syntaxmatrix 2.5.5.5__py3-none-any.whl → 2.5.6.1__py3-none-any.whl

syntaxmatrix/__init__.py

@@ -44,8 +44,8 @@ get_widget_value = _app_instance.get_widget_value
 save_embed_model = _app_instance.save_embed_model
 load_embed_model = _app_instance.load_embed_model
 delete_embed_key = _app_instance.delete_embed_key
-set_prompt_profile = _app_instance.set_prompt_profile
-set_prompt_instructions = _app_instance.set_prompt_instructions
+set_smxai_identity = _app_instance.set_smxai_identity
+set_smxai_instructions = _app_instance.set_smxai_instructions
 set_website_description = _app_instance.set_website_description
 smiv_index = _app_instance.smiv_index
 smpv_search  = _app_instance.smpv_search
@@ -54,6 +54,7 @@ process_query_stream = _app_instance.process_query_stream
 process_query = _app_instance.process_query
 embed_query = _app_instance.embed_query
 enable_user_files = _app_instance.enable_user_files
+enable_registration = _app_instance.enable_registration
 stream_write = _app_instance.stream_write
 enable_stream = _app_instance.enable_stream
 stream = _app_instance.stream

syntaxmatrix/agentic/agents.py

@@ -425,11 +425,12 @@ def refine_question_agent(raw_question: str, dataset_context: str | None = None)
 
         return "Configure LLM Profiles or contact your administrator."
 
-    system_prompt = (
-        "You rewrite user questions into specification Machine Learning (ML) job description. "
-        "If a dataset summary is provided, use it to respect column and help you redefine the question. "
-        "DO NOT write andy prelude or preamble"
-    )
+    system_prompt = ("""
+        - You are a Machine Learning (ML) and Data Science (DS) expert.
+        - You rewrite user questions into clear ML job specifications to help AI assistant generate Python code that provides solution to the user question when it is run. Most user questions are vague. So, your goal is to ensure that your output guards the assistant agains making potential errors that you anticipated could arise due to the nature of the question. 
+        - If a dataset summary is provided, use it to respect column and help you rewrite the question properly.
+        - DO NOT write andy prelude or preamble"
+    """)
 
     user_prompt = f"User question:\n{raw_question}\n\n"
     if dataset_context:
@@ -446,20 +447,7 @@ def refine_question_agent(raw_question: str, dataset_context: str | None = None)
   
 
 def classify_ml_job_agent(refined_question, dataset_profile):   
-    """
-    Instructs an LLM (gemini-2.5-flash) to analyze a task description 
-    and return a list of associated machine learning job/task types. 
-    This version uses a highly extensive, generalized list of ML jobs 
-    to ensure robustness across all domains (NLP, CV, RL, etc.).
-
-    Args:
-        task_description: The detailed description of the statistical/ML task.
-
-    Returns:
-        A list of strings identifying the relevant ML jobs. Returns an empty 
-        list if the API call fails or the output cannot be parsed.
-    """
-
+    
     def ml_response(user_prompt, system_prompt, profile):
         _profile = profile  # _prof.get_profile["admin"]
         
@@ -571,10 +559,13 @@ def classify_ml_job_agent(refined_question, dataset_profile):
 
         return "Configure LLM Profiles or contact your administrator."
 
-    system_prompt = (
-        "You are a strict machine learning task classifier for an ML workbench.\n"
-        "Your job is to label the user's task desc. with all relevant tags from a fixed list.\n\n"
-    )
+    system_prompt = ("""
+        You are a strict machine learning task classifier for an ML workbench.
+        Your goal is to correctly label the user's task specifications with the most relevant tags from a fixed list.
+        You Must always have 'data_preprocessing' as the 1st tag. Then add up to 4 to make 5 max. Your list, therefore, should have 1-5 tags. If you think a task is too complext for the given context, even if relevant, exclude it.
+        If no relevant tag, default to "data_preprocessing" and return that alone.
+        You should return only your list of tags, no prelude or preamble.
+    """)
 
     # --- 1. Define the Master List of ML Tasks (Generalized) ---
     ml_task_list = [

syntaxmatrix/auth.py

@@ -41,6 +41,19 @@ def init_auth_db():
     );
     """)
 
