Refactor cache, logger paths and clean up imports

Updated cache and log directory paths to use '.mllmcelltype' for consistency. Removed unused LangExtract-related code from __init__.py. Cleaned up imports in functions.py. Improved consensus handling and logging in consensus.py. Added a warning for ignored base_url in Gemini provider.

Author: cafferychen777

python/mllmcelltype/__init__.py

@@ -38,18 +38,6 @@
     save_to_cache,
 )
 
-# LangExtract components (simplified)
-
-# LangExtract parser module (optional dependency)
-try:
-    # LangExtract parser is available but not exposed at package level
-    # Individual components can be imported directly if needed
-    import mllmcelltype.langextract_parser  # noqa: F401
-
-    LANGEXTRACT_AVAILABLE = True
-except ImportError:
-    LANGEXTRACT_AVAILABLE = False
-
 __version__ = "1.3.3"
 
 __all__ = [
@@ -90,24 +78,4 @@
     "resolve_provider_base_url",
     "get_default_api_url",
     "validate_base_url",
-    # LangExtract configuration (removed undefined functions)
 ]
-
-# Add LangExtract functionality if available
-if LANGEXTRACT_AVAILABLE:
-    __all__.extend(
-        [
-            # LangExtract Parser
-            "LangextractParser",
-            "ParsingConfig",
-            "CellTypeAnnotation",
-            "ConsensusMetrics",
-            "BatchAnnotationResult",
-            "DiscussionAnalysis",
-            "ParsingComplexity",
-            "create_parser",
-            "parse_cell_types",
-            "analyze_consensus",
-            "LANGEXTRACT_AVAILABLE",
-        ]
-    )

python/mllmcelltype/cache_manager.py

@@ -15,7 +15,7 @@
 
 def clear_mllmcelltype_cache():
     """Clear the mLLMCelltype cache directory."""
-    cache_dir = os.path.join(os.path.expanduser("~"), ".llmcelltype", "cache")
+    cache_dir = os.path.join(os.path.expanduser("~"), ".mllmcelltype", "cache")
 
     if os.path.exists(cache_dir):
         print(f"Found cache directory: {cache_dir}")
@@ -38,7 +38,7 @@ def clear_mllmcelltype_cache():
 
 def get_cache_info():
     """Get information about the current cache state."""
-    cache_dir = os.path.join(os.path.expanduser("~"), ".llmcelltype", "cache")
+    cache_dir = os.path.join(os.path.expanduser("~"), ".mllmcelltype", "cache")
 
     if not os.path.exists(cache_dir):
         return {"exists": False, "path": cache_dir, "file_count": 0, "total_size": 0}

python/mllmcelltype/consensus.py

@@ -549,6 +549,7 @@ def process_controversial_clusters(
         final_decision = None
         rounds_history = []
         current_votes = model_votes.copy()
+        consensus_response = None  # Initialize for later use
 
         # Create initial consensus check prompt for LLM to calculate metrics
 
@@ -728,7 +729,7 @@ def process_controversial_clusters(
                 # Try to extract majority_prediction from the last consensus check
                 # Only try to access consensus_response if it was actually created in this iteration
                 last_consensus_check = None
-                if rounds_history and len(rounds_history) >= 1 and "consensus_response" in locals():
+                if rounds_history and consensus_response is not None:
                     # Get the response from the last consensus check
                     last_consensus_check = consensus_response
 
@@ -1121,6 +1122,7 @@ def interactive_consensus_annotation(
 
     # If there are controversial clusters, resolve them
     resolved = {}
+    discussion_logs = {}
     if controversial:
         # Choose model for discussion
         discussion_model = None
@@ -1232,7 +1234,7 @@ def interactive_consensus_annotation(
         "controversial_clusters": controversial,
         "resolved": resolved,
         "model_annotations": model_results,
-        "discussion_logs": discussion_logs if "discussion_logs" in locals() else {},
+        "discussion_logs": discussion_logs,
         "metadata": {
             "timestamp": time.strftime("%Y-%m-%d %H:%M:%S"),
             "models": models,

python/mllmcelltype/functions.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-from .logger import write_log
 from .providers import (
     process_anthropic,
     process_deepseek,
@@ -13,7 +12,7 @@
     process_stepfun,
     process_zhipu,
 )
-from .utils import clean_annotation, find_agreement
+from .utils import find_agreement
 
 # Global provider function mapping for reuse across modules
 PROVIDER_FUNCTIONS = {

python/mllmcelltype/logger.py

@@ -18,7 +18,7 @@
 logger = logging.getLogger("llmcelltype")
 
 # Default log directory
-DEFAULT_LOG_DIR = os.path.join(os.path.expanduser("~"), ".llmcelltype", "logs")
+DEFAULT_LOG_DIR = os.path.join(os.path.expanduser("~"), ".mllmcelltype", "logs")
 
 
 def setup_logging(log_dir: Optional[str] = None, log_level: str = "INFO") -> None:

python/mllmcelltype/providers/gemini.py

@@ -26,6 +26,13 @@ def process_gemini(
     """
     write_log(f"Starting Gemini API request with model: {model}")
 
+    # Warn if base_url is provided (Gemini SDK doesn't support custom URLs)
+    if base_url:
+        write_log(
+            "base_url parameter is ignored for Gemini (SDK doesn't support custom URLs)",
+            level="warning",
+        )
+
     # Check if API key is provided and not empty
     if not api_key:
         error_msg = "Google API key is missing or empty"