[Bug]: ChromaDB OpenAI embedding function is not working

[StamKavid]

Describe the bug

I have the below code:

openai_ef = OpenAIEmbeddingFunction(
api_key= os.getenv("AZURE_OPENAI_API_KEY"),
api_base= os.getenv("AZURE_OPENAI_ENDPOINT"),
model_name= "text-embedding-ada-002"
)

While I am passing it to RetrieveUserProxyAgent as "embedding_function" : openai_ef, i am still getting the below error:

autogen.agentchat.contrib.vectordb.chromadb - INFO - No content embedding is provided. Will use the VectorDB's embedding function to generate the content embedding.

Also in the logs, the link: https://docs.trychroma.com/embeddings is not existing anymore (404)

Steps to reproduce

No response

Model Used

text-embeddings-ada-002

Expected Behavior

TO pass the embedding function into ChromaDB and overpass the default state

Screenshots and logs

Additional Information

No response

[nithinkodithala]

Verify Compatibility: Ensure that the RetrieveUserProxyAgent accepts the embedding function in the manner you're providing it. There might be specific requirements or ways to pass the embedding function.
also try this method
{chromadb_client = ChromaDB(embedding_function=openai_ef)}

[StamKavid]

@nithinkodithyala Thank you very much for the prompt response.

I tried "client": Chroma(client = chromadb.PersistentClient(path="/tmp/chromadb/test"), collection_name="groupchat-collection", embedding_function=embeddings),

but still doesn't seem to solve this.

[StamKavid]

I can confirm that it is working with the below settings:

"client": chromadb.PersistentClient().get_or_create_collection(name = 'autogen_agent', embedding_function=openai_ef)

[Tanv-1]

I tried this:
import chromadb.utils.embedding_functions as embedding_functions
import chromadb

import autogen
from autogen import AssistantAgent
from autogen.agentchat.contrib.retrieve_user_proxy_agent import RetrieveUserProxyAgent

openai_ef = embedding_functions.OpenAIEmbeddingFunction(
api_key="XXXXXXXXXXXXXXXXXXX",
api_base="XXXXXXXXXXXXXXX",
api_type="azure",
api_version="2023-12-01-preview",
model_name="embeddings-large"
)

assistant = AssistantAgent(
name="assistant",
system_message="You are a helpful assistant.",
llm_config={
"timeout": 600,
"cache_seed": 42,
"config_list": config_list,
},
)
ragproxyagent = RetrieveUserProxyAgent(
name="ragproxyagent",
human_input_mode="NEVER",
max_consecutive_auto_reply=3,
retrieve_config={
"task": "qa",
"docs_path": [
<directory_path containing 1000 md files>
],
"model": config_list[0]["model"],
"client": chromadb.PersistentClient(path="C:/Users/Datasemantics/Documents/DS/RAG1/venvPython12/chroma_db_autogen").get_or_create_collection(name = 'danube_phase3', embedding_function=openai_ef),
"collection_name": "danube_phase3",
"embedding_model": "text-embedding-3-large",
"embedding_function": openai_ef,
"overwrite": True, # set to True if you want to overwrite an existing collection
"get_or_create": False, # set to False if don't want to reuse an existing collection
},
code_execution_config=False, # set to False if you don't want to execute the code
)

qa_problem = "How many agents are there?"
chat_result = ragproxyagent.initiate_chat(assistant, message=ragproxyagent.message_generator, problem=qa_problem)

ERROR:
Trying to create collection.
2025-01-04 12:43:18,147 - autogen.agentchat.contrib.retrieve_user_proxy_agent - INFO - Found 4424 chunks.
2025-01-04 12:43:18,202 - autogen.agentchat.contrib.vectordb.chromadb - INFO - No content embedding is provided. Will use the VectorDB's embedding function to generate the content embedding

RateLimitError Traceback (most recent call last)
File c:\Users\Datasemantics\Documents\DS\RAG1.autogen\lib\site-packages\chromadb\api\models\CollectionCommon.py:90, in validation_context..decorator..wrapper(self, *args, **kwargs)
89 try:
---> 90 return func(self, *args, **kwargs)
91 except Exception as e:

File c:\Users\Datasemantics\Documents\DS\RAG1.autogen\lib\site-packages\chromadb\api\models\CollectionCommon.py:406, in CollectionCommon._validate_and_prepare_upsert_request(self, ids, embeddings, metadatas, documents, images, uris)
403 validate_record_set_for_embedding(
404 record_set=upsert_records, embeddable_fields={"documents", "images"}
405 )
--> 406 upsert_embeddings = self._embed_record_set(record_set=upsert_records)
407 else:

File c:\Users\Datasemantics\Documents\DS\RAG1.autogen\lib\site-packages\chromadb\api\models\CollectionCommon.py:526, in CollectionCommon._embed_record_set(self, record_set, embeddable_fields)
525 else:
--> 526 return self._embed(input=record_set[field]) # type: ignore[literal-required]
527 raise ValueError(
528 "Record does not contain any non-None fields that can be embedded."
529 f"Embeddable Fields: {embeddable_fields}"
530 f"Record Fields: {record_set}"
531 )

File c:\Users\Datasemantics\Documents\DS\RAG1.autogen\lib\site-packages\chromadb\api\models\CollectionCommon.py:539, in CollectionCommon._embed(self, input)
535 raise ValueError(
...
91 except Exception as e:
92 msg = f"{str(e)} in {name}."
---> 93 raise type(e)(msg).with_traceback(e.traceback)

TypeError: init() missing 2 required keyword-only arguments: 'response' and 'body'