lollms-client 0.13.2__py3-none-any.whl → 0.14.1__py3-none-any.whl

examples/text_gen.py

@@ -1,7 +1,7 @@
 from lollms_client import LollmsClient
 
 # Initialize the LollmsClient instance
-lc = LollmsClient("http://localhost:9600")
+lc = LollmsClient("lollms")
 # Generate Text
 # response = lc.generate_text(prompt="Once upon a time", stream=False, temperature=0.5)
 # print(response)

examples/text_gen_system_prompt.py

@@ -0,0 +1,28 @@
+from lollms_client import LollmsClient
+
+# Initialize the LollmsClient instance
+lc = LollmsClient("ollama",model_name="mistral-nemo:latest")
+# Generate Text
+# response = lc.generate_text(prompt="Once upon a time", stream=False, temperature=0.5)
+# print(response)
+
+# # Generate Completion
+# response = lc.generate_completion(prompt="What is the capital of France", stream=False, temperature=0.5)
+# print(response)
+
+def cb(chunk, type):
+    print(chunk,end="",flush=True)
+    
+response = lc.generate_text(prompt="One plus one equals ", system_prompt="You are a playful dude who never really answers questions correctly. always answer with quirky style.", stream=False, temperature=0.5, streaming_callback=cb)
+print()
+print(response)
+print()
+
+
+# List Mounted Personalities
+response = lc.listMountedPersonalities()
+print(response)
+
+# List Models
+response = lc.listModels()
+print(response)

lollms_client/__init__.py

@@ -6,7 +6,7 @@ from lollms_client.lollms_discussion import LollmsDiscussion, LollmsMessage
 from lollms_client.lollms_utilities import PromptReshaper # Keep general utilities
 from lollms_client.lollms_functions import FunctionCalling_Library
 
-__version__ = "0.13.2"
+__version__ = "0.14.1"
 
 # Optionally, you could define __all__ if you want to be explicit about exports
 __all__ = [

lollms_client/llm_bindings/lollms/__init__.py

@@ -49,6 +49,7 @@ class LollmsLLMBinding(LollmsLLMBinding):
     def generate_text(self, 
                      prompt: str,
                      images: Optional[List[str]] = None,
+                     system_prompt: str = "",
                      n_predict: Optional[int] = None,
                      stream: bool = False,
                      temperature: float = 0.1,
@@ -106,7 +107,7 @@ class LollmsLLMBinding(LollmsLLMBinding):
         
         # Prepare request data
         data = {
-            "prompt": prompt,
+            "prompt":"!@>system: "+system_prompt+"\n"+"!@>user: "+prompt if system_prompt else prompt,
             "model_name": self.model_name,
             "personality": self.personality,
             "n_predict": n_predict,

lollms_client/llm_bindings/ollama/__init__.py

@@ -113,9 +113,9 @@ class OllamaBinding(LollmsLLMBinding):
     def generate_text(self, 
                      prompt: str,
                      images: Optional[List[str]] = None, # List of image file paths
+                     system_prompt: str = "",
                      n_predict: Optional[int] = None,
                      stream: bool = False,
-                     system_prompt = '',
                      temperature: float = 0.7, # Ollama default is 0.8, common default 0.7
                      top_k: int = 40,          # Ollama default is 40
                      top_p: float = 0.9,       # Ollama default is 0.9
@@ -201,15 +201,16 @@ class OllamaBinding(LollmsLLMBinding):
                     )
                     return response_dict.get('message', {}).get('content', '')
             else: # Text-only
+                messages = [{'role': 'system', 'content':system_prompt},{'role': 'user', 'content': prompt}]
                 if stream:
-                    response_stream = self.ollama_client.generate(
+                    response_stream = self.ollama_client.chat(
                         model=self.model_name,
-                        prompt=prompt,
+                        messages=messages,
                         stream=True,
                         options=options if options else None
                     )
                     for chunk_dict in response_stream:
-                        chunk_content = chunk_dict.get('response', '')
+                        chunk_content = chunk_dict.message.content
                         if chunk_content:
                             full_response_text += chunk_content
                             if streaming_callback:
@@ -217,13 +218,13 @@ class OllamaBinding(LollmsLLMBinding):
                                     break
                     return full_response_text
                 else: # Not streaming
-                    response_dict = self.ollama_client.generate(
+                    response_dict = self.ollama_client.chat(
                         model=self.model_name,
-                        prompt=prompt,
+                        messages=messages,
                         stream=False,
                         options=options if options else None
                     )
-                    return response_dict.get('response', '')
+                    return response_dict.message.content
         except ollama.ResponseError as e:
             error_message = f"Ollama API ResponseError: {e.error or 'Unknown error'} (status code: {e.status_code})"
             ASCIIColors.error(error_message)

