abstractcore 2.3.3__py3-none-any.whl → 2.3.5__py3-none-any.whl

{abstractcore-2.3.3.dist-info → abstractcore-2.3.5.dist-info}/METADATA

@@ -1,11 +1,11 @@
 Metadata-Version: 2.4
 Name: abstractcore
-Version: 2.3.3
+Version: 2.3.5
 Summary: Unified interface to all LLM providers with essential infrastructure for tool calling, streaming, and model management
 Author: Laurent-Philippe Albou
 Maintainer: Laurent-Philippe Albou
 License: MIT
-Project-URL: Homepage, https://github.com/lpalbou/AbstractCore
+Project-URL: Homepage, https://lpalbou.github.io/AbstractCore
 Project-URL: Documentation, https://github.com/lpalbou/AbstractCore#readme
 Project-URL: Repository, https://github.com/lpalbou/AbstractCore
 Project-URL: Bug Tracker, https://github.com/lpalbou/AbstractCore/issues
@@ -89,7 +89,7 @@ Dynamic: license-file
 
 # AbstractCore
 
-A unified, powerful Python library for seamless interaction with multiple Large Language Model (LLM) providers.
+A unified Python library for interaction with multiple Large Language Model (LLM) providers.
 
 **Write once, run everywhere.**
 
@@ -167,12 +167,12 @@ loaded_session = BasicSession.load('conversation.json', provider=llm)
 
 | Provider | Status | Setup |
 |----------|--------|-------|
-| **OpenAI** | ✅ Full | [Get API key](docs/prerequisites.md#openai-setup) |
-| **Anthropic** | ✅ Full | [Get API key](docs/prerequisites.md#anthropic-setup) |
-| **Ollama** | ✅ Full | [Install guide](docs/prerequisites.md#ollama-setup) |
-| **LMStudio** | ✅ Full | [Install guide](docs/prerequisites.md#lmstudio-setup) |
-| **MLX** | ✅ Full | [Setup guide](docs/prerequisites.md#mlx-setup) |
-| **HuggingFace** | ✅ Full | [Setup guide](docs/prerequisites.md#huggingface-setup) |
+| **OpenAI** | Full | [Get API key](docs/prerequisites.md#openai-setup) |
+| **Anthropic** | Full | [Get API key](docs/prerequisites.md#anthropic-setup) |
+| **Ollama** | Full | [Install guide](docs/prerequisites.md#ollama-setup) |
+| **LMStudio** | Full | [Install guide](docs/prerequisites.md#lmstudio-setup) |
+| **MLX** | Full | [Setup guide](docs/prerequisites.md#mlx-setup) |
+| **HuggingFace** | Full | [Setup guide](docs/prerequisites.md#huggingface-setup) |
 
 ## Server Mode (Optional HTTP REST API)
 
@@ -275,7 +275,7 @@ python -m abstractllm.utils.cli --provider anthropic --model claude-3-5-haiku-la
 
 ### Architecture & Advanced
 - **[Architecture](docs/architecture.md)** - System design and architecture overview
-- **[Tool Syntax Rewriting](docs/tool-syntax-rewriting.md)** - Format conversion for agentic CLIs
+- **[Tool Calling](docs/tool-calling.md)** - Universal tool system and format conversion
 
 ## Use Cases
 
@@ -355,15 +355,15 @@ curl -X POST http://localhost:8000/v1/chat/completions \
 
 ## Why AbstractCore?
 
-✅ **Unified Interface**: One API for all LLM providers  
-✅ **Production Ready**: Robust error handling, retries, timeouts  
-✅ **Type Safe**: Full Pydantic integration for structured outputs  
-✅ **Local & Cloud**: Run models locally or use cloud APIs  
-✅ **Tool Calling**: Consistent function calling across providers  
-✅ **Streaming**: Real-time responses for better UX  
-✅ **Embeddings**: Built-in vector embeddings for RAG  
-✅ **Server Mode**: Optional OpenAI-compatible API server  
-✅ **Well Documented**: Comprehensive guides and examples  
+- **Unified Interface**: One API for all LLM providers  
+- **Production Ready**: Robust error handling, retries, timeouts  
+- **Type Safe**: Full Pydantic integration for structured outputs  
+- **Local & Cloud**: Run models locally or use cloud APIs  
+- **Tool Calling**: Consistent function calling across providers  
+- **Streaming**: Real-time responses for interactive applications  
+- **Embeddings**: Built-in vector embeddings for RAG  
+- **Server Mode**: Optional OpenAI-compatible API server  
+- **Well Documented**: Comprehensive guides and examples  
 