+    # Ensure new must_reset_password flag exists for mandatory first-login reset
+    try:
+        cur = conn.execute("PRAGMA table_info(users)")
+        cols = [row[1] for row in cur.fetchall()]
+        if "must_reset_password" not in cols:
+            conn.execute(
+                "ALTER TABLE users "
+                "ADD COLUMN must_reset_password INTEGER NOT NULL DEFAULT 0"
+            )
+    except Exception:
+        # Best-effort migration; if this fails we still let the app start
+        pass
+
     # --- Roles table ---
     conn.execute("""
     CREATE TABLE IF NOT EXISTS roles (
@@ -302,6 +315,60 @@ def register_user(email:str, username:str, password:str, role:str = "user") -> b
     finally:
         conn.close()
 
+def set_must_reset_by_email(email: str, must_reset: bool = True) -> None:
+    """
+    Mark a user account as requiring a password reset (or clear the flag) by email.
+    Used when an admin creates a user with a temporary password.
+    """
+    if not email:
+        return
+    conn = _get_conn()
+    try:
+        conn.execute(
+            "UPDATE users SET must_reset_password = ? WHERE email = ?",
+            (1 if must_reset else 0, email),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+
+def user_must_reset_password(user_id: int) -> bool:
+    """
+    Check whether this user is currently forced to change their password.
+    """
+    if not user_id:
+        return False
+    conn = _get_conn()
+    try:
+        cur = conn.execute(
+            "SELECT must_reset_password FROM users WHERE id = ?",
+            (user_id,),
+        )
+        row = cur.fetchone()
+    finally:
+        conn.close()
+    if not row:
+        return False
+    return bool(row[0])
+
+
+def clear_must_reset(user_id: int) -> None:
+    """
+    Clear the mandatory-reset flag (called after the user has changed their password).
+    """
+    if not user_id:
+        return
+    conn = _get_conn()
+    try:
+        conn.execute(
+            "UPDATE users SET must_reset_password = 0 WHERE id = ?",
+            (user_id,),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
 def authenticate(email:str, password:str) -> Optional[Dict]:
     """Return user dict if creds match, else None."""
     conn = _get_conn()
@@ -315,15 +382,85 @@ def authenticate(email:str, password:str) -> Optional[Dict]:
         return {"id": row[0], "email":row[1], "username": row[2], "role": row[4]}
     return None
 
+def verify_password(user_id: int, candidate: str) -> bool:
+    """
+    Check whether `candidate` matches the current password of the user.
+    Used by the change-password flow.
+    """
+    if not user_id or not candidate:
+        return False
+
+    conn = _get_conn()
+    try:
+        cur = conn.execute(
+            "SELECT password FROM users WHERE id = ?",
+            (user_id,),
+        )
+        row = cur.fetchone()
+    finally:
+        conn.close()
+
+    if not row:
+        return False
+    return check_password_hash(row[0], candidate)
+
+
+def update_password(user_id: int, new_password: str) -> None:
+    """
+    Overwrite the user's password with a new hash.
+    """
+    if not user_id or not new_password:
+        return
+
+    hashed = generate_password_hash(new_password)
+    conn = _get_conn()
+    try:
+        conn.execute(
+            "UPDATE users SET password = ? WHERE id = ?",
+            (hashed, user_id),
+        )
+        conn.commit()
+    finally:
+        conn.close()
+
+def update_password(user_id: int, new_password: str) -> bool:
+    """
+    Update the stored password hash for a given user id.
+    Returns True on success, False if something goes wrong.
+    """
+    hashed = generate_password_hash(new_password)
+    conn = _get_conn()
+    try:
+        conn.execute(
+            "UPDATE users SET password = ? WHERE id = ?",
+            (hashed, user_id),
+        )
+        conn.commit()
+        return True
+    except Exception:
+        # We do not raise inside auth; caller shows a friendly message instead.
+        return False
+    finally:
+        conn.close()
+
 def login_required(f):
     @wraps(f)
-    def decorated(*args, **kwargs):
-        if not session.get("user_id"):
-            flash("Please log in to access this page.")
+    def wrapper(*args, **kwargs):
+        if "user_id" not in session:
+            flash("Please log in.")
             return redirect(url_for("login", next=request.path))
-        return f(*args, **kwargs)
-    return decorated
 
