fix: adjust logprob response logic

router/src/lib.rs

@@ -385,23 +385,33 @@ impl From<(Token, Vec<Token>)> for ChatCompletionLogprobs {
 impl From<(Vec<Token>, Vec<Vec<Token>>)> for ChatCompletionLogprobs {
     fn from(value: (Vec<Token>, Vec<Vec<Token>>)) -> Self {
         let (tokens, top_tokens) = value;
-        Self {
+
-            content: tokens
+        // Create an iterator that produces None for top_tokens once it's exhausted
-                .into_iter()
+        let top_tokens_iter = top_tokens
-                .zip(top_tokens)
+            .into_iter()
-                .map(|(t, top_t)| ChatCompletionLogprob {
+            .map(Some)
-                    token: t.text,
+            .chain(std::iter::repeat(None));
-                    logprob: t.logprob,
+
-                    top_logprobs: top_t
+        let content = tokens
+            .into_iter()
+            .zip(top_tokens_iter)
+            .map(|(t, top_t_option)| ChatCompletionLogprob {
+                token: t.text,
+                logprob: t.logprob,
+                top_logprobs: match top_t_option {
+                    Some(top_t) => top_t
                         .into_iter()
                         .map(|t| ChatCompletionTopLogprob {
                             token: t.text,
                             logprob: t.logprob,
                         })
                         .collect(),
-                })
+                    None => vec![], // Handle the case where there are no top tokens
-                .collect(),
+                },
-        }
+            })
+            .collect();
+
+        Self { content }
     }
 }
 

router/src/server.rs

@@ -857,7 +857,7 @@ async fn chat_completions(
             details: true,
             decoder_input_details: !stream,
             seed,
-            top_n_tokens: None,
+            top_n_tokens: req.top_logprobs,
             grammar: tool_grammar.clone(),
         },
     };