New provider: Amazon Bedrock (AWS)

[dgallitelli]

Feature description
Please include support for Amazon Bedrock models. These models can be from Amazon, Anthropic, AI21, Cohere, Mistral, or Meta Llama 2.

Your Feature

1. Create a new LLM Provides under metagpt/provider for Amazon Bedrock
2. Include it in the LLMType available

[seehi]

It might be supported later, and if possible, you can also provide a PR.
FYI: @better629

[chosenlu]

could you tell me when will metagpt add bedrock, and how to add provider by myself

[better629]

about next week, you can refs any llm provider under metagpt/provider and complete methods from BaseLLM

[chosenlu]

Thank you very much and I look forward to your updates. I have just started to contact metagpt and I am still familiar with some concepts. I will read the code carefully.

I found another problem examples/rag_search.py can not run
it looks like the llama-index update cause
Traceback (most recent call last):
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/user/rag_search/rag_search.py", line 7, in
from metagpt.rag.engines import SimpleEngine
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/lib/python3.9/site-packages/metagpt/rag/engines/init.py", line 3, in
from metagpt.rag.engines.simple import SimpleEngine
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/lib/python3.9/site-packages/metagpt/rag/engines/simple.py", line 31, in
from metagpt.rag.factories import (
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/lib/python3.9/site-packages/metagpt/rag/factories/init.py", line 3, in
from metagpt.rag.factories.retriever import get_retriever
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/lib/python3.9/site-packages/metagpt/rag/factories/retriever.py", line 15, in
from metagpt.rag.retrievers.bm25_retriever import DynamicBM25Retriever
File "/Users/chosenlu/Job/Python/meta_gpt_agent_newst/lib/python3.9/site-packages/metagpt/rag/retrievers/bm25_retriever.py", line 8, in
from llama_index.retrievers.bm25 import BM25Retriever
ModuleNotFoundError: No module named 'llama_index.retrievers'

[chosenlu]

I resolved
it looks like I need install:
pip install llama-index-retrievers-bm25
pip install llama-index-embeddings-azure-openai
pip install llama-index-llms-azure-openai
pip install llama-index-embeddings-gemini
pip install llama-index-embeddings-ollama

[better629]

Yes, rag is extra module, so you need install it. You can ref to https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html#install-submodules if meet problems.

[chosenlu]

Thank you！I still waiting the bedrock support，Is the bedrock ok now？

[better629]

@chosenlu @dgallitelli
see the pr #1231

[dgallitelli]

Nice, thank you! I'll give it a try and provide some feedback. Quick question: are the Access key and Secret access key strict requirements? Or can they be inferred from environmental variables and therefore omitted in the LLM config?

[better629]

@dgallitelli no, for now, you should config them in LLM config.

[xdstone1on163]

this is the sample code error message:

Input should be 'openai', 'anthropic', 'claude', 'spark', 'zhipuai', 'fireworks', 'open_llm', 'gemini', 'metagpt', 'azure', 'ollama', 'qianfan', 'dashscope', 'moonshot', 'mistral' or 'yi' [type=enum, input_value='bedrock', input_type=str]

why I am seeing no 'bedrock' support by now?

[seehi]

Use the main branch: https://docs.deepwisdom.ai/main/en/guide/get_started/configuration/llm_api_configuration.html

[qingerVT]

what does this mean
"botocore.errorfactory.ValidationException: An error occurred (ValidationException) whe
n calling the InvokeModelWithResponseStream operation: Invocation of model ID anthropi
c.claude-3-opus-20240229-v1:0 with on-demand throughput isn’t supported. Retry your re
quest with the ID or ARN of an inference profile that contains this model"