[BUG] Bedrock feedback loop is throwing the error

[Yeswanth-gif]

Description

With bedrock llms the feedback loop seems to be impossible as it is happening for each and every task and even when i am training the crew as this will not giving how the bedrock llms are expecting the input that it is giving role as message. I believe this is given based on the requirement for the openai but i guess we need to customize this according to the llm provider. Please look after this. Thanks !

Steps to Reproduce

1. Replace the llm for the agent to one of the bedrock llm
2. give human_input=True for one of the Task
3. kickoff the crew

Expected behavior

Expected to take the feedback.

Screenshots/Code snippets

2025-01-06 10:52:36,967 - 8561692480 - llm.py-llm:187 - ERROR: LiteLLM call failed: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message Error during LLM call to classify human feedback: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message. Retrying... (1/3) 2025-01-06 10:52:36,977 - 8561692480 - llm.py-llm:187 - ERROR: LiteLLM call failed: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message Error during LLM call to classify human feedback: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message. Retrying... (2/3) 2025-01-06 10:52:36,986 - 8561692480 - llm.py-llm:187 - ERROR: LiteLLM call failed: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message Error during LLM call to classify human feedback: litellm.BadRequestError: Invalid Message bedrock requires at least one non-system message. Retrying... (3/3) Error processing feedback after multiple attempts.

Operating System

macOS Sonoma

Python Version

3.12

crewAI Version

0.95.0

crewAI Tools Version

0.25.8

Virtual Environment

Venv

Evidence

None

Possible Solution

None

Additional context

None