aait 2.3.12.1__tar.gz → 2.3.12.3__tar.gz

aait-2.3.12.3/PKG-INFO

@@ -0,0 +1,32 @@
+Metadata-Version: 2.1
+Name: aait
+Version: 2.3.12.3
+Summary: Advanced Artificial Intelligence Tools is a package meant to develop and enable advanced AI functionalities in Orange
+Home-page: 
+Author: Orange community
+Author-email: 
+License: UNKNOWN
+Keywords: orange3 add-on
+Platform: UNKNOWN
+License-File: License.txt
+Requires-Dist: torch
+Requires-Dist: sentence-transformers==5.0.0
+Requires-Dist: gpt4all[all]==2.8.2
+Requires-Dist: sacremoses==0.1.1
+Requires-Dist: transformers==4.51.3
+Requires-Dist: sentencepiece==0.2.0
+Requires-Dist: optuna
+Requires-Dist: spacy==3.7.6
+Requires-Dist: markdown
+Requires-Dist: python-multipart
+Requires-Dist: PyMuPDF==1.24.14
+Requires-Dist: chonkie==0.4.1
+Requires-Dist: GPUtil==1.4.0
+Requires-Dist: unidecode==1.3.8
+Requires-Dist: python-docx==1.1.2
+Requires-Dist: psutil
+Requires-Dist: thefuzz==0.22.1
+Requires-Dist: beautifulsoup4==4.12.3
+Requires-Dist: CATEGORIT
+
+UNKNOWN

aait-2.3.12.3/aait.egg-info/PKG-INFO

@@ -0,0 +1,32 @@
+Metadata-Version: 2.1
+Name: aait
+Version: 2.3.12.3
+Summary: Advanced Artificial Intelligence Tools is a package meant to develop and enable advanced AI functionalities in Orange
+Home-page: 
+Author: Orange community
+Author-email: 
+License: UNKNOWN
+Keywords: orange3 add-on
+Platform: UNKNOWN
+License-File: License.txt
+Requires-Dist: torch
+Requires-Dist: sentence-transformers==5.0.0
+Requires-Dist: gpt4all[all]==2.8.2
+Requires-Dist: sacremoses==0.1.1
+Requires-Dist: transformers==4.51.3
+Requires-Dist: sentencepiece==0.2.0
+Requires-Dist: optuna
+Requires-Dist: spacy==3.7.6
+Requires-Dist: markdown
+Requires-Dist: python-multipart
+Requires-Dist: PyMuPDF==1.24.14
+Requires-Dist: chonkie==0.4.1
+Requires-Dist: GPUtil==1.4.0
+Requires-Dist: unidecode==1.3.8
+Requires-Dist: python-docx==1.1.2
+Requires-Dist: psutil
+Requires-Dist: thefuzz==0.22.1
+Requires-Dist: beautifulsoup4==4.12.3
+Requires-Dist: CATEGORIT
+
+UNKNOWN

