local-deep-research 0.1.24__py3-none-any.whl → 0.1.26__py3-none-any.whl

local_deep_research/config.py

@@ -6,10 +6,12 @@ from platformdirs import user_documents_dir
 import os
 # Setup logging
 logger = logging.getLogger(__name__)
+from dotenv import load_dotenv
+import platform
 
 # Get config directory
 def get_config_dir():
-    import platform
+
     
     if platform.system() == "Windows":
         # Windows: Use Documents directory
@@ -33,15 +35,25 @@ SEARCH_ENGINES_FILE = CONFIG_DIR / "search_engines.toml"
 LOCAL_COLLECTIONS_FILE = CONFIG_DIR / "local_collections.toml"
 
 
-# Set environment variable for Dynaconf to use
-docs_base = Path(user_documents_dir()) / "local_deep_research"
-os.environ["DOCS_DIR"] = str(docs_base)
-
-
-
+env_file = CONFIG_DIR / ".env"
 
+if env_file.exists():
+    logger.info(f"Loading environment variables from: {env_file}")
+    load_dotenv(dotenv_path=env_file)
+else:
+    logger.warning(f"Warning: .env file not found at {env_file}. Trying secondary location.")
+    env_file_secondary = get_config_dir() / ".env"
+    if env_file_secondary.exists():
+        get_config_dir() / "config"
+        logger.info(f"Loading environment variables from: {env_file_secondary}")
+        load_dotenv(dotenv_path=env_file_secondary)
+    else:
+        logger.warning(f"Warning: .env file also not found at {env_file_secondary}.")
 
 
+# Set environment variable for Dynaconf to use
+docs_base = Path(user_documents_dir()) / "local_deep_research"
+os.environ["DOCS_DIR"] = str(docs_base)
 
 
 # Expose get_llm function

local_deep_research/defaults/.env.template

@@ -1,11 +1,15 @@
 # API Keys
+# OPENAI_ENDPOINT_API_KEY=your-api-key-here
 # ANTHROPIC_API_KEY=your-api-key-here
 # OPENAI_API_KEY=your-openai-key-here
+
+# BRAVE_API_KEY=
 # SERP_API_KEY=your-api-key-here
 # GUARDIAN_API_KEY=your-api-key-here
 # GOOGLE_PSE_API_KEY=your-google-api-key-here
 # GOOGLE_PSE_ENGINE_ID=your-programmable-search-engine-id-here
 
+
 # SearXNG Configuration, add at least SEARXNG_INSTANCE to .env file to use this search engine
 # SEARXNG_INSTANCE = "http://localhost:8080"
-# SEARXNG_DELAY = 2.0
+# SEARXNG_DELAY = 2.0

local_deep_research/defaults/llm_config.py

@@ -9,6 +9,7 @@ from langchain_anthropic import ChatAnthropic
 from langchain_openai import ChatOpenAI
 from langchain_ollama import ChatOllama
 from langchain_community.llms import VLLM
+from local_deep_research.utilties.search_utilities import remove_think_tags
 from local_deep_research.config import settings
 import os
 import logging
@@ -23,6 +24,8 @@ VALID_PROVIDERS = ["ollama", "openai", "anthropic", "vllm", "openai_endpoint", "
 # LLM FUNCTIONS
 # ================================
 
+
+
 def get_llm(model_name=None, temperature=None, provider=None):
     """
     Get LLM instance based on model name and provider.
@@ -33,7 +36,7 @@ def get_llm(model_name=None, temperature=None, provider=None):
         provider: Provider to use (if None, uses settings.llm.provider)
     
     Returns:
-        A LangChain LLM instance
+        A LangChain LLM instance with automatic think-tag removal
     """
     # Use settings values for parameters if not provided
     if model_name is None:
@@ -56,31 +59,42 @@ def get_llm(model_name=None, temperature=None, provider=None):
     