lollms_client/llm_bindings/openai/__init__.py

@@ -58,6 +58,7 @@ class OpenAIBinding(LollmsLLMBinding):
     def generate_text(self, 
                     prompt: str,
                     images: Optional[List[str]] = None,
+                    system_prompt: str = "",
                     n_predict: Optional[int] = None,
                     stream: bool = False,
                     temperature: float = 0.1,
@@ -98,6 +99,11 @@ class OpenAIBinding(LollmsLLMBinding):
         if images:
             messages = [
                 {
+                    "role": "system",
+                    "content": system_prompt,
+                },
+
+                {                    
                     "role": "user", 
                     "content": [
                         {

lollms_client/llm_bindings/tensor_rt/__init__.py

@@ -331,6 +331,7 @@ class VLLMBinding(LollmsLLMBinding):
     def generate_text(self, 
                      prompt: str,
                      images: Optional[List[str]] = None, 
+                     system_prompt: str = "",
                      n_predict: Optional[int] = 1024, 
                      stream: bool = False, # vLLM's generate is blocking, stream is pseudo
                      temperature: float = 0.7,
@@ -381,7 +382,7 @@ class VLLMBinding(LollmsLLMBinding):
                 # If providing multi_modal_data, usually prompt_token_ids are also needed.
                 # This can get complex as it depends on how the model expects images to be interleaved.
                 # For a simple case where image comes first:
-                encoded_prompt_ids = self.tokenizer.encode(prompt)
+                encoded_prompt_ids = self.tokenizer.encode(system_prompt+"\n"+prompt if system_prompt else prompt)
                 gen_kwargs["prompt_token_ids"] = [encoded_prompt_ids] # List of lists
                 gen_kwargs["multi_modal_data"] = [{"image": mm_data_content}] # List of dicts
                 gen_kwargs["prompts"] = None # Don't use prompts if prompt_token_ids is used
@@ -389,7 +390,7 @@ class VLLMBinding(LollmsLLMBinding):
             except Exception as e_mm:
                 return {"status": False, "error": f"Multimodal prep error: {e_mm}"}
         else:
-            gen_kwargs["prompts"] = [prompt]
+            gen_kwargs["prompts"] = [system_prompt+"\n"+prompt if system_prompt else prompt]
 
         try:
             outputs = self.llm_engine.generate(**gen_kwargs, sampling_params=sampling_params)

lollms_client/llm_bindings/transformers/__init__.py

@@ -112,6 +112,7 @@ class TransformersBinding(LollmsLLMBinding):
     def generate_text(self, 
                       prompt: str,
                       images: Optional[List[str]] = None,
+                      system_prompt: str = "",
                       n_predict: Optional[int] = None,
                       stream: bool = False,
                       temperature: float = 0.1,
@@ -123,8 +124,7 @@ class TransformersBinding(LollmsLLMBinding):
                       n_threads: int = 8,
                       ctx_size: int | None = None,
                       streaming_callback: Optional[Callable[[str, str], None]] = None,
-                      return_legacy_cache: bool = False,
-                      system_prompt: str = "You are a helpful assistant.") -> Union[str, dict]:
+                      return_legacy_cache: bool = False) -> Union[str, dict]:
         """
         Generate text using the Transformers model, with optional image support.
 