{aait-2.3.12.1 → aait-2.3.12.3}/aait.egg-info/SOURCES.txt

@@ -50,6 +50,8 @@ orangecontrib/AAIT/widgets/OWAAITResourcesManager.py
 orangecontrib/AAIT/widgets/OWAccumulator.py
 orangecontrib/AAIT/widgets/OWAddColumns.py
 orangecontrib/AAIT/widgets/OWApplyRules.py
+orangecontrib/AAIT/widgets/OWAutoShowCreateInstance.py
+orangecontrib/AAIT/widgets/OWAutoShowTable.py
 orangecontrib/AAIT/widgets/OWCN2rule_view.py
 orangecontrib/AAIT/widgets/OWChunking.py
 orangecontrib/AAIT/widgets/OWConcatRules.py
@@ -101,6 +103,7 @@ orangecontrib/AAIT/widgets/OWOptimisation.py
 orangecontrib/AAIT/widgets/OWOptimisationSelection.py
 orangecontrib/AAIT/widgets/OWOptimisationSendScore.py
 orangecontrib/AAIT/widgets/OWProcessDocumentsFromPath.py
+orangecontrib/AAIT/widgets/OWQuadrantclicker.py
 orangecontrib/AAIT/widgets/OWQueryLLM.py
 orangecontrib/AAIT/widgets/OWRandomData.py
 orangecontrib/AAIT/widgets/OWReranking.py
@@ -115,8 +118,6 @@ orangecontrib/AAIT/widgets/OWTrigger.py
 orangecontrib/AAIT/widgets/OWUnzipFolder.py
 orangecontrib/AAIT/widgets/POW_Wfactory.py
 orangecontrib/AAIT/widgets/__init__.py
-orangecontrib/AAIT/widgets/owautoshowcreateinstance.py
-orangecontrib/AAIT/widgets/owautoshowtable.py
 orangecontrib/AAIT/widgets/designer/owMarkdown.ui
 orangecontrib/AAIT/widgets/designer/ow_OperationSystem.ui
 orangecontrib/AAIT/widgets/designer/ow_in_or_out_path.ui
@@ -170,6 +171,7 @@ orangecontrib/AAIT/widgets/designer/owmodel_spacymd_en.ui
 orangecontrib/AAIT/widgets/designer/owmodel_spacymd_fr.ui
 orangecontrib/AAIT/widgets/designer/ownumberpointinrules.ui
 orangecontrib/AAIT/widgets/designer/owprocessdocuments.ui
+orangecontrib/AAIT/widgets/designer/owquadrant_clicker.ui
 orangecontrib/AAIT/widgets/designer/owqueryllm.ui
 orangecontrib/AAIT/widgets/designer/owreranking.ui
 orangecontrib/AAIT/widgets/designer/owsavewithpath.ui
@@ -255,6 +257,7 @@ orangecontrib/AAIT/widgets/icons/owtable2corpus.svg
 orangecontrib/AAIT/widgets/icons/owtranslation.svg
 orangecontrib/AAIT/widgets/icons/owtrigger.svg
 orangecontrib/AAIT/widgets/icons/processdocuments.svg
+orangecontrib/AAIT/widgets/icons/quadrantclicker.svg
 orangecontrib/AAIT/widgets/icons/qwen-color.png
 orangecontrib/AAIT/widgets/icons/select_dynamic_row.png
 orangecontrib/AAIT/widgets/icons/startloop.png

{aait-2.3.12.1 → aait-2.3.12.3}/orangecontrib/AAIT/__init__.py

@@ -93,7 +93,10 @@ else: # Execute the file
         """
         gc.collect()
         objects = gc.get_objects()
-        widget_discoveries = [obj for obj in objects if isinstance(obj, WidgetDiscovery)]
+        try:
+            widget_discoveries = [obj for obj in objects if isinstance(obj, WidgetDiscovery)]
+        except Exception as e:
+            print(e)
         return widget_discoveries
 
     def import_proprietary_categories():

{aait-2.3.12.1 → aait-2.3.12.3}/orangecontrib/AAIT/llm/answers_llama.py

@@ -142,7 +142,7 @@ def load_model(model_path, use_gpu, n_ctx=10000):
         return
 
 
-def generate_answers(table, model_path, use_gpu=False, n_ctx=4096, query_parameters=None, workflow_id="", progress_callback=None):
+def generate_answers(table, model_path, use_gpu=False, n_ctx=4096, query_parameters=None, workflow_id="", progress_callback=None, argself=None):
     """
     Identique en signature/comportement, mais utilise llama_cpp sous le capot.
     """
@@ -179,6 +179,8 @@ def generate_answers(table, model_path, use_gpu=False, n_ctx=4096, query_paramet
                 system_prompt=system_prompt
             )
 