     # Handle different providers
     if provider == "anthropic":
-        api_key = settings.get('ANTHROPIC_API_KEY', '')
+        api_key_name = 'ANTHROPIC_API_KEY'
+        api_key = settings.get(api_key_name, '')
         if not api_key:
-            api_key = os.getenv('ANTHROPIC_API_KEY')
+            api_key = os.getenv(api_key_name)
+        if not api_key:
+            api_key = os.getenv("LDR_" + api_key_name)
         if not api_key:
             logger.warning("ANTHROPIC_API_KEY not found. Falling back to default model.")
             return get_fallback_model(temperature)
         
-        return ChatAnthropic(
+        llm = ChatAnthropic(
             model=model_name, anthropic_api_key=api_key, **common_params
         )
+        return wrap_llm_without_think_tags(llm)
     
     elif provider == "openai":
-        api_key = settings.get('OPENAI_API_KEY', '')
+        api_key_name = 'OPENAI_API_KEY'
+        api_key = settings.get(api_key_name, '')
         if not api_key:
-            api_key = os.getenv('OPENAI_API_KEY')
+            api_key = os.getenv(api_key_name)
+        if not api_key:
+            api_key = os.getenv("LDR_" + api_key_name)
         if not api_key:
             logger.warning("OPENAI_API_KEY not found. Falling back to default model.")
             return get_fallback_model(temperature)
         
-        return ChatOpenAI(model=model_name, api_key=api_key, **common_params)
+        llm = ChatOpenAI(model=model_name, api_key=api_key, **common_params)
+        return wrap_llm_without_think_tags(llm)
     
     elif provider == "openai_endpoint":
-        api_key = settings.get('OPENAI_ENDPOINT_API_KEY', '')
+        api_key_name = 'OPENAI_ENDPOINT_API_KEY'
+        api_key = settings.get(api_key_name, '')
         if not api_key:
-            api_key = os.getenv('OPENAI_ENDPOINT_API_KEY')
+            api_key = os.getenv(api_key_name)
+        if not api_key:
+            api_key = os.getenv("LDR_" + api_key_name)
         if not api_key:
             logger.warning("OPENAI_ENDPOINT_API_KEY not found. Falling back to default model.")
             return get_fallback_model(temperature)
@@ -88,16 +102,17 @@ def get_llm(model_name=None, temperature=None, provider=None):
         # Get endpoint URL from settings
         openai_endpoint_url = settings.llm.openai_endpoint_url
         
-        return ChatOpenAI(
+        llm = ChatOpenAI(
             model=model_name, 
             api_key=api_key,
             openai_api_base=openai_endpoint_url, 
             **common_params
         )
+        return wrap_llm_without_think_tags(llm)
     
     elif provider == "vllm":
         try:
-            return VLLM(
+            llm = VLLM(
                 model=model_name,
                 trust_remote_code=True,
                 max_new_tokens=128,
@@ -105,6 +120,7 @@ def get_llm(model_name=None, temperature=None, provider=None):
                 top_p=0.95,
                 temperature=temperature,
             )
+            return wrap_llm_without_think_tags(llm)
         except Exception as e:
             logger.error(f"Error loading VLLM model: {e}")
             logger.warning("Falling back.")
@@ -114,54 +130,54 @@ def get_llm(model_name=None, temperature=None, provider=None):
         try:
             # Use the configurable Ollama base URL
             base_url = settings.get('OLLAMA_BASE_URL', settings.llm.get('ollama_base_url', 'http://localhost:11434'))
-            return ChatOllama(model=model_name, base_url=base_url, **common_params)
+            llm = ChatOllama(model=model_name, base_url=base_url, **common_params)
+            return wrap_llm_without_think_tags(llm)
         except Exception as e:
             logger.error(f"Error loading Ollama model: {e}")
             return get_fallback_model(temperature)
     
