Merge branch 'main' into dotnet

.github/workflows/dotnet-build.yml

@@ -22,13 +22,15 @@ permissions:
 
 jobs:
   build:
-    name: Build
+    name: Dotnet Build
     runs-on: ubuntu-latest
     defaults:
       run:
         working-directory: dotnet
     steps:
     - uses: actions/checkout@v4
+      with:
+        lfs: true
     - name: Setup .NET
       uses: actions/setup-dotnet@v4
       with:
@@ -54,6 +56,8 @@ jobs:
     needs: build
     steps:
     - uses: actions/checkout@v4
+      with:
+        lfs: true
     - name: Setup .NET
       uses: actions/setup-dotnet@v4
       with:

.github/workflows/dotnet-release.yml

@@ -28,6 +28,8 @@ jobs:
         working-directory: dotnet
     steps:
     - uses: actions/checkout@v4
+      with:
+        lfs: true
     - name: Setup .NET
       uses: actions/setup-dotnet@v4
       with:
@@ -61,3 +63,12 @@ jobs:
         # remove AutoGen.SourceGenerator.snupkg because it's an empty package
         rm ./output/release/AutoGen.SourceGenerator.*.snupkg
         dotnet nuget push --api-key ${{ secrets.AUTOGEN_NUGET_API_KEY }} --source https://api.nuget.org/v3/index.json ./output/release/*.nupkg --skip-duplicate
+    - name: Tag commit
+      run: |
+        Write-Host "Tag commit"
+        # version = eng/MetaInfo.props.Project.PropertyGroup.VersionPrefix
+        $metaInfoContent = cat ./eng/MetaInfo.props
+        $version = $metaInfoContent | Select-String -Pattern "<VersionPrefix>(.*)</VersionPrefix>" | ForEach-Object { $_.Matches.Groups[1].Value }
+        git tag -a "$version" -m "AutoGen.Net release $version"
+        git push origin --tags
+      shell: pwsh

autogen/agentchat/contrib/gpt_assistant_agent.py

@@ -10,7 +10,7 @@
 from autogen import OpenAIWrapper
 from autogen.agentchat.agent import Agent
 from autogen.agentchat.assistant_agent import AssistantAgent, ConversableAgent
-from autogen.oai.openai_utils import retrieve_assistants_by_name
+from autogen.oai.openai_utils import create_gpt_assistant, retrieve_assistants_by_name, update_gpt_assistant
 
 logger = logging.getLogger(__name__)
 
@@ -50,7 +50,8 @@ def __init__(
                 - check_every_ms: check thread run status interval
                 - tools: Give Assistants access to OpenAI-hosted tools like Code Interpreter and Knowledge Retrieval,
                         or build your own tools using Function calling. ref https://platform.openai.com/docs/assistants/tools
-                - file_ids: files used by retrieval in run
+                - file_ids: (Deprecated) files used by retrieval in run. It is Deprecated, use tool_resources instead. https://platform.openai.com/docs/assistants/migration/what-has-changed.
+                - tool_resources: A set of resources that are used by the assistant's tools. The resources are specific to the type of tool.
             overwrite_instructions (bool): whether to overwrite the instructions of an existing assistant. This parameter is in effect only when assistant_id is specified in llm_config.
             overwrite_tools (bool): whether to overwrite the tools of an existing assistant. This parameter is in effect only when assistant_id is specified in llm_config.
             kwargs (dict): Additional configuration options for the agent.
@@ -90,7 +91,6 @@ def __init__(
                     candidate_assistants,
                     instructions,
                     openai_assistant_cfg.get("tools", []),
-                    openai_assistant_cfg.get("file_ids", []),
                 )
 