+            prompt = handle_context_length(prompt, model, n_ctx, method="truncate", progress_callback=progress_callback)
+
             answer = run_query(
                 prompt,
                 model=model,
@@ -208,7 +210,7 @@ def generate_answers(table, model_path, use_gpu=False, n_ctx=4096, query_paramet
 
             if progress_callback is not None:
                 progress_value = float(100 * (i + 1) / len(data))
-                progress_callback((progress_value, "\n\n\n\n"))
+                progress_callback(("progressBar", progress_value))
 
             if argself is not None and getattr(argself, "stop", False):
                 break
@@ -305,7 +307,7 @@ def run_query(prompt, model, max_tokens=4096, temperature=0, top_p=0, top_k=40,
             write_tokens_to_file(token, workflow_id)
 
             if progress_callback is not None:
-                progress_callback((None, token))
+                progress_callback(("assistant", token))
 
             if argself is not None and getattr(argself, "stop", False):
                 # Arrêt demandé de l'extérieur
@@ -323,7 +325,7 @@ def run_query(prompt, model, max_tokens=4096, temperature=0, top_p=0, top_k=40,
     return answer
 
 
-def generate_conversation(table, model, conversation=None, progress_callback=None, argself=None):
+def generate_conversation(table, model, conversation=None, n_ctx=32768, progress_callback=None, argself=None):
     """
     Generates a response using a language model and appends it to a conversation.
 
@@ -362,26 +364,27 @@ def generate_conversation(table, model, conversation=None, progress_callback=Non
     prompt = ""
     # If the conversation has not started, add a default system prompt
     if not conversation:
-        prompt += prompt_management.apply_system_template(model_path=model_path, system_prompt=system_prompt)
-        conversation = []
+        conversation = [{"role": "system", "content": system_prompt}]
 
     # If the conversation has started, build the complete context
     else:
         for message in conversation:
-            if message["role"] == "user":
+            if message["role"] == "system":
+                prompt += prompt_management.apply_system_template(model_path, message["content"])
+            elif message["role"] == "user":
                 prompt += prompt_management.apply_user_template(model_path, message["content"])
             elif message["role"] == "assistant":
                 prompt += prompt_management.apply_assistant_template(model_path, message["content"])
-            elif message["role"] == "system":
-                prompt += prompt_management.apply_system_template(model_path, message["content"])
 
     # Add the current user prompt & assistant prompt
     prompt += prompt_management.apply_user_template(model_path, user_prompt)
     prompt += prompt_management.apply_assistant_template(model_path, assistant_prompt)
 
+    prompt = handle_context_length(prompt, model, n_ctx, method="truncate", progress_callback=progress_callback)
+
     # Return progression to fill the user & assistant cards
     progress_callback(("user", user_prompt))
-    progress_callback((None, assistant_prompt))
+    progress_callback(("assistant", assistant_prompt))
 
     # Append the user message to the conversation
     conversation.append({"role": "user", "content": user_prompt})
@@ -395,8 +398,8 @@ def generate_conversation(table, model, conversation=None, progress_callback=Non
     # Append the answer to the conversation
     conversation.append({"role": "assistant", "content": assistant_prompt + final_answer})
 
-    # Add spaces to the widget for following answers
-    progress_callback((100, ""))
+    # End of the progress bar
+    progress_callback(("progressBar", 100))
 
     # Padding to add columns to the table, TEMPORARY # TODO : process the entire table for conversation
     rows_answer = [final_answer]
@@ -427,8 +430,37 @@ def split_think(answer: str):
     final_answer = re.sub(r"<think>.*?</think>", "", answer, flags=re.DOTALL).strip()
     return think_text, final_answer
 
+
 def conversation_to_text(conversation):
     if not conversation:
         return ""
     else:
-        pass #TODO
+        pass #TODO
+
+
+def handle_context_length(prompt, model, n_ctx, method="truncate", margin=0, progress_callback=None):
+    """
+    Truncate a prompt to fit within n_ctx tokens, leaving margin for generation.
+    Safely handles edge cases where limit <= 0.
+    """
+    # Keep a margin for generated tokens
+    limit = max(n_ctx - margin, 0)  # clamp to at least 0
+
+    if method == "truncate":
+        tokens = model.tokenize(prompt.encode("utf-8"))  # pass string, not bytes
+        initial_length = len(tokens)
+        if initial_length > limit:
+            # take last `limit` tokens safely
+            tokens = tokens[-limit:] if limit > 0 else []
+            truncated_length = len(tokens)
+            prompt = model.detokenize(tokens).decode("utf-8") if tokens else ""
+            if progress_callback:
+                warning = (
+                    f"Complete prompt contains {initial_length} tokens - context limit is {limit}. "
+                    f"The {truncated_length} last tokens have been kept in the prompt."
+                )
+                progress_callback(("warning", warning))
+        return prompt
+    else:
+        return prompt
+

