local-deep-research 0.5.7__py3-none-any.whl → 0.6.0__py3-none-any.whl

local_deep_research/__version__.py

@@ -1 +1 @@
-__version__ = "0.5.7"
+__version__ = "0.6.0"

local_deep_research/advanced_search_system/candidate_exploration/progressive_explorer.py

@@ -236,6 +236,11 @@ class ProgressiveExplorer:
         """Execute searches in parallel and return results."""
         results = []
 
+        # Import context preservation utility
+        from ...utilities.thread_context import (
+            create_context_preserving_wrapper,
+        )
+
         def search_query(query):
             try:
                 search_results = self.search_engine.run(query)
@@ -244,11 +249,16 @@ class ProgressiveExplorer:
                 logger.error(f"Error searching '{query}': {str(e)}")
                 return (query, [])
 
+        # Create context-preserving wrapper for the search function
+        context_aware_search = create_context_preserving_wrapper(search_query)
+
         # Run searches in parallel
         with concurrent.futures.ThreadPoolExecutor(
             max_workers=max_workers
         ) as executor:
-            futures = [executor.submit(search_query, q) for q in queries]
+            futures = [
+                executor.submit(context_aware_search, q) for q in queries
+            ]
             for future in concurrent.futures.as_completed(futures):
                 results.append(future.result())
 

local_deep_research/advanced_search_system/questions/browsecomp_question.py

@@ -144,15 +144,25 @@ DESCRIPTORS: [entity1], [entity2], ...
         # 1. Original query (always include)
         searches.append(query)
 
+        # If only 1 question requested, return just the original query
+        if num_questions <= 1:
+            return searches[:1]
+
         # 2. Domain exploration searches (combine key entities)
-        if entities["names"]:
+        if entities["names"] and len(searches) < num_questions:
             for name in entities["names"][:2]:  # Top 2 names
+                if len(searches) >= num_questions:
+                    break
                 searches.append(f"{name}")
-                if entities["descriptors"]:
+                if entities["descriptors"] and len(searches) < num_questions:
                     searches.append(f"{name} {entities['descriptors'][0]}")
 
         # 3. Temporal searches if years are important
-        if entities["temporal"] and len(entities["temporal"]) <= 10:
+        if (
+            entities["temporal"]
+            and len(entities["temporal"]) <= 10
+            and len(searches) < num_questions
+        ):
             # For small year ranges, search each year with a key term
             key_term = (
                 entities["names"][0]
@@ -162,14 +172,18 @@ DESCRIPTORS: [entity1], [entity2], ...
                 else ""
             )
             for year in entities["temporal"][:5]:  # Limit to 5 years initially
+                if len(searches) >= num_questions:
+                    break
                 if key_term:
                     searches.append(f"{key_term} {year}")
 
         # 4. Location-based searches
-        if entities["locations"]:
+        if entities["locations"] and len(searches) < num_questions:
             for location in entities["locations"][:2]:
+                if len(searches) >= num_questions:
+                    break
                 searches.append(f"{location}")
-                if entities["descriptors"]:
+                if entities["descriptors"] and len(searches) < num_questions:
                     searches.append(f"{location} {entities['descriptors'][0]}")
 
         # Remove duplicates and limit to requested number
@@ -179,6 +193,8 @@ DESCRIPTORS: [entity1], [entity2], ...
             if s.lower() not in seen:
                 seen.add(s.lower())
                 unique_searches.append(s)
+                if len(unique_searches) >= num_questions:
+                    break
 
         return unique_searches[:num_questions]
 
@@ -238,21 +254,26 @@ Format: One search per line
                 if line:
                     searches.append(line)
 
