Mismatch between the max_tokens and thinking.budget_tokens in thinking mode

[soxfmr]

Issue description

When thinking mode is enabled for Claude Code, it assumes to consume more tokens which is restricted by thinking.budget_tokens parameter. Currently the code fixates the max_tokens to 8192 which may cap the tokens specified in thinking.budget_tokens and causes the API server failed to process the context that requires more tokens:

ClaudeHttpError(error05:55:04 [157/1874]
ge_key='claudeClient.httpError', status_code=400, context={'url': 'https://api.anthropic.com/v1/messages', 'status_code': 400, 'error_type': 'invalid_request_error', 'error_message': '`max_tokens` must be greater than `thinking.budget_tokens`. Please consult our documentation at https://docs.claude.com/en/docs/build-with-claude/extended-thinking#max-tokens-and-context-window-size'})

Logs

clove  | 2026-01-29 02:33:50.039 | INFO     | app.processors.claude_ai.claude_api_processor:process:81 - Using cached account: None...                           
clove  | 2026-01-29 02:33:50.328 | ERROR    | app.processors.claude_ai.claude_api_processor:process:145 - Claude API error: 400 - {'type': 'error', 'error': {'ty
pe': 'invalid_request_error', 'message': '`max_tokens` must be greater than `thinking.budget_tokens`. Please consult our documentation at https://docs.claude.com
/en/docs/build-with-claude/extended-thinking#max-tokens-and-context-window-size'}, 'request_id': 'req_xxxx'}                       
clove  | 2026-01-29 02:33:50.328 | ERROR    | app.processors.claude_ai.pipeline:process:78 - Pipeline processing failed: ClaudeHttpError(error_code=503130, messa
ge_key='claudeClient.httpError', status_code=400, context={'url': 'https://api.anthropic.com/v1/messages', 'status_code': 400, 'error_type': 'invalid_request_err
or', 'error_message': '`max_tokens` must be greater than `thinking.budget_tokens`. Please consult our documentation at https://docs.claude.com/en/docs/build-with
-claude/extended-thinking#max-tokens-and-context-window-size'})                                                                                                  
clove  | 2026-01-29 02:33:50.329 | WARNING  | app.core.error_handler:handle_app_exception:71 - AppException: ClaudeHttpError - Code: 503130, Message: HTTP error 
occurred when calling Claude AI: invalid_request_error - `max_tokens` must be greater than `thinking.budget_tokens`. Please consult our documentation at https://
docs.claude.com/en/docs/build-with-claude/extended-thinking#max-tokens-and-context-window-size (Status: 400), Context: {'url': 'https://api.anthropic.com/v1/mess
ages', 'status_code': 400, 'error_type': 'invalid_request_error', 'error_message': '`max_tokens` must be greater than `thinking.budget_tokens`. Please consult ou
r documentation at https://docs.claude.com/en/docs/build-with-claude/extended-thinking#max-tokens-and-context-window-size'}

Feasible solution

Increase the hard-code max_tokens, or assign the thinking.budget_tokens retrieved from the context to the max_tokens on runtime.