From afef600dd86232b14646abb4b0a0c99082fa7f4b Mon Sep 17 00:00:00 2001
From: efecanceliksoy <efecanceliksoy@gmail.com>
Date: Wed, 6 May 2026 14:53:52 +0300
Subject: [PATCH] fix: increase Anthropic max_tokens default and warn on
 truncation

max_tokens was hardcoded at 8192 for all Anthropic models. Newer models
(Opus, Sonnet) support much higher output. Truncated responses could
produce incomplete tool_use JSON that silently breaks tool execution.

- Default to 16384 for opus/sonnet models, 8192 for others
- Add with_max_tokens() builder for explicit configuration
- Log warning when stop_reason is max_tokens so truncation is visible
---
 src/provider.rs | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)

diff --git a/src/provider.rs b/src/provider.rs
index 1e44c9b..7d5ceff 100644
--- a/src/provider.rs
+++ b/src/provider.rs
@@ -83,12 +83,18 @@ pub struct AnthropicProvider {
 
 impl AnthropicProvider {
     pub fn new(base_url: &str, api_key: &str, model: &str) -> Self {
+        // Use a higher default for newer models that support longer output
+        let max_tokens = if model.contains("opus") || model.contains("sonnet") {
+            16384
+        } else {
+            8192
+        };
         Self {
             api_key: api_key.to_string(),
             model: model.to_string(),
             base_url: base_url.to_string(),
             temperature: 0.3,
-            max_tokens: 8192,
+            max_tokens,
             client: build_reqwest_client(),
         }
     }
@@ -98,6 +104,11 @@ impl AnthropicProvider {
         self
     }
 
+    pub fn with_max_tokens(mut self, max_tokens: u32) -> Self {
+        self.max_tokens = max_tokens;
+        self
+    }
+
     /// Convert OpenAI-format tool definitions to Anthropic format.
     fn convert_tools(tools: &Value) -> Value {
         // OpenAI: [{ "type": "function", "function": { "name", "description", "parameters" } }]
@@ -366,6 +377,16 @@ impl Provider for AnthropicProvider {
             }
         }
 
+        // Check if response was truncated due to max_tokens
+        let stop_reason = json_resp["stop_reason"].as_str().unwrap_or("");
+        if stop_reason == "max_tokens" {
+            log::warn!(
+                "Anthropic response truncated (stop_reason=max_tokens, max_tokens={}). \
+                 Tool calls may be incomplete.",
+                self.max_tokens
+            );
+        }
+
         let usage = json_resp.get("usage").map(|u| TokenUsage {
             prompt_tokens: u["input_tokens"].as_u64().unwrap_or(0) as u32,
             completion_tokens: u["output_tokens"].as_u64().unwrap_or(0) as u32,