     elif provider == "lmstudio":
-
-            # LM Studio supports OpenAI API format, so we can use ChatOpenAI directly
-            lmstudio_url = settings.llm.get('lmstudio_url', "http://localhost:1234")
-            
-            return ChatOpenAI(
-                model=model_name,
-                api_key="lm-studio",  # LM Studio doesn't require a real API key
-                base_url=f"{lmstudio_url}/v1",  # Use the configured URL with /v1 endpoint
-                temperature=temperature,
-                max_tokens=settings.llm.max_tokens
-            )
-
+        # LM Studio supports OpenAI API format, so we can use ChatOpenAI directly
+        lmstudio_url = settings.llm.get('lmstudio_url', "http://localhost:1234")
+        
+        llm = ChatOpenAI(
+            model=model_name,
+            api_key="lm-studio",  # LM Studio doesn't require a real API key
+            base_url=f"{lmstudio_url}/v1",  # Use the configured URL with /v1 endpoint
+            temperature=temperature,
+            max_tokens=settings.llm.max_tokens
+        )
+        return wrap_llm_without_think_tags(llm)
  
     elif provider == "llamacpp":
-
-            # Import LlamaCpp
-            from langchain_community.llms import LlamaCpp
-            
-            # Get LlamaCpp model path from settings
-            model_path = settings.llm.get('llamacpp_model_path', "")
-            if not model_path:
-                logger.error("llamacpp_model_path not set in settings")
-                raise ValueError("llamacpp_model_path not set in settings.toml")
-                
-            # Get additional LlamaCpp parameters
-            n_gpu_layers = settings.llm.get('llamacpp_n_gpu_layers', 1)
-            n_batch = settings.llm.get('llamacpp_n_batch', 512)
-            f16_kv = settings.llm.get('llamacpp_f16_kv', True)
+        # Import LlamaCpp
+        from langchain_community.llms import LlamaCpp
+        
+        # Get LlamaCpp model path from settings
+        model_path = settings.llm.get('llamacpp_model_path', "")
+        if not model_path:
+            logger.error("llamacpp_model_path not set in settings")
+            raise ValueError("llamacpp_model_path not set in settings.toml")
             
-            # Create LlamaCpp instance
-            return LlamaCpp(
-                model_path=model_path,
-                temperature=temperature,
-                max_tokens=settings.llm.max_tokens,
-                n_gpu_layers=n_gpu_layers,
-                n_batch=n_batch,
-                f16_kv=f16_kv,
-                verbose=True
-            )
+        # Get additional LlamaCpp parameters
+        n_gpu_layers = settings.llm.get('llamacpp_n_gpu_layers', 1)
+        n_batch = settings.llm.get('llamacpp_n_batch', 512)
+        f16_kv = settings.llm.get('llamacpp_f16_kv', True)
+        
+        # Create LlamaCpp instance
+        llm = LlamaCpp(
+            model_path=model_path,
+            temperature=temperature,
+            max_tokens=settings.llm.max_tokens,
+            n_gpu_layers=n_gpu_layers,
+            n_batch=n_batch,
+            f16_kv=f16_kv,
+            verbose=True
+        )
+        return wrap_llm_without_think_tags(llm)
     
     else:
-        return get_fallback_model(temperature)
+        return wrap_llm_without_think_tags(get_fallback_model(temperature))
 
 def get_fallback_model(temperature=None):
     """Create a dummy model for when no providers are available"""
@@ -174,6 +190,31 @@ def get_fallback_model(temperature=None):
 # COMPATIBILITY FUNCTIONS
 # ================================
 