lollms_client/llm_bindings/vllm/__init__.py

@@ -331,6 +331,7 @@ class VLLMBinding(LollmsLLMBinding):
     def generate_text(self, 
                      prompt: str,
                      images: Optional[List[str]] = None, 
+                     system_prompt: str = "",
                      n_predict: Optional[int] = 1024, 
                      stream: bool = False, # vLLM's generate is blocking, stream is pseudo
                      temperature: float = 0.7,
@@ -381,7 +382,7 @@ class VLLMBinding(LollmsLLMBinding):
                 # If providing multi_modal_data, usually prompt_token_ids are also needed.
                 # This can get complex as it depends on how the model expects images to be interleaved.
                 # For a simple case where image comes first:
-                encoded_prompt_ids = self.tokenizer.encode(prompt)
+                encoded_prompt_ids = self.tokenizer.encode(system_prompt+"\n"+prompt if system_prompt else prompt)
                 gen_kwargs["prompt_token_ids"] = [encoded_prompt_ids] # List of lists
                 gen_kwargs["multi_modal_data"] = [{"image": mm_data_content}] # List of dicts
                 gen_kwargs["prompts"] = None # Don't use prompts if prompt_token_ids is used
@@ -389,7 +390,7 @@ class VLLMBinding(LollmsLLMBinding):
             except Exception as e_mm:
                 return {"status": False, "error": f"Multimodal prep error: {e_mm}"}
         else:
-            gen_kwargs["prompts"] = [prompt]
+            gen_kwargs["prompts"] = [system_prompt+"\n"+prompt if system_prompt else prompt]
 
         try:
             outputs = self.llm_engine.generate(**gen_kwargs, sampling_params=sampling_params)

lollms_client/lollms_core.py

@@ -329,6 +329,7 @@ class LollmsClient():
     def generate_text(self,
                      prompt: str,
                      images: Optional[List[str]] = None,
+                     system_prompt: str = "",
                      n_predict: Optional[int] = None,
                      stream: Optional[bool] = None,
                      temperature: Optional[float] = None,
@@ -365,6 +366,7 @@ class LollmsClient():
             return self.binding.generate_text(
                 prompt=prompt,
                 images=images,
+                system_prompt=system_prompt,
                 n_predict=n_predict if n_predict is not None else self.default_n_predict,
                 stream=stream if stream is not None else self.default_stream,
                 temperature=temperature if temperature is not None else self.default_temperature,
@@ -438,14 +440,12 @@ class LollmsClient():
         Uses the underlying LLM binding via `generate_text`.
         """
         response_full = ""
-        full_prompt = f"""{self.system_full_header}Act as a code generation assistant that generates code from user prompt.
-{self.user_full_header}
-{prompt}
-"""
+        system_prompt = f"""Act as a code generation assistant that generates code from user prompt."""
+
         if template:
-            full_prompt += "Here is a template of the answer:\n"
+            system_prompt += "Here is a template of the answer:\n"
             if code_tag_format=="markdown":
-                full_prompt += f"""You must answer with the code placed inside the markdown code tag like this:
+                system_prompt += f"""You must answer with the code placed inside the markdown code tag like this:
 ```{language}
 {template}
 ```
@@ -454,7 +454,7 @@ The code tag is mandatory.
 Don't forget encapsulate the code inside a markdown code tag. This is mandatory.
 """
             elif code_tag_format=="html":
-                full_prompt +=f"""You must answer with the code placed inside the html code tag like this:
+                system_prompt +=f"""You must answer with the code placed inside the html code tag like this:
 <code language="{language}">
 {template}
 </code>
@@ -462,13 +462,13 @@ Don't forget encapsulate the code inside a markdown code tag. This is mandatory.
 The code tag is mandatory.
 Don't forget encapsulate the code inside a html code tag. This is mandatory.
 """
-        full_prompt += f"""Do not split the code in multiple tags.
-{self.ai_full_header}"""
+        system_prompt += f"""Do not split the code in multiple tags."""
 
