fixed thinking token logic

main.go

@@ -684,6 +684,7 @@ type streamDelta struct {
     Delta struct {
       ReasoningContent string `json:"reasoning_content"`
       Reasoning        string `json:"reasoning"`
+      Thought          string `json:"thought"`
       Content          string `json:"content"`
       ToolCalls        []struct {
         Index    int    `json:"index"`
@@ -698,6 +699,19 @@ type streamDelta struct {
   Usage *Usage `json:"usage"`
 }
 
+func cleanMessagesForLLM(msgs []Message) []Message {
+  out := make([]Message, len(msgs))
+  for i, m := range msgs {
+    out[i] = Message{
+      Role:       m.Role,
+      Content:    m.Content,
+      ToolCalls:  m.ToolCalls,
+      ToolCallID: m.ToolCallID,
+    }
+  }
+  return out
+}
+
 func sanitizeMessages(msgs []Message) {
   for i := range msgs {
     if msgs[i].Role == "assistant" && len(msgs[i].ToolCalls) > 0 {
@@ -721,8 +735,9 @@ func isInvalidAssistantErr(err error) bool {
 
 func llm(ctx context.Context, cfg *Cfg, msgs []Message, tools []map[string]any) (Message, Usage, error) {
   if err := ctx.Err(); err != nil { return Message{}, Usage{}, err }
-  sanitizeMessages(msgs)
-  p := map[string]any{"model": cfg.Model, "temperature": cfg.Temperature, "messages": msgs, "stream": cfg.Stream}
+  cleanMsgs := cleanMessagesForLLM(msgs)
+  sanitizeMessages(cleanMsgs)
+  p := map[string]any{"model": cfg.Model, "temperature": cfg.Temperature, "messages": cleanMsgs, "stream": cfg.Stream}
   if tools != nil { p["tools"] = tools }
   if cfg.Stream { p["stream_options"] = map[string]any{"include_usage": true} }
   for k, v := range cfg.Raw {
@@ -798,6 +813,7 @@ func llm(ctx context.Context, cfg *Cfg, msgs []Message, tools []map[string]any)
         Message struct {
           Message
           Reasoning string `json:"reasoning"`
+          Thought   string `json:"thought"`
         } `json:"message"`
       } `json:"choices"`
       Usage Usage `json:"usage"`
@@ -809,6 +825,7 @@ func llm(ctx context.Context, cfg *Cfg, msgs []Message, tools []map[string]any)
     if len(cr.Choices) == 0 { return Message{}, Usage{}, errors.New("empty choices in LLM response") }
     m := cr.Choices[0].Message.Message
     if m.ReasoningContent == "" { m.ReasoningContent = cr.Choices[0].Message.Reasoning }
+    if m.ReasoningContent == "" { m.ReasoningContent = cr.Choices[0].Message.Thought }
     u := cr.Usage
     if u.PromptTokens == 0 {
       u.PromptTokens = estTokens(msgs)
@@ -828,7 +845,7 @@ func llm(ctx context.Context, cfg *Cfg, msgs []Message, tools []map[string]any)
 
 func parseStream(ctx context.Context, r io.Reader) (Message, Usage, error) {
   var content, reas string
-  var rh, ch bool
+  var inReasoning bool
   var lineBuf string
   var mdSt mdState
   var tblBuf []string
@@ -860,14 +877,25 @@ func parseStream(ctx context.Context, r io.Reader) (Message, Usage, error) {
     dl := d.Choices[0].Delta
     rc := dl.ReasoningContent
     if rc == "" { rc = dl.Reasoning }
+    if rc == "" { rc = dl.Thought }
     if rc != "" {
-      if !rh { fmt.Println(c("--- reasoning start ---", 36)); rh = true }
+      if !inReasoning {
+        if content != "" {
+          flushTable()
+          if lineBuf != "" { fmt.Println(renderMDLine(lineBuf, &mdSt)); lineBuf = "" }
+          fmt.Println()
+        }
+        fmt.Println(c("--- reasoning start ---", 36))
+        inReasoning = true
+      }
       fmt.Print(c(rc, 2))
       reas += rc
     }
     if dl.Content != "" {
-      if rh && !ch { fmt.Print("\n" + c("--- reasoning end ---", 36) + "\n\n") }
-      ch = true
+      if inReasoning {
+        fmt.Print("\n" + c("--- reasoning end ---", 36) + "\n\n")
+        inReasoning = false
+      }
       content += dl.Content
       lineBuf += dl.Content
       for {
@@ -911,7 +939,7 @@ func parseStream(ctx context.Context, r io.Reader) (Message, Usage, error) {
       fmt.Println(renderMDLine(lineBuf, &mdSt))
     }
   }
-  if rh && !ch {
+  if inReasoning {
     fmt.Print("\n" + c("--- reasoning end ---", 36) + "\n")
   }
   m := Message{Role: "assistant"}