 ## Installation Options
 
@@ -399,7 +399,7 @@ All tests passing as of October 12th, 2025.
 ## Quick Links
 
 - **[📚 Documentation Index](docs/)** - Complete documentation navigation guide
-- **[🚀 Getting Started](docs/getting-started.md)** - 5-minute quick start
+- **[Getting Started](docs/getting-started.md)** - 5-minute quick start
 - **[⚙️ Prerequisites](docs/prerequisites.md)** - Provider setup (OpenAI, Anthropic, Ollama, etc.)
 - **[📖 Python API](docs/api-reference.md)** - Complete Python API reference
 - **[🌐 Server Guide](docs/server.md)** - HTTP API server setup
@@ -418,4 +418,4 @@ MIT License - see [LICENSE](LICENSE) file for details.
 
 ---
 
-**AbstractCore** - One interface, all LLM providers. Focus on building, not managing API differences. 🚀
+**AbstractCore** - One interface, all LLM providers. Focus on building, not managing API differences.

{abstractcore-2.3.3.dist-info → abstractcore-2.3.5.dist-info}/RECORD

@@ -1,5 +1,5 @@
-abstractcore-2.3.3.dist-info/licenses/LICENSE,sha256=PI2v_4HMvd6050uDD_4AY_8PzBnu2asa3RKbdDjowTA,1078
-abstractllm/__init__.py,sha256=JeAt7bspfu9EG_6PpC5qEVn_S400AARivx1c0j1RXzU,1839
+abstractcore-2.3.5.dist-info/licenses/LICENSE,sha256=PI2v_4HMvd6050uDD_4AY_8PzBnu2asa3RKbdDjowTA,1078
+abstractllm/__init__.py,sha256=qoh8JJY02pFEGDkTHGiQlWjG5VBMmYn5j8pNBFxTxdE,1839
 abstractllm/apps/__init__.py,sha256=H6fOR28gyBW8bDCEAft2RUZhNmRYA_7fF91szKBjhpA,30
 abstractllm/apps/__main__.py,sha256=qul2xxC59G_ors-OeuODwfq6XY12ahyEe9GGLmrlHjk,1241
 abstractllm/apps/extractor.py,sha256=MsRgEShvsDYt-k30LdIxzSdcQh_TJGBLK7MFKOKb_4c,22682
@@ -28,7 +28,7 @@ abstractllm/processing/basic_judge.py,sha256=5SHpHD8GhgR7BPO8dm8nfnp8bDwGwN_DmfH
 abstractllm/processing/basic_summarizer.py,sha256=rl1IOwOxdgUkHXGN_S_hcXEUYsHQogRj1-wi94jM_28,22631
 abstractllm/providers/__init__.py,sha256=UTpR2Bf_ICFG7M--1kxUmNXs4gl026Tp-KI9zJlvMKU,574
 abstractllm/providers/anthropic_provider.py,sha256=BM8Vu89c974yicvFwlsJ5C3N0wR9Kkt1pOszViWCwAQ,19694
-abstractllm/providers/base.py,sha256=tSKFMBpGzADSwu00-ykRHvKhqaI72L1C6eM4HUkWOQ4,36113
+abstractllm/providers/base.py,sha256=rDxd5XglfZuZNj94iwjWc3ItPSddMQ92Y2G624mb60M,42780
 abstractllm/providers/huggingface_provider.py,sha256=4u3-Z4txonlKXE3DazvgUbxtzVfmdk-ZHsKPSaIfwD4,40785
 abstractllm/providers/lmstudio_provider.py,sha256=HcMltGyiXrLb2acA_XE4wDi1d-X2VZiT3eV1IjMF150,15701
 abstractllm/providers/mlx_provider.py,sha256=_tYg7AhPSsUfCUMqh7FJ--F3ZWr2BNNCtnvH4QtxttU,13745
@@ -53,8 +53,8 @@ abstractllm/utils/__init__.py,sha256=245wfpreb4UHXsCQBHKe9DbtjKNo-KIi-k14apBRj_Q
 abstractllm/utils/cli.py,sha256=5Dgyk6d17SKHX52n0VWa9dkfd5qsGh4jO44DZjYhsDM,55782
 abstractllm/utils/self_fixes.py,sha256=QEDwNTW80iQM4ftfEY3Ghz69F018oKwLM9yeRCYZOvw,5886
 abstractllm/utils/structured_logging.py,sha256=2_bbMRjOvf0gHsRejncel_-PrhYUsOUySX_eaPcQopc,15827
-abstractcore-2.3.3.dist-info/METADATA,sha256=KhW3cZrR28ytkKGMQwzimAmSLjTmeOhz0ng7kp4aT1I,15053
-abstractcore-2.3.3.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-abstractcore-2.3.3.dist-info/entry_points.txt,sha256=plz04HNXbCbQkmEQj_8xKWW8x7zBxus68HD_04-IARM,306
-abstractcore-2.3.3.dist-info/top_level.txt,sha256=Md-8odCjB7hTNnE5hucnifAoLrL9HvRPffZmCq2jpoI,12
-abstractcore-2.3.3.dist-info/RECORD,,
+abstractcore-2.3.5.dist-info/METADATA,sha256=3BmiLAm1e6EmpJ7QNvmqs0QFzVnAsbDx7PGqrG-Phkg,14987
+abstractcore-2.3.5.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+abstractcore-2.3.5.dist-info/entry_points.txt,sha256=plz04HNXbCbQkmEQj_8xKWW8x7zBxus68HD_04-IARM,306
+abstractcore-2.3.5.dist-info/top_level.txt,sha256=Md-8odCjB7hTNnE5hucnifAoLrL9HvRPffZmCq2jpoI,12
+abstractcore-2.3.5.dist-info/RECORD,,