-        # Ensure we have enough searches
+        # Ensure we have enough searches, but respect the limit
         while len(searches) < num_questions:
             # Generate combinations programmatically
             if iteration <= 5 and entities["temporal"]:
                 # Continue with year-based searches
+                added_any = False
                 for year in entities["temporal"]:
                     if not self._was_searched(year, questions_by_iteration):
                         base_term = (
                             entities["names"][0] if entities["names"] else ""
                         )
                         searches.append(f"{base_term} {year}".strip())
+                        added_any = True
                         if len(searches) >= num_questions:
                             break
+                if not added_any:
+                    break  # No more year searches to add
             else:
                 # Combine multiple constraints
+                added_any = False
                 if entities["names"] and entities["descriptors"]:
                     for name in entities["names"]:
                         for desc in entities["descriptors"]:
@@ -261,8 +282,13 @@ Format: One search per line
                                 combo, questions_by_iteration
                             ):
                                 searches.append(combo)
+                                added_any = True
                                 if len(searches) >= num_questions:
                                     break
+                        if len(searches) >= num_questions:
+                            break
+                if not added_any:
+                    break  # No more combinations to add
 
         return searches[:num_questions]
 

local_deep_research/advanced_search_system/strategies/focused_iteration_strategy.py

@@ -53,9 +53,9 @@ class FocusedIterationStrategy(BaseSearchStrategy):
         search=None,
         citation_handler=None,
         all_links_of_system=None,
-        max_iterations: int = 8,  # OPTIMAL FOR SIMPLEQA: 96.51% accuracy achieved
+        max_iterations: int = 8,  # OPTIMAL FOR SIMPLEQA: 90%+ accuracy achieved
         questions_per_iteration: int = 5,  # OPTIMAL FOR SIMPLEQA: proven config
-        use_browsecomp_optimization: bool = True,  # Can be False for pure SimpleQA
+        use_browsecomp_optimization: bool = True,  # True for 90%+ accuracy with forced_answer handler
     ):
         """Initialize with components optimized for focused iteration."""
         super().__init__(all_links_of_system)
@@ -63,9 +63,9 @@ class FocusedIterationStrategy(BaseSearchStrategy):
         self.model = model or get_llm()
         self.progress_callback = None
 
-        # Configuration
-        self.max_iterations = max_iterations
-        self.questions_per_iteration = questions_per_iteration
+        # Configuration - ensure these are integers
+        self.max_iterations = int(max_iterations)
+        self.questions_per_iteration = int(questions_per_iteration)
         self.use_browsecomp_optimization = use_browsecomp_optimization
 
         # Initialize specialized components
@@ -158,9 +158,11 @@ class FocusedIterationStrategy(BaseSearchStrategy):
                         questions_by_iteration=self.questions_by_iteration,
                     )
 
-                # Always include original query in first iteration
+                # Always include original query in first iteration, but respect question limit
                 if iteration == 1 and query not in questions:
                     questions = [query] + questions
+                    # Trim to respect questions_per_iteration limit
+                    questions = questions[: self.questions_per_iteration]
 
                 self.questions_by_iteration[iteration] = questions
                 logger.info(f"Iteration {iteration} questions: {questions}")
@@ -247,6 +249,7 @@ class FocusedIterationStrategy(BaseSearchStrategy):
 
                 # Accumulate all results (no filtering!)
                 self.all_search_results.extend(iteration_results)
+                self.all_links_of_system.extend(iteration_results)
 
                 # Update progress
                 self._update_progress(
@@ -356,6 +359,11 @@ class FocusedIterationStrategy(BaseSearchStrategy):
         """Execute searches in parallel (like source-based strategy)."""
         all_results = []
 
+        # Import context preservation utility
+        from ...utilities.thread_context import (
+            create_context_preserving_wrapper,
+        )
+
         def search_question(q):
             try:
                 result = self.search.run(q)
@@ -364,11 +372,18 @@ class FocusedIterationStrategy(BaseSearchStrategy):
                 logger.error(f"Error searching '{q}': {str(e)}")
                 return {"question": q, "results": [], "error": str(e)}
 