{aait-2.3.12.1 → aait-2.3.12.3}/orangecontrib/AAIT/llm/lmstudio.py

@@ -47,7 +47,7 @@ def appel_lmstudio(prompt, model, stream=False, temperature=0.7, max_tokens=4096
 
         # Vérification de la réponse
         if response.status_code != 200:
-            raise Exception("Erreur dans la requête")
+            raise Exception("Error in the request")
 
         # Initialisation de la variable de contenu
         content = ""
@@ -75,6 +75,6 @@ def appel_lmstudio(prompt, model, stream=False, temperature=0.7, max_tokens=4096
         return content
 
     except Exception as e:
-        print(f"Erreur : {e}")
+        print(f"Error : {e}")
         return None
 

aait-2.3.12.3/orangecontrib/AAIT/llm/prompt_management.py

@@ -0,0 +1,157 @@
+import os
+
+prompt_templates = {
+    "llama": {
+        "system": "<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n{system_prompt}<|eot_id|>",
+        "user": "<|start_header_id|>user<|end_header_id|>\n{user_prompt}<|eot_id|>",
+        "assistant": "<|start_header_id|>assistant<|end_header_id|>\n{assistant_prompt}"
+    },
+
+    "mistral": {
+        "system": "{system_prompt}\n",
+        "user": "<s>[INST] {user_prompt} [/INST]</s>\n",
+        "assistant": "{assistant_prompt}"
+    },
+
+    "solar": {
+        "system": "{system_prompt}\n",
+        "user": "### User: {user_prompt}\n",
+        "assistant": "### Assistant: {assistant_prompt}"
+    },
+
+    "deepseek": {
+        "system": "{system_prompt}\n",
+        "user": "### Instruction: {user_prompt}\n",
+        "assistant": "### Response: {assistant_prompt}"
+    },
+
+    "qwen": {
+        "system": "<|im_start|>system\n{system_prompt}<|im_end|>\n",
+        "user": "<|im_start|>user\n{user_prompt}<|im_end|>\n",
+        "assistant": "<|im_start|>assistant\n{assistant_prompt}"
+    },
+
+    "gemma": {
+        "system": "{system_prompt}\n",
+        "user": "<start_of_turn>user\n{user_prompt}<end_of_turn>\n",
+        "assistant": "<start_of_turn>model\n{assistant_prompt}\n"
+    },
+
+    "granite": {
+        "system": "<|system|>\n{system_prompt}\n",
+        "user": "<|user|>\n{user_prompt}\n",
+        "assistant": "<|assistant|>\n{assistant_prompt}"
+    },
+
+    "phi": {
+        "system": "<|im_start|>system<|im_sep|>\n{system_prompt}<|im_end|>\n",
+        "user": "<|im_start|>user<|im_sep|>\n{user_prompt}<|im_end|>\n",
+        "assistant": "<|im_start|>assistant<|im_sep|>\n{assistant_prompt}"
+    },
+
+    "default": {
+        "system": "{system_prompt}",
+        "user": "{user_prompt}",
+        "assistant": "{assistant_prompt}"
+    }
+}
+
+
+
+stop_tokens = {
+    "llama": "<|eot_id|>",
+    "mistral": "</s>",
+    "qwen": "<|im_end|>",
+    "gemma": "<end_of_turn>",
+    "granite": "<|endoftext|>",
+    "phi": "<|im_end|>"
+}
+
+
+model_types = {
+    "solar-10.7b-instruct-v1.0.Q6_K.gguf": "solar",
+    "solar-10.7b-instruct-v1.0-uncensored.Q6_K.gguf": "solar",
+    "Mistral-7B-Instruct-v0.3.Q6_K.gguf": "mistral",
+    "Qwen2.5.1-Coder-7B-Instruct-Q6_K.gguf": "qwen",
+    "qwen2.5-3b-instruct-q4_k_m.gguf": "qwen",
+    "deepseek-coder-6.7b-instruct.Q6_K.gguf": "deepseek"
+}
+
+model_keywords = ["qwen", "solar", "mistral", "llama", "deepseek", "gemma", "granite", "phi"]
+
+
+
+def get_model_type(model_path):
+    model_name = os.path.basename(model_path)
