diff --git a/TODO.org b/TODO.org
index a4db6aa..9de86ad 100644
--- a/TODO.org
+++ b/TODO.org
@@ -1 +1,4 @@
-* TODO Make it possible to stop the model from generating
+* DONE Make it possible to stop the model from generating
+* TODO Add system prompt configuration
+* TODO Add multiple chats
+* TODO Make a ChatSession class with all chat data
diff --git a/chat_session.py b/chat_session.py
new file mode 100644
index 0000000..008acf5
--- /dev/null
+++ b/chat_session.py
@@ -0,0 +1,28 @@
+class ChatSettings():
+ def __init__(self):
+ self.num_gpu: int = 15
+ self.low_vram: bool = False
+ self.num_thread: int = 0
+ self.temperature: int = 1.0
+ self.system_prompt: str = None
+
+ def convert_to_ollama(self):
+ return {
+ "num_gpu": self.num_gpu,
+ "low_vram": self.low_vram,
+ "num_thread": self.num_thread,
+ "temperature": self.temperature,
+ }
+
+class ChatSession():
+ def __init__(self, model_name):
+ self.model_name: str = model_name
+ self.history = []
+ self.settings: ChatSettings = ChatSettings()
+
+ def push_history(self, text: str, role: str = "user"):
+ if not self.history or self.history[-1]["role"] != role:
+ self.history.append({"role": role, "content": text})
+ else:
+ self.history[-1]["content"] += text
+
diff --git a/main.py b/main.py
index 77c41d0..b6359a2 100644
--- a/main.py
+++ b/main.py
@@ -3,6 +3,7 @@ gi.require_version("Gtk", "4.0")
from gi.repository import GLib, Gtk, Gdk
import util
+from chat_session import ChatSession
@Gtk.Template(filename="pyllama.ui")
class PyLlamaWindow(Gtk.ApplicationWindow):
@@ -18,6 +19,8 @@ class PyLlamaWindow(Gtk.ApplicationWindow):
cancel_button = Gtk.Template.Child("cancel_button")
gpu_layers_spin = Gtk.Template.Child("gpu_layers_spin")
low_vram_switch = Gtk.Template.Child("low_vram_switch")
+ threads_spin = Gtk.Template.Child("threads_spin")
+ temperature_spin = Gtk.Template.Child("temperature_spin")
def __init__(self, application=None, title=None):
super().__init__(application=application, title=title)
@@ -97,9 +100,11 @@ class PyLlamaWindow(Gtk.ApplicationWindow):
options = {
"num_gpu": self.gpu_layers_spin.get_value_as_int(),
- "low_vram": self.low_vram_switch.get_state()
+ "low_vram": self.low_vram_switch.get_state(),
+ "num_thread": self.threads_spin.get_value_as_int(),
+ "temperature": self.temperature_spin.get_value_as_int()
}
-
+
async def task():
await app.send_prompt(model, options)
@@ -112,6 +117,22 @@ class PyLlamaWindow(Gtk.ApplicationWindow):
def cancel_button_clicked(self, *args):
app = self.get_application()
app.stop_generation()
+
+ @Gtk.Template.Callback()
+ def gpu_layers_spin_changed(self, *args):
+ pass
+
+ @Gtk.Template.Callback()
+ def low_vram_switch_set(self, *args):
+ pass
+
+ @Gtk.Template.Callback()
+ def threads_spin_changed(self, *args):
+ pass
+
+ @Gtk.Template.Callback()
+ def temperature_spin_changed(self, *args):
+ pass
class PyLlama(Gtk.Application):
def __init__(self):
diff --git a/pyllama.cmb b/pyllama.cmb
index f53c030..de4fd39 100644
--- a/pyllama.cmb
+++ b/pyllama.cmb
@@ -2,5 +2,5 @@
-
+
diff --git a/pyllama.ui b/pyllama.ui
index 48f35da..2015c31 100644
--- a/pyllama.ui
+++ b/pyllama.ui
@@ -46,7 +46,7 @@ along with this program. If not, see . -->
12
horizontal
False
- 6
+ 0
+
+
+
+
+
+
+
+
+
+
+
+