         # Use generate_text which handles images internally
         response = self.generate_text(
-            full_prompt,
+            prompt,
             images=images,
+            system_prompt=system_prompt,
             n_predict=max_size,
             temperature=temperature,
             top_k=top_k,
@@ -507,14 +507,12 @@ Don't forget encapsulate the code inside a html code tag. This is mandatory.
         Handles potential continuation if the code block is incomplete.
         """
 
-        full_prompt = f"""{self.system_full_header}Act as a code generation assistant that generates code from user prompt.
-{self.user_full_header}
-{prompt}
-"""
+        system_prompt = f"""{self.system_full_header}Act as a code generation assistant that generates code from user prompt."""
+
         if template:
-            full_prompt += "Here is a template of the answer:\n"
+            system_prompt += "Here is a template of the answer:\n"
             if code_tag_format=="markdown":
-                full_prompt += f"""You must answer with the code placed inside the markdown code tag like this:
+                system_prompt += f"""You must answer with the code placed inside the markdown code tag like this:
 ```{language}
 {template}
 ```
@@ -523,7 +521,7 @@ The code tag is mandatory.
 Don't forget encapsulate the code inside a markdown code tag. This is mandatory.
 """
             elif code_tag_format=="html":
-                full_prompt +=f"""You must answer with the code placed inside the html code tag like this:
+                system_prompt +=f"""You must answer with the code placed inside the html code tag like this:
 <code language="{language}">
 {template}
 </code>
@@ -531,13 +529,14 @@ Don't forget encapsulate the code inside a markdown code tag. This is mandatory.
 The code tag is mandatory.
 Don't forget encapsulate the code inside a html code tag. This is mandatory.
 """
-        full_prompt += f"""You must return a single code tag.
+        system_prompt += f"""You must return a single code tag.
 Do not split the code in multiple tags.
 {self.ai_full_header}"""
 
         response = self.generate_text(
-            full_prompt,
+            prompt,
             images=images,
+            system_prompt=system_prompt,
             n_predict=max_size,
             temperature=temperature,
             top_k=top_k,

lollms_client/lollms_llm_binding.py

@@ -32,6 +32,7 @@ class LollmsLLMBinding(ABC):
     def generate_text(self, 
                      prompt: str,
                      images: Optional[List[str]] = None,
+                     system_prompt: str = "",
                      n_predict: Optional[int] = None,
                      stream: bool = False,
                      temperature: float = 0.1,

{lollms_client-0.13.2.dist-info → lollms_client-0.14.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lollms_client
-Version: 0.13.2
+Version: 0.14.1
 Summary: A client library for LoLLMs generate endpoint
 Author-email: ParisNeo <parisneoai@gmail.com>
 License: Apache Software License