+        # Create context-preserving wrapper for the search function
+        context_aware_search = create_context_preserving_wrapper(
+            search_question
+        )
+
         # Run searches in parallel
         with concurrent.futures.ThreadPoolExecutor(
             max_workers=len(queries)
         ) as executor:
-            futures = [executor.submit(search_question, q) for q in queries]
+            futures = [
+                executor.submit(context_aware_search, q) for q in queries
+            ]
 
             for future in concurrent.futures.as_completed(futures):
                 result_dict = future.result()
@@ -384,6 +399,11 @@ class FocusedIterationStrategy(BaseSearchStrategy):
         completed_searches = 0
         total_searches = len(queries)
 
+        # Import context preservation utility
+        from ...utilities.thread_context import (
+            create_context_preserving_wrapper,
+        )
+
         def search_question_with_progress(q):
             nonlocal completed_searches
             try:
@@ -439,12 +459,17 @@ class FocusedIterationStrategy(BaseSearchStrategy):
                     "result_count": 0,
                 }
 
+        # Create context-preserving wrapper for the search function
+        context_aware_search_with_progress = create_context_preserving_wrapper(
+            search_question_with_progress
+        )
+
         # Run searches in parallel
         with concurrent.futures.ThreadPoolExecutor(
             max_workers=min(len(queries), 5)
         ) as executor:
             futures = [
-                executor.submit(search_question_with_progress, q)
+                executor.submit(context_aware_search_with_progress, q)
                 for q in queries
             ]
 

local_deep_research/advanced_search_system/strategies/source_based_strategy.py

@@ -8,6 +8,7 @@ from ...config.llm_config import get_llm
 from ...config.search_config import get_search
 from ...utilities.db_utils import get_db_setting
 from ...utilities.threading_utils import thread_context, thread_with_app_context
+from ...utilities.thread_context import preserve_research_context
 from ..filters.cross_engine_filter import CrossEngineFilter
 from ..findings.repository import FindingsRepository
 from ..questions.atomic_fact_question import AtomicFactQuestionGenerator
@@ -211,6 +212,7 @@ class SourceBasedSearchStrategy(BaseSearchStrategy):
 
                 # Function for thread pool
                 @thread_with_app_context
+                @preserve_research_context
                 def search_question(q):
                     try:
                         result = self.search.run(q)

local_deep_research/api/__init__.py

@@ -5,12 +5,14 @@ API module for programmatic access to Local Deep Research functionality.
 
 from .research_functions import (
     analyze_documents,
+    detailed_research,
     generate_report,
     quick_summary,
 )
 
 __all__ = [
     "quick_summary",
+    "detailed_research",
     "generate_report",
     "analyze_documents",
 ]

local_deep_research/api/research_functions.py

@@ -3,7 +3,8 @@ API module for Local Deep Research.
 Provides programmatic access to search and research capabilities.
 """
 
-from typing import Any, Callable, Dict, Optional
+from datetime import datetime
+from typing import Any, Callable, Dict, Optional, Union
 
 from loguru import logger
 
@@ -24,6 +25,8 @@ def _init_search_system(
     search_strategy: str = "source_based",
     iterations: int = 1,
     questions_per_iteration: int = 1,
+    retrievers: Optional[Dict[str, Any]] = None,
+    llms: Optional[Dict[str, Any]] = None,
 ) -> AdvancedSearchSystem:
     """
     Initializes the advanced search system with specified parameters. This function sets up
@@ -43,11 +46,30 @@ def _init_search_system(
         iterations: Number of research cycles to perform
         questions_per_iteration: Number of questions to generate per cycle
         search_strategy: The name of the search strategy to use.
+        retrievers: Optional dictionary of {name: retriever} pairs to use as search engines
+        llms: Optional dictionary of {name: llm} pairs to use as language models
 
     Returns:
         AdvancedSearchSystem: An instance of the configured AdvancedSearchSystem.
 
