From f29c4c943545ca27aa09657fc285eb8048936202 Mon Sep 17 00:00:00 2001
From: "Renato Xavier da Silveira Rosa (RenatoXSR)" <renatoxsr@gmail.com>
Date: Sun, 27 Sep 2026 23:08:45 -0300
Subject: [PATCH] prompt streaming working

---
 src-py/renatoxsr/scripts/prompt.py | 40 ++++++++++++++++++++----------
 1 file changed, 27 insertions(+), 13 deletions(-)

diff --git a/src-py/renatoxsr/scripts/prompt.py b/src-py/renatoxsr/scripts/prompt.py
index 45f63b8..7227f9b 100644
--- a/src-py/renatoxsr/scripts/prompt.py
+++ b/src-py/renatoxsr/scripts/prompt.py
@@ -16,6 +16,7 @@ class bb(enum.StrEnum):
     content="bold blue"
     model="bold red"
     prompt="bold green"
+    line="italic"
 
 def write_rich(*msg, end='\n', stream=sys.stdout):
     output[stream].print(*msg, end=end)
@@ -40,8 +41,8 @@ def parse_args():
                 "qwen3.5:4b",
             ])
     p.add_argument("--prompt")
-    p.add_argument("--system-prompt",
-                   default="")
+    p.add_argument("--system",
+                   default="You are a helpful assistant. Write concisely.")
 
     mxg = p.add_mutually_exclusive_group()
     mxg.add_argument("--pull")
@@ -77,17 +78,21 @@ def parse_args():
     return p.parse_args()
 
 
+def s3f(ns):
+    return f"{(ns // 1_000_000) / 1000:.3f}"
+
+
 async def do_async_turn(conf):
     """Do user->assistant turn streaming responses"""
     if not conf.prompt:
-        prompt = Prompt.ask(f"[{bb.model}]Ask {conf.model}»[/] [{bb.prompt}]")
+        prompt = Prompt.ask(f"[{str(bb.model)}]Ask {conf.model}»[/] [{str(bb.prompt)}]")
     if conf.response:
         system_and_user = {
             "input": conf.prompt or prompt,
-            "instructions": conf.system_prompt}
+            "instructions": conf.system}
     else:
         system_and_user = {"messages": [
-            {"role":"system", "content": conf.system_prompt},
+            {"role":"system", "content": conf.system},
             {"role": "user", "content": conf.prompt or prompt}]}
         
     client = httpx.AsyncClient(
@@ -114,23 +119,32 @@ async def do_async_turn(conf):
                 continue
             data = json.loads(line)
             if data.get("done", False):
-                write(data)
+                write(f"\n[Load: {s3f(data['load_duration'])}s | "
+                      f"Analyze: {s3f(data['prompt_eval_duration'])}s | "
+                      f"Generate: {s3f(data['eval_duration'])}s] "
+                      f"Total: {s3f(data['total_duration'])}s"
+                    f"\nToken usage: "
+                    f"{data['prompt_eval_count']} input, "
+                    f"{data['prompt_eval_cached_count']} cached, "
+                    f"{data['eval_count']} output.\n")
                 msg['done'] = data
                 return msg
             #write(f"[i]{data}[/i]")
             for i in ["thinking","content"]:
                 text = data.get("message", {}).get(i, "")
                 msg[i][msg[i]['len']].append(text)
-                write(f"[i]{text}[/i]", end='')     
-                if '\\n' in msg[i][msg[i]['len']]:
-                    write(f"[{bb.line}]{msg[i]['len']:>3d}:[/] "
-                          f"[{bb(i)}]{Markdown(msg[i][msg[i]['len']])}[/]")
+                write(f"[{str(bb[i])}]{text}[/]", end='')     
+                if '\n' in text:
+                    #write(f"[{str(bb.line)}]{msg[i]['len']:>3d}:[/] "
+                    #      f"[{str(bb[i])}]")
+                    #write(Markdown("".join(msg[i][msg[i]['len']])))
                     msg[i]['len'] += 1
+                    msg[i][msg[i]['len']] = []
 
 def main():
     conf = parse_args()
     if conf.prompt:
-        write(f"[{bb.model}]Ask {conf.model}»[/] [{bb.prompt}]{conf.prompt}[/]")
+        write(f"[{str(bb.model)}]Ask {conf.model}»[/] [{str(bb.prompt)}]{conf.prompt}[/]")
     try:
         while True:
             if conf.stream:
@@ -138,8 +152,8 @@ def main():
             else:
                 data = res.json()
                 if "thinking" in data.get("message",{}):
-                    write(f"[{bb.model}]{conf.model}'s thinking:[/] [{bb.thinking}] {Markdown(data['message']['thinking'])}[/]")
-                write(f"[{bb.model}]{conf.model}'s response: [/][{bb.content}]{Markdown(data['message']['content'])}[/]")
+                    write(f"[{str(bb.model)}]{conf.model}'s thinking:[/] [{str(bb.thinking)}] {Markdown(data['message']['thinking'])}[/]")
+                write(f"[{str(bb.model)}]{conf.model}'s response: [/][{str(bb.content)}]{Markdown(data['message']['content'])}[/]")
                 #response = response['choices'][0]['message']
             if conf.prompt:
                 sys.exit(0)