             if len(candidate_assistants) == 0:
@@ -101,12 +101,12 @@ def __init__(
                         "No instructions were provided for new assistant. Using default instructions from AssistantAgent.DEFAULT_SYSTEM_MESSAGE."
                     )
                     instructions = AssistantAgent.DEFAULT_SYSTEM_MESSAGE
-                self._openai_assistant = self._openai_client.beta.assistants.create(
+                self._openai_assistant = create_gpt_assistant(
+                    self._openai_client,
                     name=name,
                     instructions=instructions,
-                    tools=openai_assistant_cfg.get("tools", []),
                     model=model_name,
-                    file_ids=openai_assistant_cfg.get("file_ids", []),
+                    assistant_config=openai_assistant_cfg,
                 )
             else:
                 logger.warning(
@@ -127,9 +127,12 @@ def __init__(
                 logger.warning(
                     "overwrite_instructions is True. Provided instructions will be used and will modify the assistant in the API"
                 )
-                self._openai_assistant = self._openai_client.beta.assistants.update(
+                self._openai_assistant = update_gpt_assistant(
+                    self._openai_client,
                     assistant_id=openai_assistant_id,
-                    instructions=instructions,
+                    assistant_config={
+                        "instructions": instructions,
+                    },
                 )
             else:
                 logger.warning(
@@ -154,9 +157,13 @@ def __init__(
                 logger.warning(
                     "overwrite_tools is True. Provided tools will be used and will modify the assistant in the API"
                 )
-                self._openai_assistant = self._openai_client.beta.assistants.update(
+                self._openai_assistant = update_gpt_assistant(
+                    self._openai_client,
                     assistant_id=openai_assistant_id,
-                    tools=openai_assistant_cfg.get("tools", []),
+                    assistant_config={
+                        "tools": specified_tools,
+                        "tool_resources": openai_assistant_cfg.get("tool_resources", None),
+                    },
                 )
             else:
                 # Tools are specified but overwrite_tools is False; do not update the assistant's tools
@@ -198,6 +205,8 @@ def _invoke_assistant(
         assistant_thread = self._openai_threads[sender]
         # Process each unread message
         for message in pending_messages:
+            if message["content"].strip() == "":
+                continue
             self._openai_client.beta.threads.messages.create(
                 thread_id=assistant_thread.id,
                 content=message["content"],
@@ -426,22 +435,23 @@ def delete_assistant(self):
         logger.warning("Permanently deleting assistant...")
         self._openai_client.beta.assistants.delete(self.assistant_id)
 
-    def find_matching_assistant(self, candidate_assistants, instructions, tools, file_ids):
+    def find_matching_assistant(self, candidate_assistants, instructions, tools):
         """
         Find the matching assistant from a list of candidate assistants.
-        Filter out candidates with the same name but different instructions, file IDs, and function names.
-        TODO: implement accurate match based on assistant metadata fields.
+        Filter out candidates with the same name but different instructions, and function names.
         """
         matching_assistants = []
 
         # Preprocess the required tools for faster comparison
-        required_tool_types = set(tool.get("type") for tool in tools)
+        required_tool_types = set(
+            "file_search" if tool.get("type") in ["retrieval", "file_search"] else tool.get("type") for tool in tools
+        )
+
         required_function_names = set(
             tool.get("function", {}).get("name")
             for tool in tools
-            if tool.get("type") not in ["code_interpreter", "retrieval"]
+            if tool.get("type") not in ["code_interpreter", "retrieval", "file_search"]
         )
-        required_file_ids = set(file_ids)  # Convert file_ids to a set for unordered comparison
 
         for assistant in candidate_assistants:
             # Check if instructions are similar
@@ -454,11 +464,12 @@ def find_matching_assistant(self, candidate_assistants, instructions, tools, fil
                 continue
 
             # Preprocess the assistant's tools
-            assistant_tool_types = set(tool.type for tool in assistant.tools)
+            assistant_tool_types = set(
+                "file_search" if tool.type in ["retrieval", "file_search"] else tool.type for tool in assistant.tools
+            )
             assistant_function_names = set(tool.function.name for tool in assistant.tools if hasattr(tool, "function"))
-            assistant_file_ids = set(getattr(assistant, "file_ids", []))  # Convert to set for comparison
 
-            # Check if the tool types, function names, and file IDs match
+            # Check if the tool types, function names match
             if required_tool_types != assistant_tool_types or required_function_names != assistant_function_names:
                 logger.warning(
                     "tools not match, skip assistant(%s): tools %s, functions %s",
@@ -467,9 +478,6 @@ def find_matching_assistant(self, candidate_assistants, instructions, tools, fil
                     assistant_function_names,
                 )
                 continue
-            if required_file_ids != assistant_file_ids:
-                logger.warning("file_ids not match, skip assistant(%s): %s", assistant.id, assistant_file_ids)
-                continue
 
             # Append assistant to matching list if all conditions are met
             matching_assistants.append(assistant)
@@ -496,7 +504,7 @@ def _process_assistant_config(self, llm_config, assistant_config):
 
