SimplerLLM 0.3.4__tar.gz → 0.3.6__tar.gz

simplerllm-0.3.6/MANIFEST.in

@@ -0,0 +1,3 @@
+include requirements.txt
+include README.md
+include LICENSE

{simplerllm-0.3.4 → simplerllm-0.3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SimplerLLM
-Version: 0.3.4
+Version: 0.3.6
 Summary: An easy-to-use Library for interacting with language models.
 Home-page: https://github.com/hassancs91/SimplerLLM
 Author: Hasan Aboul Hasan
@@ -18,6 +18,8 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.6
 Description-Content-Type: text/markdown
@@ -43,7 +45,6 @@ Requires-Dist: colorama>=0.4
 Requires-Dist: scipy>=1.15
 Requires-Dist: tiktoken>=0.9
 Requires-Dist: qdrant-client==1.14.3
-Requires-Dist: voyageai==0.3.3
 Requires-Dist: pydub>=0.25.1
 Requires-Dist: elevenlabs>=2.22.0
 Requires-Dist: moviepy>=1.0.3
@@ -52,6 +53,8 @@ Provides-Extra: transformers
 Requires-Dist: transformers>=4.36.0; extra == "transformers"
 Requires-Dist: torch>=2.0.0; extra == "transformers"
 Requires-Dist: accelerate>=0.25.0; extra == "transformers"
+Provides-Extra: voyage
+Requires-Dist: voyageai>=0.3.3; extra == "voyage"
 Dynamic: author
 Dynamic: author-email
 Dynamic: classifier

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/language/llm/wrappers/anthropic_wrapper.py

@@ -89,6 +89,7 @@ class AnthropicLLM(LLM):
                     max_tokens=max_tokens,
                     full_response=full_response,
                     api_key=self.api_key,
+                    json_mode=json_mode,
                 )
                 if self.verbose:
                     verbose_print("Web search response received successfully", "info")
@@ -241,6 +242,7 @@ class AnthropicLLM(LLM):
                     max_tokens=max_tokens,
                     full_response=full_response,
                     api_key=self.api_key,
+                    json_mode=json_mode,
                 )
                 if self.verbose:
                     verbose_print("Web search response received successfully", "info")

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/language/llm/wrappers/openai_wrapper.py

@@ -245,6 +245,7 @@ class OpenAILLM(LLM):
                     max_tokens=max_tokens,
                     full_response=full_response,
                     api_key=self.api_key,
+                    json_mode=json_mode,
                 )
                 if self.verbose:
                     verbose_print("Web search response received successfully", "info")
@@ -407,6 +408,7 @@ class OpenAILLM(LLM):
                     max_tokens=max_tokens,
                     full_response=full_response,
                     api_key=self.api_key,
+                    json_mode=json_mode,
                 )
                 if self.verbose:
                     verbose_print("Web search response received successfully", "info")

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/language/llm_addons/json_generation.py

@@ -47,6 +47,7 @@ from SimplerLLM.tools.json_helpers import (
     validate_json_with_pydantic_model,
     generate_json_example_from_pydantic,
     extract_schema_constraints,
+    try_auto_wrap_for_nested_model,
     HAS_ROOT_MODEL,
     RootModel,
 )
@@ -279,6 +280,9 @@ def generate_pydantic_json_model(
                 # Handle RootModel list unwrapping (when LLM wraps list in object)
                 json_object = _unwrap_rootmodel_list(json_object, model_class)
 
+                # Handle auto-wrapping for nested models (when LLM returns unwrapped objects)
+                json_object = try_auto_wrap_for_nested_model(model_class, json_object)
+
                 validated, errors = validate_json_with_pydantic_model(
                     model_class, json_object
                 )
@@ -411,6 +415,9 @@ def generate_pydantic_json_model_reliable(
                 # Handle RootModel list unwrapping (when LLM wraps list in object)
                 json_object = _unwrap_rootmodel_list(json_object, model_class)
 
+                # Handle auto-wrapping for nested models (when LLM returns unwrapped objects)
+                json_object = try_auto_wrap_for_nested_model(model_class, json_object)
+
                 validated, errors = validate_json_with_pydantic_model(
                     model_class, json_object
                 )
@@ -541,6 +548,9 @@ async def generate_pydantic_json_model_async(
                 # Handle RootModel list unwrapping (when LLM wraps list in object)
                 json_object = _unwrap_rootmodel_list(json_object, model_class)
 
