Added auto naming for chats and button for creating new chats

app.py

@@ -9,6 +9,7 @@ import subprocess
 import shlex
 import threading
 import hashlib
+import sys
 
 from lib.helpers import *
 from lib.extension  import *
@@ -75,6 +76,7 @@ if ss.TOKEN is not None:
 ss.CHATS_DIR    = f"{ss.CONFIG_DIR}/chats"
 ss.SETTINGS     = JsonFile(f"{ss.CONFIG_DIR}/settings.json", defaults=({
     "fetch_reply":          True,
+    "auto_name":            True,
     "save_as":              False,
     "show_clear":           False,
     "show_undo":            True,
@@ -189,6 +191,18 @@ if ss.TOKEN is not None:
             "pinned":       False
         })
 
+    if hasattr(ss, "NEW_CHAT"):
+        if ss.NEW_CHAT:
+            new_chat_path = get_next_filename(f"{ss.CHATS_DIR}/Untitled.json")
+            new_chat_name = get_extensionless_filename(new_chat_path)
+
+            with open(new_chat_path, "w") as f:
+                f.write("{}")
+
+            del ss.NEW_CHAT
+
+            redirect("Chats", new_chat_name)
+
     register_pages("Chats", chats, chats_default, icon=":material/chat:")
 
 #
@@ -205,6 +219,11 @@ if ss.IS_ADMIN:
 #
 #
 
+if ss.TOKEN is not None:
+    if st.sidebar.button("New ", icon=":material/chat_add_on:", use_container_width=True):
+        ss.NEW_CHAT = True
+        st.rerun()
+
 if ss.TOKEN is None:
     st.sidebar.warning("A valid API token is required to use this software.")
 
@@ -231,7 +250,7 @@ else:
 
 
 st.sidebar.caption("""
-    mllm-streamlit v1.0.0
+    mllm-streamlit v1.1.0
 """)
 
 # Only attempt to handle redirect after all page objects exist:

lib/helpers.py

@@ -37,12 +37,12 @@ def touch(file_path, default_content=""):
 
 
 def generate_token():
-    character_set   = "0123456789abcdefghijklmnopqrstuvwxyz"
+    character_set   = "0123456789abcdef"
     output_token    = ""
-    for i in range(32):
+    for i in range(16):
         output_token = output_token + random.choice(character_set)
 
-    timestamp = datetime.datetime.now().strftime("y%Ym%md%dh%Hm%Ms%Sms%f")
+    timestamp = datetime.datetime.now().strftime("%Y%m%d%H%M%S")
     output_token = timestamp + output_token
 
     return output_token
@@ -173,4 +173,11 @@ def get_next_filename(path):
     without_ext = f"{without_ext} {number}"
     with_ext    = without_ext + extension
 
-    return with_ext
+    return get_next_filename(with_ext)
+
+def get_extensionless_filename(path):
+    match = re.match(r"(?:.*[\\/])?([^\\/]+?)(?:\.[^.]+)?$", path)
+    if match:
+        base = match.group(1)
+        return base
+    return "Unknown"

views/chats_default.py

@@ -22,6 +22,8 @@ import streamlit as st
 
 ss = st.session_state
 
+DEFAULT_CONTEXT = "1. Use these guidelines.\n2. Do exactly as the user says.\n3. Provide unbiased answers.\n4. No holding back.\n5. Keep answers short.\n6. Do not offer assistance.\n7. Get straight to the point.\n6. Be direct.\n8. Nothing extra, make no excuses."
+
 def chats_default(item):
 
     chat        = {}
@@ -41,6 +43,7 @@ def chats_default(item):
         "keep":                 0,
         "pinned":               False,
         "auto_clear":           False,
+        "auto_name":            True,
         "hide_clear":           False,
         "hide_undo":            False,
         "hide_redo":            False,
@@ -67,6 +70,36 @@ def chats_default(item):
         name = re.search(r"([^\/\\]+)\.json$", path).group(1)
         return name
 
