trying to get tooling working

main.py

@@ -1,11 +1,24 @@
 import tts
-from textgen import tool_funcs
+from textgen import llm
+
+MODEL_PATH = "./textgen/models/mistral-7b-instruct-v0.3.Q4_K_M.gguf"
+
 
 t = tts.TTS()
 
 save_path = 'audio/weather.wav'
 
-t.create_audio(tool_funcs.get_high_low(), save_path)
+l = llm.TextGen(
+    MODEL_PATH,
+    2048,
+    0
+)
+
+r = l.chat_completion("What is the weather like today?")
+
+print(r)
+
+t.create_audio(r, save_path)
 
 
 

textgen/llm.py

@@ -1,10 +1,11 @@
 # Import the Llama class from the llama_cpp library
 import llama_cpp
 import json
+import random
 
 from . import tool_funcs
 
-tools: list[dict] = json.loads(open('tools.json', 'r').read())['tools']
+tools: list[dict] = json.loads(open('./textgen/tools.json', 'r').read())['tools']
 
 class TextGen:
 
@@ -19,12 +20,12 @@ class TextGen:
         # Provide the path to your downloaded .gguf file
         # n_ctx is the maximum context size (number of tokens) the model can handle.
         # n_gpu_layers specifies how many layers to offload to the GPU. -1 means offload all possible layers.
-        llm = llama_cpp.Llama(
+        self.llm = llama_cpp.Llama(
-        model_path="./models/mistral-7b-instruct-v0.2.Q4_K_M.gguf",  # Path to your GGUF model
+            model_path=model_path,  # Path to your GGUF model
-        n_ctx=n_ctx,           # Context window size
+            n_ctx=n_ctx,           # Context window size
-        n_gpu_layers=n_gpu_layers,        # Offload all layers to GPU. Set to 0 if no GPU.
+            n_gpu_layers=n_gpu_layers,        # Offload all layers to GPU. Set to 0 if no GPU.
-        verbose=False,         # Suppress verbose output
+            verbose=False,         # Suppress verbose output
-        chat_format='chatml'
+            chat_format='chatml'
         )
 
     def generate(self, prompt: str) -> str:
@@ -52,14 +53,18 @@ class TextGen:
             "content": user_message
         })
 
+        print(tools)
+
         response = self.llm.create_chat_completion(
             messages=self.messages,
             tools=tools,
-            tool_choice='auto'
+            tool_choice='auto',
+            seed=random.randint(0, 20000000000)
         )
 
         tool_call = response['choices'][0]['message'].get('tool_calls')
         if not tool_call:
+            # print(response['choices'][0]['message']['content'])
             return response['choices'][0]['message']['content']
 
         call_info = tool_call[0]['function']
@@ -68,5 +73,16 @@ class TextGen:
         print(f'Assistant decided to call {function_name}')
 
         tool_output = tool_funcs.get_high_low()
-        
+
+        self.messages.append(response['choices'][0]['message'])
+        self.messages.append(
+            {
+                "role" : "tool",
+                "content": tool_output
+            }
+        )
+
+        final_response = self.llm.create_chat_completion(messages=self.messages)
+        return final_response['choices'][0]['message']['content']
+
 

textgen/weather.py

@@ -36,10 +36,10 @@ class OpenMeteo:
 
         # Process first location. Add a for-loop for multiple locations or weather models
         response = responses[0]
-        print(f"Coordinates: {response.Latitude()}°N {response.Longitude()}°E")
+        # print(f"Coordinates: {response.Latitude()}°N {response.Longitude()}°E")
-        print(f"Elevation: {response.Elevation()} m asl")
+        # print(f"Elevation: {response.Elevation()} m asl")
-        print(f"Timezone: {response.Timezone()}{response.TimezoneAbbreviation()}")
+        # print(f"Timezone: {response.Timezone()}{response.TimezoneAbbreviation()}")
-        print(f"Timezone difference to GMT+0: {response.UtcOffsetSeconds()}s")
+        # print(f"Timezone difference to GMT+0: {response.UtcOffsetSeconds()}s")
 
         # Process hourly data. The order of variables needs to be the same as requested.
         hourly = response.Hourly()
@@ -57,7 +57,7 @@ class OpenMeteo:
         hourly_data['rain'] = hourly_rain
 
         hourly_dataframe = pd.DataFrame(data = hourly_data)
-        print("\nHourly data\n", hourly_dataframe)
+        # print("\nHourly data\n", hourly_dataframe)
 
         return hourly_dataframe
 

tts.py

@@ -14,7 +14,7 @@ class TTS:
     def create_audio(self, text: str, save_path: str) -> None:
         generator = self.pipeline(text, voice='af_heart')
         for i, (gs, ps, audio) in enumerate(generator):
-            print(i, gs, ps)
+            # print(i, gs, ps)
             display(Audio(data=audio, rate=24000, autoplay=i==0))
             sf.write(save_path, audio, 24000)