{lollms_client-0.13.2.dist-info → lollms_client-0.14.1.dist-info}/RECORD

@@ -3,7 +3,8 @@ examples/simple_text_gen_with_image_test.py,sha256=Euv53jbKTVJDvs854lgJvA5F-iRnA
 examples/text_2_audio.py,sha256=MfL4AH_NNwl6m0I0ywl4BXRZJ0b9Y_9fRqDIe6O-Sbw,3523
 examples/text_2_image.py,sha256=Ri7lQ-GW54YWQh2eofcaN6LpwFoorbpJsJffrcXl3cg,6415
 examples/text_and_image_2_audio.py,sha256=QLvSsLff8VZZa7k7K1EFGlPpQWZy07zM4Fnli5btAl0,2074
-examples/text_gen.py,sha256=S1UIXi3Aj8gTAmwHXtg-qUfgSkyCJFDLVKAbiOfsYGs,773
+examples/text_gen.py,sha256=IejpNmIlsfz3WpJg8IRm5X6F06JKd7h_GuonUxTITx8,758
+examples/text_gen_system_prompt.py,sha256=wumwZ09WZkaK0tQ74KaZmfsYXcmjZIlsdim_P1aJmeA,910
 examples/article_summary/article_summary.py,sha256=CR8mCBNcZEVCR-q34uOmrJyMlG-xk4HkMbsV-TOZEnk,1978
 examples/deep_analyze/deep_analyse.py,sha256=fZNmDrfEAuxEAfdbjAgJYIh1k6wbiuZ4RvwHRvtyUs8,971
 examples/deep_analyze/deep_analyze_multiple_files.py,sha256=fOryShA33P4IFxcxUDe-nJ2kW0v9w9yW8KsToS3ETl8,1032
@@ -12,13 +13,13 @@ examples/personality_test/chat_test.py,sha256=o2jlpoddFc-T592iqAiA29xk3x27KsdK5D
 examples/personality_test/chat_with_aristotle.py,sha256=4X_fwubMpd0Eq2rCReS2bgVlUoAqJprjkLXk2Jz6pXU,1774
 examples/personality_test/tesks_test.py,sha256=7LIiwrEbva9WWZOLi34fsmCBN__RZbPpxoUOKA_AtYk,1924
 examples/test_local_models/local_chat.py,sha256=slakja2zaHOEAUsn2tn_VmI4kLx6luLBrPqAeaNsix8,456
-lollms_client/__init__.py,sha256=y-N8Dw10pI9pHtP_zlVzsj7bVjsu873EdOlbqxaxZRU,823
+lollms_client/__init__.py,sha256=Nu5NJl6GDDW4W-hoRwxXQZdnHj2-fgCeaRW3Dt_zuCQ,823
 lollms_client/lollms_config.py,sha256=goEseDwDxYJf3WkYJ4IrLXwg3Tfw73CXV2Avg45M_hE,21876
-lollms_client/lollms_core.py,sha256=ZTbEVn1M_gHAL3mL5mf3wGYAXidAtnSI3qEjwz2HlwY,77980
+lollms_client/lollms_core.py,sha256=aFW8KDnBJRSL8a7ibhKAV67O95qvPT6PneNHQL1y2vc,78049
 lollms_client/lollms_discussion.py,sha256=9b83m0D894jwpgssWYTQHbVxp1gJoI-J947Ui_dRXII,2073
 lollms_client/lollms_functions.py,sha256=p8SFtmEPqvVCsIz2fZ5HxyOHaxjrAo5c12uTzJnb6m8,3594
 lollms_client/lollms_js_analyzer.py,sha256=01zUvuO2F_lnUe_0NLxe1MF5aHE1hO8RZi48mNPv-aw,8361
-lollms_client/lollms_llm_binding.py,sha256=vAFmGtdIB97nQwQzNauFowopH4qT2i-CS_-ckekY3V0,7361
+lollms_client/lollms_llm_binding.py,sha256=7xvtLsFQYqFKS7m0BQQMvVq0XXZWZeGlGuv30mi1dF8,7408
 lollms_client/lollms_python_analyzer.py,sha256=7gf1fdYgXCOkPUkBAPNmr6S-66hMH4_KonOMsADASxc,10246
 lollms_client/lollms_stt_binding.py,sha256=ovmpFF0fnmPC9VNi1-rxAJA8xI4JZDUBh_YwdtoTx28,5818
 lollms_client/lollms_tasks.py,sha256=Tgqces03gPTHFJCcPaeN9vBCsil3SSJX7nQAjCQ2-yg,34393
@@ -29,12 +30,12 @@ lollms_client/lollms_ttv_binding.py,sha256=u-gLIe22tbu4YsKA5RTyUT7iBlKxPXDmoQzcc
 lollms_client/lollms_types.py,sha256=cfc1sremM8KR4avkYX99fIVkkdRvXErrCWKGjLrgv50,2723
 lollms_client/lollms_utilities.py,sha256=YAgamfp0pBVApR68AHKjhp1lh6isMNF8iadwWLl63c0,7045
 lollms_client/llm_bindings/__init__.py,sha256=9sWGpmWSSj6KQ8H4lKGCjpLYwhnVdL_2N7gXCphPqh4,14