abstractllm/__init__.py

@@ -24,7 +24,7 @@ Quick Start:
     print(response.content)
 """
 
-__version__ = "2.3.3"
+__version__ = "2.3.5"
 
 from .core.factory import create_llm
 from .core.session import BasicSession

abstractllm/providers/base.py

@@ -227,6 +227,22 @@ class BaseProvider(AbstractLLMInterface, ABC):
                     "Install with: pip install pydantic>=2.0.0"
                 )
 
+            # Handle hybrid case: tools + structured output
+            if tools is not None:
+                return self._handle_tools_with_structured_output(
+                    prompt=prompt,
+                    messages=messages,
+                    system_prompt=system_prompt,
+                    tools=tools,
+                    response_model=response_model,
+                    retry_strategy=retry_strategy,
+                    tool_call_tags=tool_call_tags,
+                    execute_tools=execute_tools,
+                    stream=stream,
+                    **kwargs
+                )
+
+            # Standard structured output (no tools)
             from ..structured import StructuredOutputHandler
             handler = StructuredOutputHandler(retry_strategy=retry_strategy)
             return handler.generate_structured(
@@ -235,7 +251,7 @@ class BaseProvider(AbstractLLMInterface, ABC):
                 response_model=response_model,
                 messages=messages,
                 system_prompt=system_prompt,
-                tools=tools,
+                tools=None,  # No tools in this path
                 stream=stream,
                 **kwargs
             )
@@ -824,4 +840,140 @@ class BaseProvider(AbstractLLMInterface, ABC):
             self.logger.debug(f"Non-streaming tag rewriting failed: {e}")
 