+    model_type = model_types.get(model_name)
+    if model_type is None:
+        model_type = next((keyword for keyword in model_keywords if keyword in model_name.lower()), None)
+    if model_type is None:
+        model_type = "default"
+    return model_type
+
+
+def apply_prompt_template(model_path, user_prompt, assistant_prompt="", system_prompt=""):
+    """
+    Apply a prompt template based on the given model name and user input.
+
+    Parameters:
+        model_path (str): The name of the model used to determine its type.
+        user_prompt (str): The user input or request to embed into the prompt.
+        assistant_prompt (str, optional): The assistant's beginning of response, if any. Defaults to an empty string.
+        system_prompt (str, optional): A system-level instruction or context to include in the prompt. Defaults to an empty string.
+
+    Returns:
+        str: The formatted prompt that is ready to be passed to the model.
+    """
+    # Try to identify the model's type
+    model_type = get_model_type(model_path)
+    # Retrieve the template
+    template = prompt_templates.get(model_type, prompt_templates["default"])  # Default template if none found
+    # Apply the template
+    prompt = ""
+    if system_prompt is not None:
+        prompt += template["system"].format(system_prompt=system_prompt)
+    prompt += template["user"].format(user_prompt=user_prompt)
+    prompt += template["assistant"].format(assistant_prompt=assistant_prompt)
+    return prompt
+
+
+def apply_system_template(model_path, system_prompt):
+    model_type = get_model_type(model_path)
+    template = prompt_templates.get(model_type, prompt_templates["default"])
+    return template["system"].format(system_prompt=system_prompt)
+
+def apply_user_template(model_path, user_prompt):
+    model_type = get_model_type(model_path)
+    template = prompt_templates.get(model_type, prompt_templates["default"])
+    return template["user"].format(user_prompt=user_prompt)
+
+def apply_assistant_template(model_path, assistant_prompt):
+    model_type = get_model_type(model_path)
+    template = prompt_templates.get(model_type, prompt_templates["default"])
+    return template["assistant"].format(assistant_prompt=assistant_prompt)
+
+
+def get_stop_token(model_name):
+    """
+    Get the stop token according to the model name / type.
+
+    Parameters:
+        model_name (str): The name of the model used to determine its type.
+    """
+    # If there is a stop token
+    try:
+        # Get the model type
+        model_type = model_types[model_name]
+        # Get the template for the model type
+        stop_token = stop_tokens[model_type]
+    except KeyError as e:
+        print(f"Your model {model_name} has no stop token defined. See prompt_management.py. (detail: {e})")
+        return None
+    return stop_token
+
+
+
+
+

{aait-2.3.12.1 → aait-2.3.12.3}/orangecontrib/AAIT/utils/shared_functions.py

@@ -21,8 +21,11 @@ def get_ow_file_widget():
 
     # Iterate through the objects in the garbage collector
     for obj in gc.get_objects():
-        if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
-            le_self = obj
+        try:
+            if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
+                le_self = obj
+        except Exception as e:
+            print(e)
     return le_self
 