     """
+    # Register retrievers if provided
+    if retrievers:
+        from ..web_search_engines.retriever_registry import retriever_registry
+
+        retriever_registry.register_multiple(retrievers)
+        logger.info(
+            f"Registered {len(retrievers)} retrievers: {list(retrievers.keys())}"
+        )
+
+    # Register LLMs if provided
+    if llms:
+        from ..llm import register_llm
+
+        for name, llm_instance in llms.items():
+            register_llm(name, llm_instance)
+        logger.info(f"Registered {len(llms)} LLMs: {list(llms.keys())}")
+
     # Get language model with custom temperature
     llm = get_llm(
         temperature=temperature,
@@ -84,6 +106,9 @@ def _init_search_system(
 
 def quick_summary(
     query: str,
+    research_id: Optional[Union[int, str]] = None,
+    retrievers: Optional[Dict[str, Any]] = None,
+    llms: Optional[Dict[str, Any]] = None,
     **kwargs: Any,
 ) -> Dict[str, Any]:
     """
@@ -91,6 +116,9 @@ def quick_summary(
 
     Args:
         query: The research query to analyze
+        research_id: Optional research ID (int or UUID string) for tracking metrics
+        retrievers: Optional dictionary of {name: retriever} pairs to use as search engines
+        llms: Optional dictionary of {name: llm} pairs to use as language models
         **kwargs: Configuration for the search system. Will be forwarded to
             `_init_search_system()`.
 
@@ -103,7 +131,46 @@ def quick_summary(
     """
     logger.info("Generating quick summary for query: %s", query)
 
-    system = _init_search_system(**kwargs)
+    # Generate a research_id if none provided
+    if research_id is None:
+        import uuid
+
+        research_id = str(uuid.uuid4())
+        logger.debug(f"Generated research_id: {research_id}")
+
+    # Register retrievers if provided
+    if retrievers:
+        from ..web_search_engines.retriever_registry import retriever_registry
+
+        retriever_registry.register_multiple(retrievers)
+        logger.info(
+            f"Registered {len(retrievers)} retrievers: {list(retrievers.keys())}"
+        )
+
+    # Register LLMs if provided
+    if llms:
+        from ..llm import register_llm
+
+        for name, llm_instance in llms.items():
+            register_llm(name, llm_instance)
+        logger.info(f"Registered {len(llms)} LLMs: {list(llms.keys())}")
+
+    # Set search context with research_id
+    from ..metrics.search_tracker import set_search_context
+
+    search_context = {
+        "research_id": research_id,  # Pass UUID or integer directly
+        "research_query": query,
+        "research_mode": kwargs.get("research_mode", "quick"),
+        "research_phase": "init",
+        "search_iteration": 0,
+        "search_engine_selected": kwargs.get("search_tool"),
+    }
+    set_search_context(search_context)
+
+    # Remove research_mode from kwargs before passing to _init_search_system
+    init_kwargs = {k: v for k, v in kwargs.items() if k != "research_mode"}
+    system = _init_search_system(llms=llms, **init_kwargs)
 
     # Perform the search and analysis
     results = system.analyze_topic(query)
@@ -130,6 +197,8 @@ def generate_report(
     output_file: Optional[str] = None,
     progress_callback: Optional[Callable] = None,
     searches_per_section: int = 2,
+    retrievers: Optional[Dict[str, Any]] = None,
+    llms: Optional[Dict[str, Any]] = None,
     **kwargs: Any,
 ) -> Dict[str, Any]:
     """
@@ -141,6 +210,8 @@ def generate_report(
         progress_callback: Optional callback function to receive progress updates
         searches_per_section: The number of searches to perform for each
             section in the report.
+        retrievers: Optional dictionary of {name: retriever} pairs to use as search engines
+        llms: Optional dictionary of {name: llm} pairs to use as language models
 