         # Return original response if rewriting fails
-        return response
+        return response
+
+    def _handle_tools_with_structured_output(self,
+                                           prompt: str,
+                                           messages: Optional[List[Dict[str, str]]] = None,
+                                           system_prompt: Optional[str] = None,
+                                           tools: Optional[List] = None,
+                                           response_model: Optional[Type[BaseModel]] = None,
+                                           retry_strategy=None,
+                                           tool_call_tags: Optional[str] = None,
+                                           execute_tools: Optional[bool] = None,
+                                           stream: bool = False,
+                                           **kwargs) -> BaseModel:
+        """
+        Handle the hybrid case: tools + structured output.
+        
+        Strategy: Sequential execution
+        1. First, generate response with tools (may include tool calls)
+        2. If tool calls are generated, execute them
+        3. Then generate structured output using tool results as context
+        
+        Args:
+            prompt: Input prompt
+            messages: Optional message history
+            system_prompt: Optional system prompt
+            tools: List of available tools
+            response_model: Pydantic model for structured output
+            retry_strategy: Optional retry strategy for structured output
+            tool_call_tags: Optional tool call tag format
+            execute_tools: Whether to execute tools automatically
+            stream: Whether to use streaming (not supported for hybrid mode)
+            **kwargs: Additional parameters
+            
+        Returns:
+            Validated instance of response_model
+            
+        Raises:
+            ValueError: If streaming is requested (not supported for hybrid mode)
+        """
+        if stream:
+            raise ValueError(
+                "Streaming is not supported when combining tools with structured output. "
+                "Please use either stream=True OR response_model, but not both."
+            )
+            
+        # Step 1: Generate response with tools (normal tool execution flow)
+        self.logger.info("Hybrid mode: Executing tools first, then structured output",
+                        model=self.model,
+                        response_model=response_model.__name__,
+                        num_tools=len(tools) if tools else 0)
+        
+        # Force tool execution for hybrid mode
+        should_execute_tools = execute_tools if execute_tools is not None else True
+        
+        # Generate response with tools using the normal flow (without response_model)
+        tool_response = self.generate_with_telemetry(
+            prompt=prompt,
+            messages=messages,
+            system_prompt=system_prompt,
+            tools=tools,
+            stream=False,  # Never stream in hybrid mode
+            response_model=None,  # No structured output in first pass
+            tool_call_tags=tool_call_tags,
+            execute_tools=should_execute_tools,
+            **kwargs
+        )
+        
+        # Step 2: Generate structured output using tool results as context
+        # Create enhanced prompt with tool execution context
+        if hasattr(tool_response, 'content') and tool_response.content:
+            enhanced_prompt = f"""{prompt}
+
+Based on the following tool execution results:
+{tool_response.content}
+
+Please provide a structured response."""
+        else:
+            enhanced_prompt = prompt
+            
+        self.logger.info("Hybrid mode: Generating structured output with tool context",
+                        model=self.model,
+                        response_model=response_model.__name__,
+                        has_tool_context=bool(hasattr(tool_response, 'content') and tool_response.content))
+        
+        # Generate structured output using the enhanced prompt
+        from ..structured import StructuredOutputHandler
+        handler = StructuredOutputHandler(retry_strategy=retry_strategy)
+        
+        structured_result = handler.generate_structured(
+            provider=self,
+            prompt=enhanced_prompt,
+            response_model=response_model,
+            messages=messages,
+            system_prompt=system_prompt,
+            tools=None,  # No tools in structured output pass
+            stream=False,
+            **kwargs
+        )
+        
+        self.logger.info("Hybrid mode: Successfully completed tools + structured output",
+                        model=self.model,
+                        response_model=response_model.__name__,
+                        success=True)
+        
+        return structured_result
+
+    def generate(self,
+                prompt: str,
+                messages: Optional[List[Dict[str, str]]] = None,
+                system_prompt: Optional[str] = None,
+                tools: Optional[List[Dict[str, Any]]] = None,
+                stream: bool = False,
+                **kwargs) -> Union[GenerateResponse, Iterator[GenerateResponse], BaseModel]:
+        """
+        Generate response from the LLM.
+        
+        This method implements the AbstractLLMInterface and delegates to generate_with_telemetry.
+        
+        Args:
+            prompt: The input prompt
+            messages: Optional conversation history
+            system_prompt: Optional system prompt
+            tools: Optional list of available tools
+            stream: Whether to stream the response
+            **kwargs: Additional provider-specific parameters (including response_model)
+            
+        Returns:
+            GenerateResponse, iterator of GenerateResponse for streaming, or BaseModel for structured output
+        """
+        return self.generate_with_telemetry(
+            prompt=prompt,
+            messages=messages,
+            system_prompt=system_prompt,
+            tools=tools,
+            stream=stream,
+            **kwargs
+        )