feat: first cut at openai compatible server

src/settings.rs

@@ -443,6 +443,8 @@ impl GenerationPreferences {
             temperature: self.temperature.unwrap_or(1.0),
             top_p: self.top_p.unwrap_or(if glm { 0.95 } else { 1.0 }),
             min_p: self.min_p.unwrap_or(if glm { 0.0 } else { 0.05 }),
+            top_k: 0,
+            stops: Vec::new(),
             seed: self.seed,
             reasoning_mode: if self.reasoning_mode == ReasoningMode::Max
                 && self.context_tokens < THINK_MAX_MIN_CONTEXT
@@ -489,6 +491,8 @@ pub(crate) struct TurnSettings {
     pub(crate) temperature: f32,
     pub(crate) top_p: f32,
     pub(crate) min_p: f32,
+    pub(crate) top_k: i32,
+    pub(crate) stops: Vec<String>,
     pub(crate) seed: Option<u64>,
     pub(crate) reasoning_mode: ReasoningMode,
 }