+                # Handle auto-wrapping for nested models (when LLM returns unwrapped objects)
+                json_object = try_auto_wrap_for_nested_model(model_class, json_object)
+
                 validated, errors = validate_json_with_pydantic_model(
                     model_class, json_object
                 )
@@ -674,6 +684,9 @@ async def generate_pydantic_json_model_reliable_async(
                 # Handle RootModel list unwrapping (when LLM wraps list in object)
                 json_object = _unwrap_rootmodel_list(json_object, model_class)
 
+                # Handle auto-wrapping for nested models (when LLM returns unwrapped objects)
+                json_object = try_auto_wrap_for_nested_model(model_class, json_object)
+
                 validated, errors = validate_json_with_pydantic_model(
                     model_class, json_object
                 )

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/language/llm_providers/anthropic_llm.py

@@ -443,6 +443,7 @@ def generate_response_with_web_search(
     max_tokens: int = 300,
     full_response: bool = False,
     api_key: Optional[str] = None,
+    json_mode: bool = False,
 ) -> Optional[Dict]:
     """
     Generate a response using Anthropic's Messages API with web search enabled.
@@ -454,6 +455,7 @@ def generate_response_with_web_search(
         max_tokens: Maximum tokens for the response
         full_response: If True, returns LLMFullResponse with web_sources
         api_key: Anthropic API key
+        json_mode: If True, enhances system prompt to request JSON output
 
     Returns:
         str or LLMFullResponse: Generated text or full response with web sources
@@ -475,6 +477,11 @@ def generate_response_with_web_search(
 
     client = Anthropic(api_key=api_key)
 
+    # Enhance system prompt for JSON mode
+    effective_system_prompt = system_prompt
+    if json_mode:
+        effective_system_prompt = f"{system_prompt}\n\nIMPORTANT: You MUST respond with valid JSON only. No additional text or explanation outside the JSON structure."
+
     # Execute with retry logic
     response = None
     last_error = None
@@ -484,7 +491,7 @@ def generate_response_with_web_search(
             response = client.messages.create(
                 model=model_name,
                 max_tokens=max_tokens,
-                system=system_prompt,
+                system=effective_system_prompt,
                 messages=messages,
                 tools=[{
                     "type": "web_search_20250305",
@@ -550,6 +557,7 @@ async def generate_response_with_web_search_async(
     max_tokens: int = 300,
     full_response: bool = False,
     api_key: Optional[str] = None,
+    json_mode: bool = False,
 ) -> Optional[Dict]:
     """
     Asynchronously generate a response using Anthropic's Messages API with web search enabled.
@@ -561,6 +569,7 @@ async def generate_response_with_web_search_async(
         max_tokens: Maximum tokens for the response
         full_response: If True, returns LLMFullResponse with web_sources
         api_key: Anthropic API key
+        json_mode: If True, enhances system prompt to request JSON output
 
     Returns:
         str or LLMFullResponse: Generated text or full response with web sources
@@ -584,6 +593,11 @@ async def generate_response_with_web_search_async(
 
     client = AsyncAnthropic(api_key=api_key)
 
+    # Enhance system prompt for JSON mode
+    effective_system_prompt = system_prompt
+    if json_mode:
+        effective_system_prompt = f"{system_prompt}\n\nIMPORTANT: You MUST respond with valid JSON only. No additional text or explanation outside the JSON structure."
+
     # Execute with retry logic
     response = None
     last_error = None
@@ -593,7 +607,7 @@ async def generate_response_with_web_search_async(
             response = await client.messages.create(
                 model=model_name,
                 max_tokens=max_tokens,
-                system=system_prompt,
+                system=effective_system_prompt,
                 messages=messages,
                 tools=[{
                     "type": "web_search_20250305",

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/language/llm_providers/openai_llm.py

@@ -713,6 +713,7 @@ def generate_response_with_web_search(
     max_tokens=300,
     full_response=False,
     api_key=None,
+    json_mode=False,
 ):
     """
     Generate a response using OpenAI's Responses API with web search enabled.
@@ -723,6 +724,7 @@ def generate_response_with_web_search(
         max_tokens: Maximum tokens for the response
         full_response: If True, returns LLMFullResponse with web_sources
         api_key: OpenAI API key
+        json_mode: If True, enables JSON output format
 
