🪙 feat: Add check_embedding_ctx_length Flag (#161)

* Add environment variable for disabling 'check_embedding_ctx_length'.

* Update README.md

---------

Co-authored-by: duncan <[email protected]>

Author: Fjf

README.md

@@ -85,6 +85,7 @@ The following environment variables are required to run the application:
 - `AWS_ACCESS_KEY_ID`: (Optional) needed for bedrock embeddings
 - `AWS_SECRET_ACCESS_KEY`: (Optional) needed for bedrock embeddings
 - `GOOGLE_APPLICATION_CREDENTIALS`: (Optional) needed for Google VertexAI embeddings
+- `RAG_CHECK_EMBEDDING_CTX_LENGTH` (Optional) Default is true, disabling this will send raw input to the embedder, use this for custom embedding models.
 
 Make sure to set these environment variables before running the application. You can set them in a `.env` file or as system environment variables.
 

app/config.py

@@ -179,6 +179,8 @@ async def dispatch(self, request, call_next):
 AWS_ACCESS_KEY_ID = get_env_variable("AWS_ACCESS_KEY_ID", "")
 AWS_SECRET_ACCESS_KEY = get_env_variable("AWS_SECRET_ACCESS_KEY", "")
 GOOGLE_APPLICATION_CREDENTIALS = get_env_variable("GOOGLE_APPLICATION_CREDENTIALS", "")
+env_value = get_env_variable("RAG_CHECK_EMBEDDING_CTX_LENGTH", "True").lower()
+RAG_CHECK_EMBEDDING_CTX_LENGTH = True if env_value == "true" else False
 
 ## Embeddings
 
@@ -193,6 +195,7 @@ def init_embeddings(provider, model):
             openai_api_base=RAG_OPENAI_BASEURL,
             openai_proxy=RAG_OPENAI_PROXY,
             chunk_size=EMBEDDINGS_CHUNK_SIZE,
+            check_embedding_ctx_length=RAG_CHECK_EMBEDDING_CTX_LENGTH,
         )
     elif provider == EmbeddingsProvider.AZURE:
         from langchain_openai import AzureOpenAIEmbeddings
@@ -203,6 +206,7 @@ def init_embeddings(provider, model):
             azure_endpoint=RAG_AZURE_OPENAI_ENDPOINT,
             api_version=RAG_AZURE_OPENAI_API_VERSION,
             chunk_size=EMBEDDINGS_CHUNK_SIZE,
+            check_embedding_ctx_length=RAG_CHECK_EMBEDDING_CTX_LENGTH,
         )
     elif provider == EmbeddingsProvider.HUGGINGFACE:
         from langchain_huggingface import HuggingFaceEmbeddings