fix: increase Anthropic max_tokens and warn on truncation

[efecnc]

Summary

• Increases default max_tokens for Anthropic Opus/Sonnet models from 8192 to 16384
• Adds with_max_tokens() builder method for explicit configuration
• Logs a warning when stop_reason == "max_tokens" so truncation is visible

Root Cause

max_tokens was hardcoded at 8192 for all Anthropic models. When output was truncated, stop_reason was never checked, so incomplete tool_use JSON was silently parsed with unwrap_or_default() — tools received empty args and failed mysteriously.

Test plan

• Verify cargo check passes
• Test with Claude Sonnet — long tool-heavy responses should no longer be truncated
• Verify truncation warning appears in logs when output hits the limit

[gemini-code-assist]

Code Review

This pull request updates the AnthropicProvider to dynamically set default max_tokens based on the model name, adds a with_max_tokens builder method, and introduces logging for truncated responses. Feedback indicates that the proposed token limits for Opus and Sonnet models exceed current API constraints and may result in errors, suggesting lower defaults and case-insensitive model name matching.

[gemini-code-assist]

The proposed default of 16384 tokens for Opus and Sonnet models is likely to cause 400 Bad Request errors from the Anthropic API. As of the current API version (2023-06-01), the maximum output tokens for Claude 3 Opus is 4096, and for Claude 3.5 Sonnet it is 8192. While a beta feature exists for 16384 tokens on Sonnet 3.5, it requires the anthropic-beta: max-tokens-3-5-sonnet-2024-07-15 header, which is not currently included in the request. Additionally, the model name check is case-sensitive; using to_lowercase() would be more robust.

        let model_lower = model.to_lowercase();
        let max_tokens = if model_lower.contains("sonnet") {
            8192
        } else if model_lower.contains("opus") {
            4096
        } else {
            8192
        };