     Returns:
         str or LLMFullResponse: Generated text or full response with web sources
@@ -732,11 +734,17 @@ def generate_response_with_web_search(
 
     for attempt in range(MAX_RETRIES):
         try:
-            response = openai_client.responses.create(
-                model=model_name,
-                input=input_text,
-                tools=[{"type": "web_search"}],
-            )
+            # Build request parameters
+            # Note: JSON mode is NOT supported with web search in OpenAI's Responses API
+            # The json_mode parameter is accepted but ignored - JSON extraction happens
+            # post-response via extract_json_from_text() in json_generation.py
+            request_params = {
+                "model": model_name,
+                "input": input_text,
+                "tools": [{"type": "web_search"}],
+            }
+
+            response = openai_client.responses.create(**request_params)
 
             # Extract text and citations from the response
             generated_text = ""
@@ -784,6 +792,7 @@ async def generate_response_with_web_search_async(
     max_tokens=300,
     full_response=False,
     api_key=None,
+    json_mode=False,
 ):
     """
     Asynchronously generate a response using OpenAI's Responses API with web search enabled.
@@ -794,6 +803,7 @@ async def generate_response_with_web_search_async(
         max_tokens: Maximum tokens for the response
         full_response: If True, returns LLMFullResponse with web_sources
         api_key: OpenAI API key
+        json_mode: If True, enables JSON output format
 
     Returns:
         str or LLMFullResponse: Generated text or full response with web sources
@@ -803,11 +813,17 @@ async def generate_response_with_web_search_async(
 
     for attempt in range(MAX_RETRIES):
         try:
-            response = await async_openai_client.responses.create(
-                model=model_name,
-                input=input_text,
-                tools=[{"type": "web_search"}],
-            )
+            # Build request parameters
+            # Note: JSON mode is NOT supported with web search in OpenAI's Responses API
+            # The json_mode parameter is accepted but ignored - JSON extraction happens
+            # post-response via extract_json_from_text() in json_generation.py
+            request_params = {
+                "model": model_name,
+                "input": input_text,
+                "tools": [{"type": "web_search"}],
+            }
+
+            response = await async_openai_client.responses.create(**request_params)
 
             # Extract text and citations from the response
             generated_text = ""

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM/tools/json_helpers.py

@@ -15,6 +15,38 @@ except ImportError:
     RootModel = None
 
 
+def _is_decimal_type(field_type) -> bool:
+    """Check if a type is Decimal, handling various import scenarios."""
+    if field_type is Decimal:
+        return True
+    if field_type == Decimal:
+        return True
+    # Fallback: check by name for cases where Decimal is imported differently
+    type_name = getattr(field_type, '__name__', None)
+    if type_name == 'Decimal':
+        return True
+    return False
+
+
+def _is_origin_type(origin, target_types: tuple) -> bool:
+    """
+    Check if origin matches any of the target types.
+    Handles differences between Python versions where get_origin() may return
+    different values (e.g., set vs typing.Set).
+    """
+    if origin is None:
+        return False
+    # Direct comparison
+    if origin in target_types:
+        return True
+    # Check by name for cross-version compatibility
+    origin_name = getattr(origin, '__name__', str(origin))
+    for target in target_types:
+        target_name = getattr(target, '__name__', str(target))
+        if origin_name == target_name:
+            return True
+    return False
+
 
 def convert_pydantic_to_json(model_instance):
     """
@@ -126,7 +158,14 @@ def validate_json_with_pydantic_model(model_class, json_data):
             try:
                 # RootModel takes value directly, not as kwargs
                 if is_root_model:
-                    model_instance = model_class(item)
+                    # Handle case where LLM wrapped the list in an object
+                    # e.g., {"items": ["a", "b"]} should become ["a", "b"]
+                    actual_item = item
+                    if isinstance(item, dict) and len(item) == 1:
+                        single_value = list(item.values())[0]
+                        if isinstance(single_value, list):
+                            actual_item = single_value
+                    model_instance = model_class(actual_item)
                 else:
                     model_instance = model_class(**item)
                 validated_data.append(model_instance.model_dump())
@@ -144,6 +183,102 @@ def validate_json_with_pydantic_model(model_class, json_data):
     return validated_data, validation_errors
 