     Returns:
         Dictionary containing the research report with keys:
@@ -149,7 +220,24 @@ def generate_report(
     """
     logger.info("Generating comprehensive research report for query: %s", query)
 
-    system = _init_search_system(**kwargs)
+    # Register retrievers if provided
+    if retrievers:
+        from ..web_search_engines.retriever_registry import retriever_registry
+
+        retriever_registry.register_multiple(retrievers)
+        logger.info(
+            f"Registered {len(retrievers)} retrievers: {list(retrievers.keys())}"
+        )
+
+    # Register LLMs if provided
+    if llms:
+        from ..llm import register_llm
+
+        for name, llm_instance in llms.items():
+            register_llm(name, llm_instance)
+        logger.info(f"Registered {len(llms)} LLMs: {list(llms.keys())}")
+
+    system = _init_search_system(retrievers=retrievers, llms=llms, **kwargs)
 
     # Set progress callback if provided
     if progress_callback:
@@ -175,6 +263,92 @@ def generate_report(
     return report
 
 
+def detailed_research(
+    query: str,
+    research_id: Optional[Union[int, str]] = None,
+    retrievers: Optional[Dict[str, Any]] = None,
+    llms: Optional[Dict[str, Any]] = None,
+    **kwargs: Any,
+) -> Dict[str, Any]:
+    """
+    Perform detailed research with comprehensive analysis.
+
+    Similar to generate_report but returns structured data instead of markdown.
+
+    Args:
+        query: The research query to analyze
+        research_id: Optional research ID (int or UUID string) for tracking metrics
+        retrievers: Optional dictionary of {name: retriever} pairs to use as search engines
+        llms: Optional dictionary of {name: llm} pairs to use as language models
+        **kwargs: Configuration for the search system
+
+    Returns:
+        Dictionary containing detailed research results
+    """
+    logger.info("Performing detailed research for query: %s", query)
+
+    # Generate a research_id if none provided
+    if research_id is None:
+        import uuid
+
+        research_id = str(uuid.uuid4())
+        logger.debug(f"Generated research_id: {research_id}")
+
+    # Register retrievers if provided
+    if retrievers:
+        from ..web_search_engines.retriever_registry import retriever_registry
+
+        retriever_registry.register_multiple(retrievers)
+        logger.info(
+            f"Registered {len(retrievers)} retrievers: {list(retrievers.keys())}"
+        )
+
+    # Register LLMs if provided
+    if llms:
+        from ..llm import register_llm
+
+        for name, llm_instance in llms.items():
+            register_llm(name, llm_instance)
+        logger.info(f"Registered {len(llms)} LLMs: {list(llms.keys())}")
+
+    # Set search context
+    from ..metrics.search_tracker import set_search_context
+
+    search_context = {
+        "research_id": research_id,
+        "research_query": query,
+        "research_mode": "detailed",
+        "research_phase": "init",
+        "search_iteration": 0,
+        "search_engine_selected": kwargs.get("search_tool"),
+    }
+    set_search_context(search_context)
+
+    # Initialize system
+    system = _init_search_system(retrievers=retrievers, llms=llms, **kwargs)
+
+    # Perform detailed research
+    results = system.analyze_topic(query)
+
+    # Return comprehensive results
+    return {
+        "query": query,
+        "research_id": research_id,
+        "summary": results.get("current_knowledge", ""),
+        "findings": results.get("findings", []),
+        "iterations": results.get("iterations", 0),
+        "questions": results.get("questions", {}),
+        "formatted_findings": results.get("formatted_findings", ""),
+        "sources": results.get("all_links_of_system", []),
+        "metadata": {
+            "timestamp": datetime.now().isoformat(),
+            "search_tool": kwargs.get("search_tool", "auto"),
+            "iterations_requested": kwargs.get("iterations", 1),
+            "strategy": kwargs.get("search_strategy", "source_based"),
+        },
+    }
+
+
 def analyze_documents(
     query: str,
     collection_name: str,