+        # If the account is flagged for a mandatory reset, force the user
+        # to the change-password screen before allowing anything else.
+        if session.get("must_reset_password") and request.endpoint not in (
+            "change_password",
+            "logout",
+        ):
+            flash("Please set a new password before continuing.")
+            return redirect(url_for("change_password"))
+
+        return f(*args, **kwargs)
+    return wrapper
 
 def admin_required(view):
     @wraps(view)

syntaxmatrix/core.py

@@ -54,7 +54,7 @@ class SyntaxMUI:
             port="5080", 
             user_icon="👩🏿‍🦲",
             bot_icon="<img src='/static/icons/favicon.png' width=20' alt='bot'/>",
-            favicon="/static/icons/favicon.png",      
+            favicon="",  # /static/icons/favicon.png",      
             site_logo="<img src='/static/icons/logo.png' width='30' alt='logo'/>",           
             site_title="SyntaxMatrix", 
             project_name="smxAI", 
@@ -75,6 +75,7 @@ class SyntaxMUI:
         self.ui_mode = ui_mode
         self.theme_toggle_enabled = False
         self.user_files_enabled = False
+        self.registration_enabled = False
         self.smxai_identity = SMXAI_CHAT_ID
         self.smxai_instructions = SMXAI_CHAT_INSTRUCTIONS    
         self.website_description = SMXAI_WEBSITE_DESCRIPTION
@@ -311,6 +312,9 @@ class SyntaxMUI:
     
     def enable_user_files(self):
         self.user_files_enabled = True
+
+    def enable_registration(self):
+        self.registration_enabled = True
     
     @staticmethod
     def columns(components):
@@ -514,12 +518,12 @@ class SyntaxMUI:
     # ──────────────────────────────────────────────────────────────
     #  *********** LLM CLIENT HELPERS  **********************
     # ──────────────────────────────────────────────────────────────
-    def set_prompt_profile(self, profile):
-        self.ai_chat_id = profile
+    def set_smxai_identity(self, profile):
+        self.set_smxai_identity = profile
     
 
-    def set_prompt_instructions(self, instructions):
-        self.ai_chat_instructions = instructions
+    def set_smxai_instructions(self, instructions):
+        self.set_smxai_instructions = instructions
 
 
     def set_website_description(self, desc):
@@ -1137,21 +1141,23 @@ class SyntaxMUI:
         tasks = [str(t).strip().lower() for t in tasks if str(t).strip()]
 
         ai_profile = """
-        - You are a Python expert specializing in data science and machine learning. 
+        - You are a Python expert specializing in Data Science (DS) and Machine Learning (ML).
         - Your task is to generate a single, complete, production-quality, executable Python script for a Jupyter-like Python kernel, based on the given instructions. 
         - The dataset is already loaded as a pandas DataFrame named `df` (no file I/O or file uploads).
-        - Make a copy of `df` and name it `df_copy`. Make sure `df_copy` is preprocessed and cleaned, named `df_cleaned`, if not already done so. Then use `df_cleaned` to perform the ML tasks described in the given context.
-        - Select your features and target, from `df_cleaned`, with care and name it `required_cols`
+        - Make a copy of `df` and name it `df_copy`. 
+        - Make sure `df_copy` is preprocessed and cleaned, and name it `df_cleaned`, if not already done so. 
+        - Work only with `df_cleaned` to perform the ML tasks described in the given context.
+        - Select your features and targets, from `df_cleaned`, with care and name it `required_cols`
         - Create your 'df_filtered by doing: df_filtered = df_cleaned[required_cols].
-        - Use the {TEMPLATE_CATALOGUE} below to educate yourself on which visualizations you will implement in the code.
-        - The final output MUST be the complete, executable Python code only, enclosed in a single markdown code block (```python ... ```), which is required to fulfill the user's request. See the {tasks} below.
+        - Use the {TEMPLATE_CATALOGUE} below to educate yourself on which visualizations you will implement in the code, and ensure the implementations are in the code you generate.
+        - The final output MUST BE the complete, executable Python code only, enclosed in a single markdown code block (```python ... ```), and MUST BE able to fulfill the user's request: {tasks}.
         - Do not include any explanatory text or markdown outside the code block.
         """
 
         TEMPLATE_CATALOGUE = """
         ### Available SyntaxMatrix templates (use these instead of inventing new helpers)
 