 
+def try_auto_wrap_for_nested_model(model_class: Type[BaseModel], json_data: list) -> list:
+    """
+    Auto-wrap unwrapped JSON objects if the model expects them in a single List field.
+
+    When an LLM returns separate objects instead of wrapping them in the expected
+    container field, this function detects and fixes the structure.
+
+    Example:
+        Input (unwrapped):  [{"name": "A", ...}, {"name": "B", ...}]
+        Output (wrapped):   [{"items": [{"name": "A", ...}, {"name": "B", ...}]}]
+
+    Args:
+        model_class: The target Pydantic model class
+        json_data: List of extracted JSON objects
+
+    Returns:
+        Modified json_data with auto-wrapping applied if applicable
+    """
+    if not json_data or len(json_data) == 0:
+        return json_data
+
+    # Skip RootModel (handled separately by _unwrap_rootmodel_list)
+    if HAS_ROOT_MODEL and RootModel is not None:
+        try:
+            if issubclass(model_class, RootModel):
+                return json_data
+        except TypeError:
+            pass
+
+    # Handle case where LLM returned a JSON array that got wrapped as [[{}, {}]]
+    # This happens when extract_json_from_text parses "[{...}, {...}]" and wraps it
+    working_data = json_data
+    if (len(json_data) == 1 and
+        isinstance(json_data[0], list) and
+        len(json_data[0]) > 0 and
+        all(isinstance(item, dict) for item in json_data[0])):
+        # json_data is [[{}, {}]] - use the inner list for processing
+        working_data = json_data[0]
+
+    # First try validation - if at least one item validates, use it
+    # This handles the case where extract_json_from_text extracts both
+    # the outer wrapper AND inner objects from pretty-printed JSON
+    validated, errors = validate_json_with_pydantic_model(model_class, working_data)
+    if validated:
+        # At least one item validated - return only the validated items
+        # Re-extract just the validated data to return
+        valid_items = []
+        for item in working_data:
+            try:
+                model_class(**item)
+                valid_items.append(item)
+            except:
+                pass
+        if valid_items:
+            return valid_items
+
+    # Find List[SomeModel] fields in the model
+    type_hints = get_type_hints(model_class)
+    list_fields = []
+
+    for field_name, field_type in type_hints.items():
+        origin = get_origin(field_type)
+        if origin is list:
+            args = get_args(field_type)
+            if args and hasattr(args[0], '__annotations__'):
+                # This is a List[SomeModel] field
+                list_fields.append((field_name, args[0]))
+
+    # Only auto-wrap if model has exactly ONE List[Model] field
+    if len(list_fields) != 1:
+        return json_data
+
+    field_name, inner_model = list_fields[0]
+    inner_fields = set(get_type_hints(inner_model).keys())
+
+    # Check if ALL extracted objects contain the inner model's required fields
+    # Allow extra keys (they'll be ignored during Pydantic validation)
+    # This makes auto-wrap more robust with web search where LLM output is less constrained
+    all_match = all(
+        isinstance(obj, dict) and inner_fields.issubset(set(obj.keys()))
+        for obj in working_data
+    )
+
+    if all_match:
+        # Wrap the objects in the expected structure
+        wrapped = [{field_name: working_data}]
+
+        # Validate the wrapped structure
+        validated, errors = validate_json_with_pydantic_model(model_class, wrapped)
+        if not errors:
+            return wrapped
+
+    # Return original if wrapping didn't help
+    return working_data
+
+
 def convert_json_to_pydantic_model(model_class, json_data):
     try:
         # Handle RootModel - it takes the value directly, not as kwargs
@@ -173,6 +308,10 @@ def get_field_constraints(field_info):
     """
     constraints = {}
 
+    # Check direct pattern attribute on FieldInfo (Pydantic v2 Field(pattern="..."))
+    if hasattr(field_info, 'pattern') and field_info.pattern:
+        constraints['pattern'] = field_info.pattern
+
     # Check if field_info has constraints attribute (Pydantic v2)
     if hasattr(field_info, 'constraints'):
         for constraint in field_info.constraints:
@@ -362,7 +501,7 @@ def example_value_for_type(field_type: Type, constraints: dict = None, _recursio
         return "12345678-1234-5678-1234-567812345678"
 
