title" in d:
+            d.pop("title")
+        if "description" in d:
+            d.pop("description")
+
+        return d
+
+    def model_dump(model: BaseModel) -> Dict[str, Any]:
+        """Convert a pydantic model to a dict
+
+        Args:
+            model (BaseModel): The model to convert
+
+        Returns:
+            Dict[str, Any]: The dict representation of the model
+
+        """
+        return model.dict()
+
+    def model_dump_json(model: BaseModel) -> str:
+        """Convert a pydantic model to a JSON string
+
+        Args:
+            model (BaseModel): The model to convert
+
+        Returns:
+            str: The JSON string representation of the model
+        """
+        return model.json()

autogen/agentchat/assistant_agent.py

@@ -70,7 +70,7 @@ def __init__(
             **kwargs,
         )
 
-        # Update the provided desciption if None, and we are using the default system_message,
+        # Update the provided description if None, and we are using the default system_message,
         # then use the default description.
         if description is None:
             if system_message == self.DEFAULT_SYSTEM_MESSAGE:

autogen/agentchat/contrib/compressible_agent.py

@@ -225,7 +225,7 @@ def _manage_history_on_token_limit(self, messages, token_used, max_token_allowed
         # 1. mode = "TERMINATE", terminate the agent if no token left.
         if self.compress_config["mode"] == "TERMINATE":
             if max_token_allowed - token_used <= 0:
-                # Teminate if no token left.
+                # Terminate if no token left.
                 print(
                     colored(
                         f'Warning: Terminate Agent "{self.name}" due to no token left for oai reply. max token for {model}: {max_token_allowed}, existing token count: {token_used}',
@@ -320,7 +320,7 @@ def on_oai_token_limit(
                         cmsg["role"] = "user"
                     sender._oai_messages[self][i] = cmsg
 
-            # sucessfully compressed, return False, None for generate_oai_reply to be called with the updated messages
+            # successfully compressed, return False, None for generate_oai_reply to be called with the updated messages
             return False, None
         return final, None
 
@@ -332,7 +332,7 @@ def compress_messages(
         """Compress a list of messages into one message.
 
         The first message (the initial prompt) will not be compressed.
-        The rest of the messages will be compressed into one message, the model is asked to distinuish the role of each message: USER, ASSISTANT, FUNCTION_CALL, FUNCTION_RETURN.
+        The rest of the messages will be compressed into one message, the model is asked to distinguish the role of each message: USER, ASSISTANT, FUNCTION_CALL, FUNCTION_RETURN.
         Check out the compress_sys_msg.
 
         TODO: model used in compression agent is different from assistant agent: For example, if original model used by is gpt-4; we start compressing at 70% of usage, 70% of 8092 = 5664; and we use gpt 3.5 here max_toke = 4096, it will raise error. choosinng model automatically?

autogen/agentchat/contrib/gpt_assistant_agent.py

@@ -203,7 +203,7 @@ def _get_run_response(self, thread, run):
                                 new_messages.append(
                                     {
                                         "role": msg.role,
-                                        "content": f"Recieved file id={content.image_file.file_id}",
+                                        "content": f"Received file id={content.image_file.file_id}",
                                     }
                                 )
                 return new_messages
@@ -219,7 +219,7 @@ def _get_run_response(self, thread, run):
                     }
 
                     logger.info(
-                        "Intermediate executing(%s, Sucess: %s) : %s",
+                        "Intermediate executing(%s, Success: %s) : %s",
                         tool_response["name"],
                         is_exec_success,
                         tool_response["content"],

autogen/agentchat/contrib/img_utils.py

@@ -26,7 +26,7 @@ def get_image_data(image_file: str, use_b64=True) -> bytes:
         return content
 
 
-def llava_formater(prompt: str, order_image_tokens: bool = False) -> Tuple[str, List[str]]:
+def llava_formatter(prompt: str, order_image_tokens: bool = False) -> Tuple[str, List[str]]:
     """
     Formats the input prompt by replacing image tags and returns the new prompt along with image locations.
 

autogen/agentchat/contrib/llava_agent.py

@@ -10,7 +10,7 @@
 from regex import R
 
 from autogen.agentchat.agent import Agent
-from autogen.agentchat.contrib.img_utils import get_image_data, llava_formater
+from autogen.agentchat.contrib.img_utils import get_image_data, llava_formatter
 from autogen.agentchat.contrib.multimodal_conversable_agent import MultimodalConversableAgent
 from autogen.code_utils import content_str
 
@@ -162,7 +162,7 @@ def llava_call(prompt: str, llm_config: dict) -> str:
     Makes a call to the LLaVA service to generate text based on a given prompt
     """
 
-    prompt, images = llava_formater(prompt, order_image_tokens=False)
+    prompt, images = llava_formatter(prompt, order_image_tokens=False)
 
     for im in images:
         if len(im) == 0:

autogen/agentchat/contrib/math_user_proxy_agent.py

@@ -4,6 +4,7 @@
 from typing import Any, Callable, Dict, List, Optional, Union, Tuple
 from time import sleep
 
+from autogen._pydantic import PYDANTIC_V1
 from autogen.agentchat import Agent, UserProxyAgent
 from autogen.code_utils import UNKNOWN, extract_code, execute_code, infer_lang
 from autogen.math_utils import get_answer
@@ -384,7 +385,8 @@ class WolframAlphaAPIWrapper(BaseModel):
     class Config:
         """Configuration for this pydantic object."""
 
-        extra = Extra.forbid
+        if PYDANTIC_V1:
+            extra = Extra.forbid
 
     @root_validator(skip_on_failure=True)
     def validate_environment(cls, values: Dict) -> Dict:
@@ -395,8 +397,8 @@ def validate_environment(cls, values: Dict) -> Dict:
         try:
             import wolframalpha
 
-        except ImportError:
-            raise ImportError("wolframalpha is not installed. " "Please install it with `pip install wolframalpha`")
+        except ImportError as e:
+            raise ImportError("wolframalpha is not installed. Please install it with `pip install wolframalpha`") from e
         client = wolframalpha.Client(wolfram_alpha_appid)
         values["wolfram_client"] = client
 

autogen/agentchat/contrib/multimodal_conversable_agent.py

@@ -16,9 +16,14 @@ def colored(x, *args, **kwargs):
 from autogen.code_utils import content_str
 
 DEFAULT_LMM_SYS_MSG = """You are a helpful AI assistant."""
+DEFAULT_MODEL = "gpt-4-vision-preview"
 
 
 class MultimodalConversableAgent(ConversableAgent):
+    DEFAULT_CONFIG = {
+        "model": DEFAULT_MODEL,
+    }
+
     def __init__(
         self,
         name: str,