Merge branch 'microsoft:main' into arjun-g-dev-tool-config

.github/workflows/contrib-tests.yml

@@ -91,6 +91,17 @@ jobs:
         image: mongodb/mongodb-atlas-local:latest
         ports:
           - 27017:27017
+      couchbase:
+        image: couchbase:enterprise-7.6.3
+        ports:
+          - "8091-8095:8091-8095"
+          - "11210:11210"
+          - "9102:9102"
+        healthcheck: # checks couchbase server is up
+          test: ["CMD", "curl", "-v", "http://localhost:8091/pools"]
+          interval: 20s
+          timeout: 20s
+          retries: 5
     steps:
       - uses: actions/checkout@v4
       - name: Set up Python ${{ matrix.python-version }}
@@ -111,6 +122,9 @@ jobs:
       - name: Install mongodb when on linux
         run: |
           pip install -e .[retrievechat-mongodb]
+      - name: Install couchbase when on linux
+        run: |
+          pip install -e .[retrievechat-couchbase]
       - name: Install unstructured when python-version is 3.9 and on linux
         if: matrix.python-version == '3.9'
         run: |

autogen/agentchat/contrib/gpt_assistant_agent.py

@@ -209,10 +209,12 @@ def _invoke_assistant(
         for message in pending_messages:
             if message["content"].strip() == "":
                 continue
+            # Convert message roles to 'user' or 'assistant', by calling _map_role_for_api, to comply with OpenAI API spec
+            api_role = self._map_role_for_api(message["role"])
             self._openai_client.beta.threads.messages.create(
                 thread_id=assistant_thread.id,
                 content=message["content"],
-                role=message["role"],
+                role=api_role,
             )
 
         # Create a new run to get responses from the assistant
@@ -240,6 +242,28 @@ def _invoke_assistant(
         self._unread_index[sender] = len(self._oai_messages[sender]) + 1
         return True, response
 
+    def _map_role_for_api(self, role: str) -> str:
+        """
+        Maps internal message roles to the roles expected by the OpenAI Assistant API.
+
+        Args:
+            role (str): The role from the internal message.
+
+        Returns:
+            str: The mapped role suitable for the API.
+        """
+        if role in ["function", "tool"]:
+            return "assistant"
+        elif role == "system":
+            return "system"
+        elif role == "user":
+            return "user"
+        elif role == "assistant":
+            return "assistant"
+        else:
+            # Default to 'assistant' for any other roles not recognized by the API
+            return "assistant"
+
     def _get_run_response(self, thread, run):
         """
         Waits for and processes the response of a run from the OpenAI assistant.

autogen/agentchat/contrib/retrieve_user_proxy_agent.py

@@ -178,7 +178,7 @@ def __init__(
                     vector db. Default is None, SentenceTransformer with the given `embedding_model`
                     will be used. If you want to use OpenAI, Cohere, HuggingFace or other embedding
                     functions, you can pass it here,
-                    follow the examples in `https://docs.trychroma.com/embeddings`.
+                    follow the examples in `https://docs.trychroma.com/guides/embeddings`.
                 - `customized_prompt` (Optional, str) - the customized prompt for the retrieve chat.
                     Default is None.
                 - `customized_answer_prefix` (Optional, str) - the customized answer prefix for the

autogen/agentchat/contrib/vectordb/base.py

@@ -201,7 +201,7 @@ class VectorDBFactory:
     Factory class for creating vector databases.
     """
 
-    PREDEFINED_VECTOR_DB = ["chroma", "pgvector", "mongodb", "qdrant"]
+    PREDEFINED_VECTOR_DB = ["chroma", "pgvector", "mongodb", "qdrant", "couchbase"]
 
     @staticmethod
     def create_vector_db(db_type: str, **kwargs) -> VectorDB:
@@ -231,6 +231,10 @@ def create_vector_db(db_type: str, **kwargs) -> VectorDB:
             from .qdrant import QdrantVectorDB
 
             return QdrantVectorDB(**kwargs)
+        if db_type.lower() in ["couchbase", "couchbasedb", "capella"]:
+            from .couchbase import CouchbaseVectorDB
+
+            return CouchbaseVectorDB(**kwargs)
         else:
             raise ValueError(
                 f"Unsupported vector database type: {db_type}. Valid types are {VectorDBFactory.PREDEFINED_VECTOR_DB}."

autogen/agentchat/contrib/vectordb/chromadb.py

@@ -14,6 +14,11 @@
 except ImportError:
     raise ImportError("Please install chromadb: `pip install chromadb`")
 
+try:
+    from chromadb.errors import ChromaError
+except ImportError:
+    ChromaError = Exception
+
 CHROMADB_MAX_BATCH_SIZE = os.environ.get("CHROMADB_MAX_BATCH_SIZE", 40000)
 logger = get_logger(__name__)
 
@@ -84,7 +89,7 @@ def create_collection(
                 collection = self.active_collection
             else:
                 collection = self.client.get_collection(collection_name, embedding_function=self.embedding_function)
-        except ValueError:
+        except (ValueError, ChromaError):
             collection = None
         if collection is None:
             return self.client.create_collection(