     # Handle Decimal type
-    if field_type == Decimal:
+    if _is_decimal_type(field_type):
         if 'ge' in constraints:
             return str(constraints['ge'])
         elif 'gt' in constraints:
@@ -406,21 +545,44 @@ def example_value_for_type(field_type: Type, constraints: dict = None, _recursio
                 for field_name, field_info in field_type.model_fields.items():
                     field_constraints = get_field_constraints(field_info)
                     field_type_hint = get_type_hints(field_type).get(field_name)
-                    example_data[field_name] = example_value_for_type(
-                        field_type_hint,
-                        field_constraints,
-                        _recursion_depth + 1,
-                        _seen_types
-                    )
+                    # Try smart example first (handles field validators like username, password, email)
+                    smart = get_smart_example_for_field(field_name, field_type_hint, field_constraints)
+                    if smart is not None:
+                        example_data[field_name] = smart
+                    else:
+                        value = example_value_for_type(
+                            field_type_hint,
+                            field_constraints,
+                            _recursion_depth + 1,
+                            _seen_types
+                        )
+                        # For list-type fields, ensure we return [] instead of None
+                        # This handles circular references in List[Model] fields
+                        if value is None and field_type_hint is not None:
+                            hint_origin = get_origin(field_type_hint)
+                            if hint_origin is list:
+                                value = []
+                        example_data[field_name] = value
             else:
                 # Fallback for older Pydantic versions
                 for field_name, field_type_hint in get_type_hints(field_type).items():
-                    example_data[field_name] = example_value_for_type(
-                        field_type_hint,
-                        None,
-                        _recursion_depth + 1,
-                        _seen_types
-                    )
+                    # Try smart example first
+                    smart = get_smart_example_for_field(field_name, field_type_hint, {})
+                    if smart is not None:
+                        example_data[field_name] = smart
+                    else:
+                        value = example_value_for_type(
+                            field_type_hint,
+                            {},
+                            _recursion_depth + 1,
+                            _seen_types
+                        )
+                        # For list-type fields, ensure we return [] instead of None
+                        if value is None and field_type_hint is not None:
+                            hint_origin = get_origin(field_type_hint)
+                            if hint_origin is list:
+                                value = []
+                        example_data[field_name] = value
             return field_type(**example_data)
         elif field_type == str:
             # Check for pattern constraint first
@@ -447,8 +609,9 @@ def example_value_for_type(field_type: Type, constraints: dict = None, _recursio
                 return base_string[:target_length]
             else:
                 # Need to pad or repeat to meet minimum length
+                # Use underscores instead of spaces to avoid breaking alphanumeric validators
                 repeats = (target_length // len(base_string)) + 1
-                repeated = (base_string + " ") * repeats
+                repeated = (base_string + "_") * repeats
                 return repeated[:target_length]
         elif field_type == int:
             # Respect numeric constraints
@@ -532,7 +695,7 @@ def example_value_for_type(field_type: Type, constraints: dict = None, _recursio
         for i in range(1, 3):  # Create 2 example key-value pairs
             example_dict[f"example_key_{i}"] = example_value_for_type(value_type, {}, _recursion_depth + 1, _seen_types)
         return example_dict
-    elif origin == set or origin == frozenset:  # Handle Set and FrozenSet
+    elif _is_origin_type(origin, (set, frozenset)):  # Handle Set and FrozenSet
         args = get_args(field_type)
         if not args:
             return []  # No type specified, return empty list
@@ -548,7 +711,7 @@ def example_value_for_type(field_type: Type, constraints: dict = None, _recursio
                     value = f"{value}_{i}"
                 items.append(value)
         return items  # Return as list since JSON doesn't support sets
-    elif origin == tuple:  # Handle Tuple types
+    elif _is_origin_type(origin, (tuple,)):  # Handle Tuple types
         args = get_args(field_type)
         if not args:
             return []  # No type args, return empty list

