crewAI
crewAI copied to clipboard
Add AWS as an Embedder Provider (required for Memory configuration)
In order to configure a Crew with memory=True
without a OpenAI API Key, one needs to configure an Embedder provider. Currently, AWS (Amazon Bedrock and Amazon SageMaker) are not supported providers.
Ask
- Include AWS in the providers list (
'openai', 'gpt4all', 'huggingface', 'vertexai', 'azure_openai', 'google', 'mistralai', 'nvidia'
). - Support Amazon Bedrock Embedding models
- Support Amazon SageMaker Real-time Endpoints
Things to note
Provider aws_bedrock
is already supported in llm configuration, e.g.:
semantic_search_resume = MDXSearchTool(
mdx='./fake_resume.md',
config=dict(
llm=dict(
provider="aws_bedrock", # <=== WORKS!
config=dict(
model="anthropic.claude-3-haiku-20240307-v1:0",
# Optional parameters can be included here.
# temperature=0.5,
# top_p=1,
# stream=true,
),
),
embedder=dict(
provider="aws_bedrock", <=== DOES NOT WORK!
config=dict(
model="cohere.embed-english-v3",
task_type="retrieval_document",
# Optional title for the embeddings can be added here.
# title="Embeddings",
),
),
)
)