-        Visualisation templates (dataset-agnostic):
+        Visualisation templates:
         - viz_pie(df, category_col=None, top_k=8): pie/donut shares within a category.
         - viz_stacked_bar(df, x=None, hue=None, normalise=True): composition across groups.
         - viz_count_bar(df, category_col=None, top_k=12): counts/denominators by category.
@@ -1191,9 +1197,9 @@ class SyntaxMUI:
             
             """
             ### Template rules
-            - You MAY call a template if it matches the task.
+            - You MAY call 1 or more templates if they matche the task.
             - Do NOT invent template names.
-            - If no template fits, write minimal direct pandas/sklearn/seaborn code instead.
+            - If no template fits, write minimal direct pandas/sklearn/seaborn code instead, for visualization.
             - Keep the solution short: avoid writing wrappers/utilities already handled by SyntaxMatrix hardener.
 
             #### Template selection hint examples:
@@ -1216,8 +1222,7 @@ class SyntaxMUI:
                 set `random_state=42` where relevant.
             4) Be defensive, but avoid hard-failing on optional fields:
             - If the primary column, needed to answer the question, is missing, review your copy of the `df` again.
-            Make sure that you selected the proper column. 
-            Never use a column/variable which isn't available or defined.
+            - Make sure that you selected the proper column. Never use a column/variable which isn't available or defined.
             - If a secondary/extra column is missing, show a warning with `show(...)` and continue using available fields.
             - Handle missing values sensibly (drop rows for simple EDA; use `ColumnTransformer` + `SimpleImputer` for modelling).
             - For categorical features in ML, use `OneHotEncoder(handle_unknown="ignore")`
@@ -1249,6 +1254,20 @@ class SyntaxMUI:
             11) You MUST NOT reference any column outside Available columns: {AVAILABLE_COLUMNS}.
             12) If asked to predict/classify, choose the target by matching the task text to Allowed columns
                 and never invent a new name.
+            13) Treat df as the primary dataset you must work with.
+            14) The dataset is already loaded as df (no file I/O or file uploads).
+            15) All outputs must be visible to the user via the provided show(...) helper.
+            16) Never use print(...); use show(...) instead.
+            17) You MUST NOT read from or write to local files, folders, or external storage.
+            - Do not call open(...), Path(...).write_text/write_bytes, or similar file APIs.
+            - Do not use df.to_csv(...), df.to_excel(...), df.to_parquet(...),
+                df.to_pickle(...), df.to_json(...), df.to_hdf(...), or any other
+                method that writes to disk.
+            - Do not call joblib.dump(...), pickle.dump(...), torch.save(...),
+                numpy.save(...), numpy.savetxt(...), or similar saving functions.
+            - Do not call plt.savefig(..., 'somefile.png') or any variant that
+                writes an image to a filename. Plots must be rendered in-memory only.
+            18) Keep everything in memory and surface results via show(...) or plots.
             
             #### Cohort rules
             When you generate plots for cohorts or categories, you MUST obey these rules:

syntaxmatrix/generate_page.py

@@ -35,15 +35,25 @@ _SMX_FADEIN_CSS = '''
 </style>
 '''.strip()
 
-_SMX_LAYOUT_CSS = '''
+_SMX_LAYOUT_CSS = """
 <style>
-.smx-hero{max-width:1200px;margin:0 auto;padding:clamp(1.5rem,3vw,2.5rem);}
-.smx-hero .smx-img{aspect-ratio:16/9;width:100%;max-height:min(56vh,560px);object-fit:cover;border-radius:1rem;}
-.smx-card{border-radius:1rem;box-shadow:0 6px 20px rgba(2,6,23,0.06);}
-#smx-root,[id^="smx-"]{margin-top:clamp(16px,2vw,32px);margin-bottom:clamp(24px,3vw,48px);padding-inline:clamp(12px,2vw,24px);}
-#smx-root .smx-container,[id^="smx-"] .smx-container{max-width:1200px;margin-inline:auto;}
+    #smx-root,
+    [id^="smx-"]{
+        margin-top: 0;
+        margin-bottom: 40px;
+        /* No side padding on desktop */
+        padding-inline: 0;
+    }
+
+    /* Keep a bit of breathing room on small screens */
+    @media (max-width: 768px) {
+        #smx-root,
+        [id^="smx-"]{
+            padding-inline: 12px;
+        }
+    }
 </style>
-'''.strip()
+""".strip()
 
 def smx_strip_fences(html: str) -> str:
     s = (html or '').strip()