+def wrap_llm_without_think_tags(llm):
+    """Create a wrapper class that processes LLM outputs with remove_think_tags"""
+
+    
+    class ProcessingLLMWrapper:
+        def __init__(self, base_llm):
+            self.base_llm = base_llm
+            
+        def invoke(self, *args, **kwargs):
+            response = self.base_llm.invoke(*args, **kwargs)
+            
+            # Process the response content if it has a content attribute
+            if hasattr(response, 'content'):
+                response.content = remove_think_tags(response.content)
+            elif isinstance(response, str):
+                response = remove_think_tags(response)
+                
+            return response
+            
+        # Pass through any other attributes to the base LLM
+        def __getattr__(self, name):
+            return getattr(self.base_llm, name)
+    
+    return ProcessingLLMWrapper(llm)
+
 def get_available_provider_types():
     """Return available model providers"""
     providers = {}

local_deep_research/web_search_engines/search_engine_factory.py

@@ -46,7 +46,10 @@ def create_search_engine(engine_name: str, llm=None, **kwargs) -> Optional[BaseS
         
         # First check environment variable
         api_key = os.getenv(api_key_env)
-        
+        if not api_key:
+            api_key = os.getenv("LDR_" + api_key_env)
+
+
         # If not found in environment, check Dynaconf settings
         if not api_key and api_key_env:
             # Convert env var name to settings path (e.g., BRAVE_API_KEY -> brave_api_key)

{local_deep_research-0.1.24.dist-info → local_deep_research-0.1.26.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: local-deep-research
-Version: 0.1.24
+Version: 0.1.26
 Summary: AI-powered research assistant with deep, iterative analysis using LLMs and web searches
 Author-email: LearningCircuit <185559241+LearningCircuit@users.noreply.github.com>, HashedViking <6432677+HashedViking@users.noreply.github.com>
 License: MIT License
@@ -65,6 +65,7 @@ Requires-Dist: xmltodict>=0.13.0
 Requires-Dist: lxml>=4.9.2
 Requires-Dist: pdfplumber>=0.9.0
 Requires-Dist: unstructured>=0.10.0
+Requires-Dist: google-search-results
 Dynamic: license-file
 
 # Local Deep Research
@@ -79,7 +80,9 @@ A powerful AI-powered research assistant that performs deep, iterative analysis
   </a>
 </div>
 
-## Quick Start
+
+
+## Quick Start (not required if installed with windows installer)
 
 ```bash
 # Install the package
@@ -107,12 +110,12 @@ Access the web interface at `http://127.0.0.1:5000` in your browser.
 
 ## Docker Support
 
-### Build the image first if you haven't already
+Build the image first if you haven't already
 ```bash
 docker build -t local-deep-research .
 ```
 
-### Quick Docker Run
+Quick Docker Run
 
 ```bash
 # Run with default settings (connects to Ollama running on the host)
@@ -126,6 +129,31 @@ For comprehensive Docker setup information, see:
 - [Docker Usage Guide](https://github.com/LearningCircuit/local-deep-research/blob/main/docs/docker-usage-readme.md)
 - [Docker Compose Guide](https://github.com/LearningCircuit/local-deep-research/blob/main/docs/docker-compose-guide.md)
 
+## Programmatic Access
+
+Local Deep Research now provides a simple API for programmatic access to its research capabilities:
+
+```python
+import os
+# Set environment variables to control the LLM
+os.environ["LDR_LLM__MODEL"] = "mistral"     # Specify model name
+
+from local_deep_research import quick_summary, generate_report, analyze_documents
+
+# Generate a quick research summary with custom parameters
+results = quick_summary(
+    query="advances in fusion energy",
+    search_tool="auto",          # Auto-select the best search engine
+    iterations=1,                # Single research cycle for speed
+    questions_per_iteration=2,   # Generate 2 follow-up questions
+    max_results=30,              # Consider up to 30 search results
+    temperature=0.7              # Control creativity of generation
+)
+print(results["summary"])
+```
+
+These functions provide flexible options for customizing the search parameters, iterations, and output formats. For more examples, see the [programmatic access tutorial](https://github.com/LearningCircuit/local-deep-research/blob/main/examples/programmatic_access.ipynb).
+
 
 ## Features
 
@@ -308,27 +336,6 @@ You can use local document search in several ways:
 3. **All collections**: Set `tool = "local_all"` to search across all collections
 4. **Query syntax**: Type `collection:project_docs your query` to target a specific collection
 
