From 374a21c7feeb1a08a143f4b86ff9f7df4d2d4164 Mon Sep 17 00:00:00 2001
From: "ilblackdragon@gmail.com" <ilblackdragon@gmail.com>
Date: Sun, 22 Mar 2026 00:50:57 -0700
Subject: [PATCH] fix(bridge): match existing LLM request format to prevent 400
 errors
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The LLM bridge was missing several defaults that the existing
Reasoning.respond_with_tools() sets:

- tool_choice: "auto" when tools are present (required by some providers)
- max_tokens: 4096 (default)
- temperature: 0.7 (default)
- When no tools (force_text): use plain complete() instead of
  complete_with_tools() with empty tools array — matches existing
  no-tools fallback path

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 src/bridge/llm_adapter.rs | 43 ++++++++++++++++++++++++++++++---------
 1 file changed, 33 insertions(+), 10 deletions(-)

diff --git a/src/bridge/llm_adapter.rs b/src/bridge/llm_adapter.rs
index b190f0af..8b17aedd 100644
--- a/src/bridge/llm_adapter.rs
+++ b/src/bridge/llm_adapter.rs
@@ -58,17 +58,40 @@ impl LlmBackend for LlmBridgeAdapter {
             actions.iter().map(action_def_to_tool_def).collect()
         };
 
-        // Build request
-        let mut request = ToolCompletionRequest::new(chat_messages, tools);
-        if let Some(max_tokens) = config.max_tokens {
-            request = request.with_max_tokens(max_tokens);
-        }
-        if let Some(temp) = config.temperature {
-            request = request.with_temperature(temp);
-        }
-        if config.force_text {
-            request = request.with_tool_choice("none");
+        // Build request — match the existing Reasoning.respond_with_tools() defaults
+        let max_tokens = config.max_tokens.unwrap_or(4096);
+        let temperature = config.temperature.unwrap_or(0.7);
+
+        if tools.is_empty() {
+            // No tools: use plain completion (matches existing no-tools path)
+            let mut request = crate::llm::CompletionRequest::new(chat_messages)
+                .with_max_tokens(max_tokens)
+                .with_temperature(temperature);
+            request.metadata = config.metadata.clone();
+
+            let response = provider
+                .complete(request)
+                .await
+                .map_err(|e| EngineError::Llm {
+                    reason: e.to_string(),
+                })?;
+
+            return Ok(LlmOutput {
+                response: LlmResponse::Text(response.content),
+                usage: TokenUsage {
+                    input_tokens: u64::from(response.input_tokens),
+                    output_tokens: u64::from(response.output_tokens),
+                    cache_read_tokens: u64::from(response.cache_read_input_tokens),
+                    cache_write_tokens: u64::from(response.cache_creation_input_tokens),
+                },
+            });
         }
+
+        // With tools: use tool completion (matches existing tools path)
+        let mut request = ToolCompletionRequest::new(chat_messages, tools)
+            .with_max_tokens(max_tokens)
+            .with_temperature(temperature)
+            .with_tool_choice("auto");
         request.metadata = config.metadata.clone();
 
         // Call provider