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SimplerLLM
-Version: 0.3.4
+Version: 0.3.6
 Summary: An easy-to-use Library for interacting with language models.
 Home-page: https://github.com/hassancs91/SimplerLLM
 Author: Hasan Aboul Hasan
@@ -18,6 +18,8 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Requires-Python: >=3.6
 Description-Content-Type: text/markdown
@@ -43,7 +45,6 @@ Requires-Dist: colorama>=0.4
 Requires-Dist: scipy>=1.15
 Requires-Dist: tiktoken>=0.9
 Requires-Dist: qdrant-client==1.14.3
-Requires-Dist: voyageai==0.3.3
 Requires-Dist: pydub>=0.25.1
 Requires-Dist: elevenlabs>=2.22.0
 Requires-Dist: moviepy>=1.0.3
@@ -52,6 +53,8 @@ Provides-Extra: transformers
 Requires-Dist: transformers>=4.36.0; extra == "transformers"
 Requires-Dist: torch>=2.0.0; extra == "transformers"
 Requires-Dist: accelerate>=0.25.0; extra == "transformers"
+Provides-Extra: voyage
+Requires-Dist: voyageai>=0.3.3; extra == "voyage"
 Dynamic: author
 Dynamic: author-email
 Dynamic: classifier

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM.egg-info/SOURCES.txt

@@ -1,5 +1,7 @@
 LICENSE
+MANIFEST.in
 README.md
+requirements.txt
 setup.py
 SimplerLLM/__init__.py
 SimplerLLM.egg-info/PKG-INFO
@@ -120,6 +122,7 @@ SimplerLLM/voice/tts/providers/lahajati_tts.py
 SimplerLLM/voice/tts/providers/openai_tts.py
 tests/__init__.py
 tests/test_anthropic.py
+tests/test_aspect_ratio_debug.py
 tests/test_brainstorm.py
 tests/test_cohere.py
 tests/test_deepseek.py

{simplerllm-0.3.4 → simplerllm-0.3.6}/SimplerLLM.egg-info/requires.txt

@@ -19,7 +19,6 @@ colorama>=0.4
 scipy>=1.15
 tiktoken>=0.9
 qdrant-client==1.14.3
-voyageai==0.3.3
 pydub>=0.25.1
 elevenlabs>=2.22.0
 moviepy>=1.0.3
@@ -29,3 +28,6 @@ yt-dlp>=2023.3.4
 transformers>=4.36.0
 torch>=2.0.0
 accelerate>=0.25.0
+
+[voyage]
+voyageai>=0.3.3

simplerllm-0.3.6/requirements.txt

@@ -0,0 +1,30 @@
+aiohttp>=3.9
+anthropic>=0.40.0
+cohere>=5.0
+duckduckgo_search>=5.3
+elevenlabs>=1.0
+lxml_html_clean>=0.1
+newspaper3k>=0.2
+numpy>=1.26
+openai>=1.59
+google-genai>=1.52.0
+Pillow>=10.0
+pydantic>=2.10
+PyPDF2>=3.0
+python-dotenv>=1.0
+python_docx>=1.1
+requests>=2.31
+youtube_transcript_api>=0.6
+colorama>=0.4
+scipy>=1.15
+tiktoken>=0.9
+qdrant-client==1.14.3
+# voyageai is optional - install via: pip install SimplerLLM[voyage]
+# (Not in default install because upstream voyageai currently requires-python <3.14)
+pydub>=0.25.1
+elevenlabs>=2.22.0
+# sounddevice>=0.4.6  # Optional - for live voice features. Requires PortAudio. Install with: pip install SimplerLLM[voice]
+# pynput>=1.7.6  # Optional - for push-to-talk. Install with: pip install SimplerLLM[voice]
+# pygame>=2.5.0  # Optional - only needed for AudioPlayer file playback. Install with: pip install pygame>=2.5.0
+moviepy>=1.0.3
+yt-dlp>=2023.3.4

{simplerllm-0.3.4 → simplerllm-0.3.6}/setup.py

@@ -15,6 +15,9 @@ extras_require = {
         "torch>=2.0.0",
         "accelerate>=0.25.0",
     ],
+    "voyage": [
+        "voyageai>=0.3.3",
+    ],
 }
 
 # Read the long description from the README file
@@ -23,7 +26,7 @@ with open("README.md", encoding="utf-8") as f:
 
 setup(
     name="SimplerLLM",
-    version="0.3.4",
+    version="0.3.6",
     author="Hasan Aboul Hasan",
     author_email="hasan@learnwithhasan.com",
     description="An easy-to-use Library for interacting with language models.",
@@ -48,6 +51,8 @@ setup(
         "Programming Language :: Python :: 3.10",
         "Programming Language :: Python :: 3.11",
         "Programming Language :: Python :: 3.12",
+        "Programming Language :: Python :: 3.13",
+        "Programming Language :: Python :: 3.14",
         "Topic :: Software Development :: Libraries :: Python Modules",
     ],