         # Move the assistant related configurations to assistant_config
         # It's important to keep forward compatibility
-        assistant_config_items = ["assistant_id", "tools", "file_ids", "check_every_ms"]
+        assistant_config_items = ["assistant_id", "tools", "file_ids", "tool_resources", "check_every_ms"]
         for item in assistant_config_items:
             if openai_client_cfg.get(item) is not None and openai_assistant_cfg.get(item) is None:
                 openai_assistant_cfg[item] = openai_client_cfg[item]

autogen/agentchat/conversable_agent.py

@@ -77,6 +77,7 @@ def __init__(
         llm_config: Optional[Union[Dict, Literal[False]]] = None,
         default_auto_reply: Union[str, Dict] = "",
         description: Optional[str] = None,
+        chat_messages: Optional[Dict[Agent, List[Dict]]] = None,
     ):
         """
         Args:
@@ -122,6 +123,9 @@ def __init__(
             default_auto_reply (str or dict): default auto reply when no code execution or llm-based reply is generated.
             description (str): a short description of the agent. This description is used by other agents
                 (e.g. the GroupChatManager) to decide when to call upon this agent. (Default: system_message)
+            chat_messages (dict or None): the previous chat messages that this agent had in the past with other agents.
+                Can be used to give the agent a memory by providing the chat history. This will allow the agent to
+                resume previous had conversations. Defaults to an empty chat history.
         """
         # we change code_execution_config below and we have to make sure we don't change the input
         # in case of UserProxyAgent, without this we could even change the default value {}
@@ -131,7 +135,11 @@ def __init__(
 
         self._name = name
         # a dictionary of conversations, default value is list
-        self._oai_messages = defaultdict(list)
+        if chat_messages is None:
+            self._oai_messages = defaultdict(list)
+        else:
+            self._oai_messages = chat_messages
+
         self._oai_system_message = [{"content": system_message, "role": "system"}]
         self._description = description if description is not None else system_message
         self._is_termination_msg = (
@@ -1210,7 +1218,6 @@ def initiate_chats(self, chat_queue: List[Dict[str, Any]]) -> List[ChatResult]:
         return self._finished_chats
 
     async def a_initiate_chats(self, chat_queue: List[Dict[str, Any]]) -> Dict[int, ChatResult]:
-
         _chat_queue = self._check_chat_queue_for_sender(chat_queue)
         self._finished_chats = await a_initiate_chats(_chat_queue)
         return self._finished_chats

autogen/coding/jupyter/base.py

@@ -10,9 +10,9 @@ class JupyterConnectionInfo:
     """`str` - Host of the Jupyter gateway server"""
     use_https: bool
     """`bool` - Whether to use HTTPS"""
-    port: int
-    """`int` - Port of the Jupyter gateway server"""
-    token: Optional[str]
+    port: Optional[int] = None
+    """`Optional[int]` - Port of the Jupyter gateway server. If None, the default port is used"""
+    token: Optional[str] = None
     """`Optional[str]` - Token for authentication. If None, no token is used"""
 
 

autogen/coding/jupyter/jupyter_client.py

@@ -41,10 +41,12 @@ def _get_headers(self) -> Dict[str, str]:
 
     def _get_api_base_url(self) -> str:
         protocol = "https" if self._connection_info.use_https else "http"
-        return f"{protocol}://{self._connection_info.host}:{self._connection_info.port}"
+        port = f":{self._connection_info.port}" if self._connection_info.port else ""
+        return f"{protocol}://{self._connection_info.host}{port}"
 
     def _get_ws_base_url(self) -> str:
-        return f"ws://{self._connection_info.host}:{self._connection_info.port}"
+        port = f":{self._connection_info.port}" if self._connection_info.port else ""
+        return f"ws://{self._connection_info.host}{port}"
 
     def list_kernel_specs(self) -> Dict[str, Dict[str, str]]:
         response = self._session.get(f"{self._get_api_base_url()}/api/kernelspecs", headers=self._get_headers())

autogen/oai/openai_utils.py

@@ -1,14 +1,17 @@
+import importlib.metadata
 import json
 import logging
 import os
 import re
 import tempfile
+import time
 from pathlib import Path
 from typing import Any, Dict, List, Optional, Set, Union
 
 from dotenv import find_dotenv, load_dotenv
 from openai import OpenAI
 from openai.types.beta.assistant import Assistant
+from packaging.version import parse
 
