agent mode: APIError

[yennor]

Issue

I've just wanted to try out agent mode:

aider-ce --model openrouter/anthropic/claude-sonnet-4.5 --agent
what ever i write, it results in:

LiteLLM API Error: litellm.BadRequestError: OpenrouterException - {"error":{"message":"Provider returned error","code":400,"metadata":{"raw":"{\"type\":\"error\",\"error\":{\"type\":\"invalid_request_error\",\"message\":\"A maximum of 4 blocks with cache_control may be provided. Found 7.\"},\"request_id\":\"req_011CVuxStX4ojGj47gXdCvJM\"}","provider_name":"Anthropic"}},"user_id":"user_30YqKDHrYOuabrJeyr0BQO3afyE"}
litellm.APIError: Error building chunks for logging/streaming usage calculation
Retrying in 0.2 seconds...
LiteLLM API Error: litellm.BadRequestError: OpenrouterException - {"error":{"message":"Provider returned error","code":400,"metadata":{"raw":"{\"type\":\"error\",\"error\":{\"type\":\"invalid_request_error\",\"message\":\"A maximum of 4 blocks with cache_control may be provided. Found 7.\"},\"request_id\":\"req_vrtx_011CVuxT7aPAEnhZ7Qjjxnpg\"}","provider_name":"Google"}},"user_id":"user_30YqKDHrYOuabrJeyr0BQO3afyE"}
litellm.APIError: Error building chunks for logging/streaming usage calculation
Retrying in 0.5 seconds...

Version and model info

Aider 0.86.1

[minger0]

@dwash96 reproduced this consistently on HEAD docker, and I have at the moment no workaround

agent> /model MiMo-V2-Flash                                                                                                                                                                   
Processing...
Warning for MiMo-V2-Flash: Unknown context window size and costs, using sane defaults.
Error initializing MCP server Local: 'object' object has no attribute 'list_tools'
cecli v0.91.1.dev+less
Main model: MiMo-V2-Flash with agent edit format
Weak model: openrouter/deepseek/deepseek-r1:free
Git repo: .git with 173 files
Repo-map: using 4096 tokens, auto refresh
Warning: map-tokens > 2048 is not recommended. Too much irrelevant code can confuse LLMs.
MCP servers configured: Local
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
design_tool/app_installer.py                                                                                                                                                                  
agent> 'python app_installer.py --test' tests app_installer.py, get the tests pass                                                                                                            
Processing...
LiteLLM API Error: litellm.BadRequestError: LLM Provider NOT provided. Pass in the LLM provider you are trying to call. You passed model=MiMo-V2-Flash
 Pass model as E.g. For 'Huggingface' inference endpoints pass in `completion(model='huggingface/starcoder',..)` Learn more: https://docs.litellm.ai/docs/providers
litellm.APIError: Error building chunks for logging/streaming usage calculation
Retrying in 0.2 seconds...

same for

/model bedrock_converse/eu.anthropic.claude-sonnet-4-5-20250929-v1:0