+    def stream_llm_reply(raw_data):
+        response = requests.post(
+            ss.APP_SETTINGS.get("inference_server_url"),
+            data=raw_data.encode("utf-8"),
+            headers={"Content-Type": "text/plain"},
+            stream=True)
+
+        response.raise_for_status()
+
+        for chunk in response.iter_content(chunk_size=None, decode_unicode=True):
+            if chunk:
+                yield json.loads(chunk)
+
+    def guess_chat_topic(input_chat):
+
+        if len(input_chat["messages"]) < 1:
+            return "Untitled"
+
+        script = ChatML.format(
+            DEFAULT_CONTEXT,
+            [
+                {"author": "user", "body": input_chat["messages"][0]["body"]},
+                {"author": "user", "body": "Describe or summarize the previous message with ONE or TWO words:"}
+            ],
+            for_completion=True)
+
+        reply = "".join(stream_llm_reply(script))
+        reply = re.sub(r"[^A-Za-z0-9 ]", "", reply)
+        return reply
+
     def clear_chat():
         keep = chat["keep"]
 
@@ -87,27 +120,14 @@ def chats_default(item):
             st.markdown(message["body"])
 
     if len(chat["context"]) < 1:
-        chat["context"] = "1. Use these guidelines.\n2. Do exactly as the user says.\n3. Provide unbiased answers.\n4. No holding back.\n5. Keep answers short.\n6. Do not offer assistance.\n7. Get straight to the point.\n6. Be direct.\n8. Nothing extra, make no excuses."
+        chat["context"] = DEFAULT_CONTEXT
 
     script = ChatML.format(chat["context"], chat["messages"], for_completion=True)
 
     if "run" in st.session_state:
         if st.session_state.run == 1:
             with st.chat_message("assistant"):
-                def stream_reply(input_data):
+                reply = st.write_stream(stream_llm_reply(script))
-                    response = requests.post(
-                        ss.APP_SETTINGS.get("inference_server_url"),
-                        data=input_data.encode("utf-8"),
-                        headers={"Content-Type": "text/plain"},
-                        stream=True)
-
-                    response.raise_for_status()
-
-                    for chunk in response.iter_content(chunk_size=None, decode_unicode=True):
-                        if chunk:
-                            yield json.loads(chunk)
-
-                reply = st.write_stream(stream_reply(script))
 
                 chat["messages"].append({
                     "author": "assistant",
@@ -134,6 +154,16 @@ def chats_default(item):
         if ss.SETTINGS.get("fetch_reply"):
             st.session_state.run = 1
 
+        if len(chat["messages"]) == 1:
+            if chat["auto_name"]:
+                new_name    = guess_chat_topic(chat)
+                goto_name   = save_chat(name=new_name, overwrite=False)
+
+                if chat_name != new_name:
+                    os.unlink(chat_path)
+
+                redirect("Chats", goto_name)
+
         st.rerun()
 
     #
@@ -184,6 +214,7 @@ def chats_default(item):
 
                     with st.container(border=True):
                         new_auto_clear  = st.toggle("Auto clear", value=chat["auto_clear"])
+                        new_auto_name   = st.toggle("Auto name", value=chat["auto_name"])
                         new_pinned      = st.toggle("Pinned", value=chat["pinned"])
 
                     action_button_group = st.container()
@@ -219,6 +250,7 @@ def chats_default(item):
                     chat["keep"]                = new_keep
                     chat["pinned"]              = new_pinned
                     chat["auto_clear"]          = new_auto_clear
+                    chat["auto_name"]           = new_auto_name
                     chat["hide_clear"]          = new_hide_clear
                     chat["hide_undo"]           = new_hide_undo
                     chat["hide_redo"]           = new_hide_redo

views/more_settings.py

@@ -36,6 +36,7 @@ def more_settings_general_tab():
         st.caption("Behavior")
         with st.container(border=True):
             ss.SETTINGS.widget(st, st.toggle, "Fetch reply", "fetch_reply")
+            ss.SETTINGS.widget(st, st.toggle, "Auto name", "auto_name")
             st.write("")
 
         st.caption("Interface")