from_documents function call results with "AttributeError: 'NoneType' object has no attribute 'embed'"

[mkemalm]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

from langchain_community.vectorstores import Chroma
vector_store = Chroma.from_documents(documents=chunks, embedding=self.embedding, persist_directory="./chroma_db")

Error Message and Stack Trace (if applicable)

No response

Description

I am trying to run RAG samples for a few weeks. Codes were running until pulling latest langchain-community. Codes run after downgrading to 0.2.17

System Info

System Information

OS: Linux
OS Version: #1 SMP PREEMPT_DYNAMIC Mon Aug 19 14:09:30 UTC 2024
Python Version: 3.12.5 (main, Aug 7 2024, 00:00:00) [GCC 14.2.1 20240801 (Red Hat 14.2.1-1)]

Package Information

langchain_core: 0.2.40
langchain: 0.2.16
langchain_community: 0.2.17
langsmith: 0.1.122
langchain_chroma: 0.1.4
langchain_text_splitters: 0.2.4

Optional packages not installed

langgraph
langserve

Other Dependencies

aiohttp: 3.9.5
async-timeout: Installed. No version info available.
chromadb: 0.5.3
dataclasses-json: 0.6.7
fastapi: 0.115.0
httpx: 0.27.2
jsonpatch: 1.33
numpy: 1.26.4
orjson: 3.10.7
packaging: 23.2
pydantic: 2.9.2
PyYAML: 6.0.1
requests: 2.31.0
SQLAlchemy: 2.0.32
tenacity: 8.5.0
typing-extensions: 4.12.2

[eyurtsev]

I am trying to run RAG samples for a few weeks. Codes were running until pulling latest langchain-community. Codes run after downgrading to 0.2.17

Could you put the env information when the code breaks? I think the sys info you provided is for langchain_community: 0.2.17

[mkemalm]

System Information

OS: Linux
OS Version: #1 SMP PREEMPT_DYNAMIC Mon Aug 19 14:09:30 UTC 2024
Python Version: 3.12.5 (main, Aug 7 2024, 00:00:00) [GCC 14.2.1 20240801 (Red Hat 14.2.1-1)]

Package Information

langchain_core: 0.3.1
langchain: 0.3.0
langchain_community: 0.3.0
langsmith: 0.1.122
langchain_chroma: 0.1.4
langchain_text_splitters: 0.3.0

Optional packages not installed

langgraph
langserve

Other Dependencies

aiohttp: 3.9.5
async-timeout: Installed. No version info available.
chromadb: 0.5.3
dataclasses-json: 0.6.7
fastapi: 0.115.0
httpx: 0.27.2
jsonpatch: 1.33
numpy: 1.26.4
orjson: 3.10.7
packaging: 23.2
pydantic: 2.9.2
pydantic-settings: 2.5.2
PyYAML: 6.0.1
requests: 2.31.0
SQLAlchemy: 2.0.32
tenacity: 8.5.0
typing-extensions: 4.12

[luon32]

I can second the observation: Latest "langchain_community" version breaks the "from_documents" call with the above mentioned error and downgrade to "langchain_community" = 0.2.17 solves the issue

[andreabellacicca]

Any news? Downgrading to "langchain_community" = 0.2.17 breaks the other packages.

If it can help, I have investigated the issue. It is an initialization issue.

This is my code:

from langchain_community.embeddings.fastembed import FastEmbedEmbeddings
from langchain_chroma import Chroma

emb_model = FastEmbedEmbeddings(model_name="Qdrant/clip-ViT-B-32-text")
vector_store = Chroma(collection_name="my_collection",
                                             embedding_function=emb_model
)

What I found is The _model protected member from FastEmbedEmbeddings was never filled with the TextEmbedding. The TextEmbedding class is initialized though.

a simple patch is:

from langchain_community.embeddings.fastembed import FastEmbedEmbeddings
from fastembed import TextEmbedding

emb_model = FastEmbedEmbeddings(model_name="Qdrant/clip-ViT-B-32-text")
text_emb = TextEmbedding(model_name="Qdrant/clip-ViT-B-32-text")
emb_model._model = text_emb
vector_store = Chroma(collection_name="my_collection",
                                             embedding_function=emb_model
)

In this way, of course, you initialize TextEmbedding two times.

If it can be useful, I also found that:

1. In langchain_community.embeddings.fastembed.FastEmbedEmbeddings the values['_model'] is loaded correctly

class FastEmbedEmbeddings(BaseModel, Embeddings):
...
    def validate_environment(cls, values: Dict) -> Dict:
        ....
        values["_model"] = fastembed.TextEmbedding(
            model_name=model_name,
            max_length=max_length,
            cache_dir=cache_dir,
            threads=threads,
        )
        return values

2. But it is lost during pydantic validation. This is the last time I can found it (pydantic.main)

3. Here _model is initalized with None

Hope it can help

[BenedictusAryo]

My current simple patch while waiting for the fix:

from langchain_community.embeddings.fastembed import FastEmbedEmbeddings

embedding_model = FastEmbedEmbeddings(model_name="your_model_name")
if embedding_model._model is None:
    embedding_model._model = embedding_model.model_extra['_model']    

so no need to initialize it twice