updated logic for vendor neutral llm provider. moved llm provider to separate file. updated requirements.txt with correct version

config/settings.yaml

@@ -2,9 +2,10 @@
 
 # LLM Settings
 llm:
+  provider: deepseek  # Supported values: openai, deepseek
   temperature: 0.5
   max_tokens: 4096
-  model: deepseek-r1
+  model: deepseek-r1  # Model name specific to the provider
 
 # ArXiv Client Settings
 arxiv:

requirements.txt

@@ -1,12 +1,14 @@
-langchain==0.2.1
+langchain==0.3.17
+langchain-community>=0.0.10
+openai>=1.0.0
 arxiv==2.1.0
-unstructured==0.12.2
-deepseek-ai==1.0.0
+unstructured==0.16.17
+deepseek==1.0.0
 faiss-cpu==1.8.0
 sqlalchemy==2.0.30
 python-dotenv==1.0.0
 pyyaml==6.0.1
-pypdf==4.2.0
+pypdf==5.2.0
 redis==5.0.3
-chromadb==0.5.0
+chromadb==0.6.3
 requests>=2.31.0

src/analysis/analysis_engine.py

@@ -1,9 +1,8 @@
 from langchain.schema import Document
 from langchain.chains.summarize import load_summarize_chain
-from langchain_community.llms import OpenAI
 from typing import List, Dict
 import yaml
-import os
+from .llm_provider import create_llm_provider
 
 class TechnicalConceptExtractor:
     def __init__(self, llm):
@@ -31,15 +30,17 @@ class PaperAnalyzer:
     def __init__(self, config_path: str = "config/settings.yaml"):
         with open(config_path) as f:
             self.config = yaml.safe_load(f)
-            
-        self.llm = OpenAI(
-            temperature=0.5,  # More conservative temperature for academic analysis
-            model_name="deepseek-r1",
-            api_key=os.getenv("DEEPSEEK_API_KEY"),
-            base_url="https://api.deepseek.com/v1",
-            max_tokens=4096  # Increased token limit for complex papers
+        
+        # Create LLM provider based on configuration
+        llm_config = self.config.get("llm", {})
+        provider_type = llm_config.get("provider", "openai")
+        self.llm = create_llm_provider(provider_type, llm_config).get_llm()
+        
+        # Initialize analysis components
+        self.summary_chain = load_summarize_chain(
+            self.llm, 
+            chain_type=self.config.get("analysis", {}).get("summary_chain_type", "map_reduce")
         )
-        self.summary_chain = load_summarize_chain(self.llm, chain_type="map_reduce")
         self.concept_extractor = TechnicalConceptExtractor(self.llm)
         self.insight_aggregator = InsightAggregator(self.llm)
 

src/analysis/llm_provider.py

@@ -0,0 +1,50 @@
+from abc import ABC, abstractmethod
+from typing import Dict, Optional
+from langchain.base_language import BaseLanguageModel
+from langchain_community.llms import OpenAI
+import os
+
+class LLMProvider(ABC):
+    """Abstract base class for LLM providers"""
+    @abstractmethod
+    def get_llm(self) -> BaseLanguageModel:
+        """Return configured LLM instance"""
+        pass
+
+class OpenAIProvider(LLMProvider):
+    def __init__(self, config: Dict):
+        self.config = config
+        
+    def get_llm(self) -> BaseLanguageModel:
+        return OpenAI(
+            temperature=self.config.get("temperature", 0.5),
+            model_name=self.config.get("model", "gpt-3.5-turbo"),
+            api_key=os.getenv("OPENAI_API_KEY"),
+            max_tokens=self.config.get("max_tokens", 2048)
+        )
+
+class DeepseekProvider(LLMProvider):
+    def __init__(self, config: Dict):
+        self.config = config
+        
+    def get_llm(self) -> BaseLanguageModel:
+        return OpenAI(
+            temperature=self.config.get("temperature", 0.5),
+            model_name=self.config.get("model", "deepseek-r1"),
+            api_key=os.getenv("DEEPSEEK_API_KEY"),
+            base_url="https://api.deepseek.com/v1",
+            max_tokens=self.config.get("max_tokens", 4096)
+        )
+
+def create_llm_provider(provider_type: str, config: Dict) -> LLMProvider:
+    """Factory function to create LLM provider instances"""
+    providers = {
+        "openai": OpenAIProvider,
+        "deepseek": DeepseekProvider
+    }
+    
+    provider_class = providers.get(provider_type.lower())
+    if not provider_class:
+        raise ValueError(f"Unsupported LLM provider: {provider_type}")
+        
+    return provider_class(config)