 NON_CACHE_KEY = ["api_key", "base_url", "api_type", "api_version"]
 DEFAULT_AZURE_API_VERSION = "2024-02-15-preview"
@@ -675,3 +678,103 @@ def retrieve_assistants_by_name(client: OpenAI, name: str) -> List[Assistant]:
         if assistant.name == name:
             candidate_assistants.append(assistant)
     return candidate_assistants
+
+
+def detect_gpt_assistant_api_version() -> str:
+    """Detect the openai assistant API version"""
+    oai_version = importlib.metadata.version("openai")
+    if parse(oai_version) < parse("1.21"):
+        return "v1"
+    else:
+        return "v2"
+
+
+def create_gpt_vector_store(client: OpenAI, name: str, fild_ids: List[str]) -> Any:
+    """Create a openai vector store for gpt assistant"""
+
+    vector_store = client.beta.vector_stores.create(name=name)
+    # poll the status of the file batch for completion.
+    batch = client.beta.vector_stores.file_batches.create_and_poll(vector_store_id=vector_store.id, file_ids=fild_ids)
+
+    if batch.status == "in_progress":
+        time.sleep(1)
+        logging.debug(f"file batch status: {batch.file_counts}")
+        batch = client.beta.vector_stores.file_batches.poll(vector_store_id=vector_store.id, batch_id=batch.id)
+
+    if batch.status == "completed":
+        return vector_store
+
+    raise ValueError(f"Failed to upload files to vector store {vector_store.id}:{batch.status}")
+
+
+def create_gpt_assistant(
+    client: OpenAI, name: str, instructions: str, model: str, assistant_config: Dict[str, Any]
+) -> Assistant:
+    """Create a openai gpt assistant"""
+
+    assistant_create_kwargs = {}
+    gpt_assistant_api_version = detect_gpt_assistant_api_version()
+    tools = assistant_config.get("tools", [])
+
+    if gpt_assistant_api_version == "v2":
+        tool_resources = assistant_config.get("tool_resources", {})
+        file_ids = assistant_config.get("file_ids")
+        if tool_resources.get("file_search") is not None and file_ids is not None:
+            raise ValueError(
+                "Cannot specify both `tool_resources['file_search']` tool and `file_ids` in the assistant config."
+            )
+
+        # Designed for backwards compatibility for the V1 API
+        # Instead of V1 AssistantFile, files are attached to Assistants using the tool_resources object.
+        for tool in tools:
+            if tool["type"] == "retrieval":
+                tool["type"] = "file_search"
+                if file_ids is not None:
+                    # create a vector store for the file search tool
+                    vs = create_gpt_vector_store(client, f"{name}-vectorestore", file_ids)
+                    tool_resources["file_search"] = {
+                        "vector_store_ids": [vs.id],
+                    }
+            elif tool["type"] == "code_interpreter" and file_ids is not None:
+                tool_resources["code_interpreter"] = {
+                    "file_ids": file_ids,
+                }
+
+        assistant_create_kwargs["tools"] = tools
+        if len(tool_resources) > 0:
+            assistant_create_kwargs["tool_resources"] = tool_resources
+    else:
+        # not support forwards compatibility
+        if "tool_resources" in assistant_config:
+            raise ValueError("`tool_resources` argument are not supported in the openai assistant V1 API.")
+        if any(tool["type"] == "file_search" for tool in tools):
+            raise ValueError(
+                "`file_search` tool are not supported in the openai assistant V1 API, please use `retrieval`."
+            )
+        assistant_create_kwargs["tools"] = tools
+        assistant_create_kwargs["file_ids"] = assistant_config.get("file_ids", [])
+
+    logging.info(f"Creating assistant with config: {assistant_create_kwargs}")
+    return client.beta.assistants.create(name=name, instructions=instructions, model=model, **assistant_create_kwargs)
+
+
+def update_gpt_assistant(client: OpenAI, assistant_id: str, assistant_config: Dict[str, Any]) -> Assistant:
+    """Update openai gpt assistant"""
+
+    gpt_assistant_api_version = detect_gpt_assistant_api_version()
+    assistant_update_kwargs = {}
+
+    if assistant_config.get("tools") is not None:
+        assistant_update_kwargs["tools"] = assistant_config["tools"]
+
+    if assistant_config.get("instructions") is not None:
+        assistant_update_kwargs["instructions"] = assistant_config["instructions"]
+
+    if gpt_assistant_api_version == "v2":
+        if assistant_config.get("tool_resources") is not None:
+            assistant_update_kwargs["tool_resources"] = assistant_config["tool_resources"]
+    else:
+        if assistant_config.get("file_ids") is not None:
+            assistant_update_kwargs["file_ids"] = assistant_config["file_ids"]
+
+    return client.beta.assistants.update(assistant_id=assistant_id, **assistant_update_kwargs)

setup.py

@@ -14,7 +14,7 @@
 __version__ = version["__version__"]
 
 install_requires = [
-    "openai>=1.3,<1.21",
+    "openai>=1.3",
     "diskcache",
     "termcolor",
     "flaml",