 
@@ -43,29 +46,31 @@ def get_duplicate_names(only_aait=False):
     widgets_lists = []  # List to store lists of widgets for each OWS
 
     print("Collecting all widgets for every OWS")
-    for obj in gc.get_objects():
-        if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
-            ows_name = obj.current_document().path().split("/")[-1]
-            ows_list.append(ows_name)
-            widgets_list = []
-
-            # Extract information about widgets for the current OWS
-            item_dict = obj.scheme_widget._SchemeEditWidget__scheme.widget_manager._OWWidgetManager__item_for_node
-            for item in item_dict.items():
-                #widget = item[1].widget  # The widget is the 2nd element of the Tuple
-                widget_name = item[0].title
-                # print("  OWS :", ows_name, "ID :", id(obj))
-                # print("  Widget :", widget)
-                # print("  Widget ID:", id(widget))
-                # print("  Widget name:", widget_name)
-                category = item[0].description.category
-
-                # Targets only aait widgets or includes all widgets based on the 'only_aait' parameter
-                if category[0:9] == "dev_workflow" or not only_aait:
-                    widgets_list.append(widget_name)
-
-            widgets_lists.append(widgets_list)
-
+    try:
+        for obj in gc.get_objects():
+            if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
+                ows_name = obj.current_document().path().split("/")[-1]
+                ows_list.append(ows_name)
+                widgets_list = []
+
+                # Extract information about widgets for the current OWS
+                item_dict = obj.scheme_widget._SchemeEditWidget__scheme.widget_manager._OWWidgetManager__item_for_node
+                for item in item_dict.items():
+                    #widget = item[1].widget  # The widget is the 2nd element of the Tuple
+                    widget_name = item[0].title
+                    # print("  OWS :", ows_name, "ID :", id(obj))
+                    # print("  Widget :", widget)
+                    # print("  Widget ID:", id(widget))
+                    # print("  Widget name:", widget_name)
+                    category = item[0].description.category
+
+                    # Targets only aait widgets or includes all widgets based on the 'only_aait' parameter
+                    if category[0:9] == "dev_workflow" or not only_aait:
+                        widgets_list.append(widget_name)
+
+                widgets_lists.append(widgets_list)
+    except Exception as e:
+        print(e)
     all_duplicates = []  # List to store tuples of OWS names and sets of duplicate widget names
     n_ows = len(ows_list)
 
@@ -102,24 +107,27 @@ def get_current_ows_gui(arg_self):
     # print("----> Created Widget ID :", widget_id)
 
     # Loop over ALL instantiated objects
-    for obj in gc.get_objects():
-        # Stop whenever a MainWindow is found (OWS)
-        if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
-            # Dictionary containing all the existing widgets (inside Tuples)
-            item_dict = obj.scheme_widget._SchemeEditWidget__scheme.widget_manager._OWWidgetManager__item_for_node
+    try:
+        for obj in gc.get_objects():
+            # Stop whenever a MainWindow is found (OWS)
+            if isinstance(obj, Orange.canvas.mainwindow.MainWindow):
+                # Dictionary containing all the existing widgets (inside Tuples)
+                item_dict = obj.scheme_widget._SchemeEditWidget__scheme.widget_manager._OWWidgetManager__item_for_node
 
-            for item in item_dict.items():
-                # The widget (object) is the 2nd element of Tuple
-                widget = item[1].widget
+                for item in item_dict.items():
+                    # The widget (object) is the 2nd element of Tuple
+                    widget = item[1].widget
 
-                if id(widget) == widget_id:
-                    ows_path = obj.current_document().path()
+                    if id(widget) == widget_id:
+                        ows_path = obj.current_document().path()
 
-                    # If the OWS hasn't been named / saved, force save-as
-                    if not ows_path:
-                        MetManagement.force_save_as(arg_self, target_canvas_id=id(obj))
+                        # If the OWS hasn't been named / saved, force save-as
+                        if not ows_path:
+                            MetManagement.force_save_as(arg_self, target_canvas_id=id(obj))
 
-                    return obj
+                        return obj
+    except Exception as e:
+        print(e)
 
     return None
 
@@ -134,10 +142,13 @@ def get_current_ows_name_no_gui(arg_self):
     Returns:
     The name of the associated OWS or an empty string if not found.
     """
-    for obj in gc.get_objects():
-        if isinstance(obj, argparse.Namespace):
-            ows_path = obj.file
-            return ows_path
+    try:
+        for obj in gc.get_objects():
+            if isinstance(obj, argparse.Namespace):
+                ows_path = obj.file
+                return ows_path
+    except Exception as e:
+        print(e)
 
     return ""