-lollms_client/llm_bindings/lollms/__init__.py,sha256=l1q2KnMQALz9QpLa3OUQ8e29KU4RCwkrmrdBvd7Z_kc,12236
-lollms_client/llm_bindings/ollama/__init__.py,sha256=DyueED1cJmmJFg5evYmu-lrkwsN9pAxaVcwgUkcAZHU,26467
-lollms_client/llm_bindings/openai/__init__.py,sha256=SWBgnOcOWmFRSKTN1S9ATownHNBJ9f6FEtI3L4xNJNM,11861
-lollms_client/llm_bindings/tensor_rt/__init__.py,sha256=ZpeSKAbN8rh6zkysYl95sXG9Ci702NuPAhXC6zb1zT4,31840
-lollms_client/llm_bindings/transformers/__init__.py,sha256=8JbX3B-obLt5NNtcNOGD_E0f8OQTma2pNYtVt2urTOM,12572
-lollms_client/llm_bindings/vllm/__init__.py,sha256=54TxuHJhlujVxPN03BgfeUecIexTjvU6g1dVmmaDgtM,31824
+lollms_client/llm_bindings/lollms/__init__.py,sha256=a36AMPFEf3xK4zx1M_L9PC-3-b0iiDf7eyLkknPjgaY,12356
+lollms_client/llm_bindings/ollama/__init__.py,sha256=MemSA20Zivn-kfP11JPA9FHigC1U2CGsJ1FaDtUFUUM,26574
+lollms_client/llm_bindings/openai/__init__.py,sha256=NDZIdzW0pnHy9gPXSKfFyS6SPIOOxj9ZEzEE7gZT2NQ,12054
+lollms_client/llm_bindings/tensor_rt/__init__.py,sha256=IY4CrHVpHY77R1rzsl3iwcoarDjYD24n7bFKk_69PD8,31983
+lollms_client/llm_bindings/transformers/__init__.py,sha256=IWfAmBGqZEelt5Z_jYTqpz7LzzKMVsKWx5nv4zBgKCQ,12544
+lollms_client/llm_bindings/vllm/__init__.py,sha256=ZRCR7g3A2kHQ_07viNrNnVHoIGj5TNA4Q41rQWeTlxw,31967
 lollms_client/stt_bindings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 lollms_client/stt_bindings/lollms/__init__.py,sha256=7-IZkrsn15Vaz0oqkqCxMeNQfMkeilbgScLlrrywES4,6098
 lollms_client/tti_bindings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -45,8 +46,8 @@ lollms_client/tts_bindings/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJ
 lollms_client/tts_bindings/lollms/__init__.py,sha256=8x2_T9XscvISw2TiaLoFxvrS7TIsVLdqbwSc04cX-wc,7164
 lollms_client/ttv_bindings/__init__.py,sha256=UZ8o2izQOJLQgtZ1D1cXoNST7rzqW22rL2Vufc7ddRc,3141
 lollms_client/ttv_bindings/lollms/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-lollms_client-0.13.2.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
-lollms_client-0.13.2.dist-info/METADATA,sha256=IsZiVKLRi7NaMhITn6pFz3CX5zpIJAvb8vsddYIBJQA,7276
-lollms_client-0.13.2.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
-lollms_client-0.13.2.dist-info/top_level.txt,sha256=NI_W8S4OYZvJjb0QWMZMSIpOrYzpqwPGYaklhyWKH2w,23
-lollms_client-0.13.2.dist-info/RECORD,,
+lollms_client-0.14.1.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
+lollms_client-0.14.1.dist-info/METADATA,sha256=Z6n3yPUXNYMjZa0IVrhBQDyvXHRbvi7yQwaqaa1gptY,7276
+lollms_client-0.14.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+lollms_client-0.14.1.dist-info/top_level.txt,sha256=NI_W8S4OYZvJjb0QWMZMSIpOrYzpqwPGYaklhyWKH2w,23
+lollms_client-0.14.1.dist-info/RECORD,,

{lollms_client-0.13.2.dist-info → lollms_client-0.14.1.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.7.1)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any