-## Programmatic Access
-
-Local Deep Research now provides a simple API for programmatic access to its research capabilities:
-
-```python
-from local_deep_research import quick_summary, generate_report
-
-# Generate a quick research summary
-results = quick_summary("advances in fusion energy")
-print(results["summary"])
-
-# Create a comprehensive structured report
-report = generate_report("impact of quantum computing on cryptography")
-print(report["content"])
-
-# Analyze documents in a local collection
-from local_deep_research import analyze_documents
-docs = analyze_documents("renewable energy", "research_papers")
-```
-
-These functions provide flexible options for customizing the search parameters, iterations, and output formats. For more examples, see the [programmatic access tutorial](https://github.com/LearningCircuit/local-deep-research/blob/programmatic-access/examples/programmatic_access.ipynb).
 
 ## Advanced Configuration
 

{local_deep_research-0.1.24.dist-info → local_deep_research-0.1.26.dist-info}/RECORD

@@ -1,14 +1,14 @@
 local_deep_research/__init__.py,sha256=Grde0sFEYyCXrPCfxd-9b9v1M6OurrzQbVYRmKQ9E7w,886
 local_deep_research/citation_handler.py,sha256=v_fwTy-2XvUuoH3OQRzmBrvaiN7mBk8jbNfySslmt5g,4357
-local_deep_research/config.py,sha256=3g8-QPMrxoIMjHvyjSJBFUELmAIyOQFHApUnd8p50a8,9881
+local_deep_research/config.py,sha256=FQxtlhw49O-f5iImst84-JwhabfYr-xV-YMEh6hkAwU,10516
 local_deep_research/main.py,sha256=uQXtGQ6LtZNd5Qw63D5ke4Q_LjYimouWVSUknVsk3JQ,3645
 local_deep_research/report_generator.py,sha256=EvaArnWirMgg42fMzmZeJczoEYujEbJ2ryHHYuuoXx8,8058
 local_deep_research/search_system.py,sha256=yY3BEzX68vdtUcYF9h6lC3yVao0YA_NSBj6W3-RwlKk,15459
 local_deep_research/api/__init__.py,sha256=H0WGFSohUR0T2QswtWngPZWoMYPs9VWQTQYaivAlrJU,440
 local_deep_research/api/research_functions.py,sha256=Z23wZYsB1x2ivdFYJ9uqIqCAwjR2RdOff7Bq30DxQYU,12099
-local_deep_research/defaults/.env.template,sha256=U4B_InwGZl4IVuAdbY_u0nKN_akHtebMBwUU_e_eljc,427
+local_deep_research/defaults/.env.template,sha256=_eVCy4d_XwpGXy8n50CG3wH9xx2oqJCFKS7IbqgInDk,491
 local_deep_research/defaults/__init__.py,sha256=2Vvlkl-gmP_qPYWegE4JBgummypogl3VXrQ1XzptFDU,1381
-local_deep_research/defaults/llm_config.py,sha256=Ql0euemgLw_Uwg5g05sA1SkVzAYK7O_ZAnnBi3rsAi4,10095
+local_deep_research/defaults/llm_config.py,sha256=1KiW9k8kmsUD5u9VgEdgWZBNMmK1BA0ZxoGbuC2spAk,11652
 local_deep_research/defaults/local_collections.toml,sha256=zNa03PVnFrZ757JdZOuW6QDxkOc6ep5tG8baGBrMmXM,1778
 local_deep_research/defaults/main.toml,sha256=6Lzbc5sVLxMwu83bLBp_tpYOZgmtThCfPL1L42eTGro,1939
 local_deep_research/defaults/search_engines.toml,sha256=g0-qrw10oMgW74z_lYpPDkGwMje25mvalfY1EJ0nL3g,8134
@@ -34,7 +34,7 @@ local_deep_research/web/templates/settings.html,sha256=S9A-tdpzMhP2Zw7kp2jxKlwaW
 local_deep_research/web/templates/settings_dashboard.html,sha256=De-v1KNdVvkXme5i3YZ6sIfU9aAKDc_N-AW9n4PZoso,9109
 local_deep_research/web_search_engines/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 local_deep_research/web_search_engines/search_engine_base.py,sha256=QmhfjuHK2deomh8tARghKuYnF-5t3wwBB661odS2VtU,8065
-local_deep_research/web_search_engines/search_engine_factory.py,sha256=Sld6bYTwcyTxgVLx04t00sD7vfJhSHFOl6iiGJ08ZUE,11118
+local_deep_research/web_search_engines/search_engine_factory.py,sha256=8REYoRdDWvB6XLhBym8rqzuULX28VQ-UKWNcRA5tLTQ,11189
 local_deep_research/web_search_engines/search_engines_config.py,sha256=5C0tCmy_Jpv1YHLZLlyS7h5B2XToYcWPAaBDEOsxMo0,2739
 local_deep_research/web_search_engines/engines/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 local_deep_research/web_search_engines/engines/full_search.py,sha256=BuOz8dX-XocazCG7gGBKFnIY99FZtNFI0-Wq3fhsfp4,4689
@@ -53,9 +53,9 @@ local_deep_research/web_search_engines/engines/search_engine_semantic_scholar.py
 local_deep_research/web_search_engines/engines/search_engine_serpapi.py,sha256=XikEYnM-pAaR70VeAJ28lbqpRzCj4bCA9xY29taTV8g,9215
 local_deep_research/web_search_engines/engines/search_engine_wayback.py,sha256=astAvSLajDZ6rwgthJ3iBcHSWuDSYPO7uilIxaJhXmU,18132
 local_deep_research/web_search_engines/engines/search_engine_wikipedia.py,sha256=KSGJECbEcxZpVK-PhYsTCtzedSK0l1AjQmvGtx8KBks,9799
-local_deep_research-0.1.24.dist-info/licenses/LICENSE,sha256=Qg2CaTdu6SWnSqk1_JtgBPp_Da-LdqJDhT1Vt1MUc5s,1072
-local_deep_research-0.1.24.dist-info/METADATA,sha256=mM-b8LezrBKyR-VjOYE0lMqoKzEwfvID2kMYgUJH9Z4,17096
-local_deep_research-0.1.24.dist-info/WHEEL,sha256=CmyFI0kx5cdEMTLiONQRbGQwjIoR1aIYB7eCAQ4KPJ0,91
-local_deep_research-0.1.24.dist-info/entry_points.txt,sha256=u-Y6Z3MWtR3dmsTDFYhXyfkPv7mALUA7YAnY4Fi1XDs,97
-local_deep_research-0.1.24.dist-info/top_level.txt,sha256=h6-uVE_wSuLOcoWwT9szhX23mBWufu77MqmM25UfbCY,20
-local_deep_research-0.1.24.dist-info/RECORD,,
+local_deep_research-0.1.26.dist-info/licenses/LICENSE,sha256=Qg2CaTdu6SWnSqk1_JtgBPp_Da-LdqJDhT1Vt1MUc5s,1072
+local_deep_research-0.1.26.dist-info/METADATA,sha256=lxkwHdoKtXh2HlxFL_XZ6NGn6aWm-mVjBQR7GclJpqE,17384
+local_deep_research-0.1.26.dist-info/WHEEL,sha256=CmyFI0kx5cdEMTLiONQRbGQwjIoR1aIYB7eCAQ4KPJ0,91
+local_deep_research-0.1.26.dist-info/entry_points.txt,sha256=u-Y6Z3MWtR3dmsTDFYhXyfkPv7mALUA7YAnY4Fi1XDs,97
+local_deep_research-0.1.26.dist-info/top_level.txt,sha256=h6-uVE_wSuLOcoWwT9szhX23mBWufu77MqmM25UfbCY,20
+local_deep_research-0.1.26.dist-info/RECORD,,