add logging

app.py

@@ -26,7 +26,7 @@ dspy.settings.configure(lm=turbo, rm=rm)
 Settings.llm = OpenAI(model='gpt-3.5-turbo')
 
 sm = SemanticCaching(model_name='sentence-transformers/all-mpnet-base-v2', dimension=768,
-                     json_file='rag_test_cache.json', cosine_threshold=.90)
+                     json_file='rag_test_cache.json')
 sm.load_cache()
 
 # Initialize RAGWorkflow with indexer
@@ -41,7 +41,7 @@ def clear_cache():
 
 async def ask_med_question(query, enable_stream):
     # Check the cache first
-    response = sm.lookup(question=query)
+    response = sm.lookup(question=query, cosine_threshold=0.9)
     if response:
         # Return cached response if found
         yield response

medirag/cache/local.py

@@ -1,75 +1,79 @@
 import faiss
 import json
 import numpy as np
+from pydantic import BaseModel, ValidationError
 from sentence_transformers import SentenceTransformer
+from loguru import logger
+
+
+class SemanticCache(BaseModel):
+    questions: list[str] = []
+    embeddings: list[list[float]] = []
+    response_text: list[str] = []
 
 
 class SemanticCaching:
     def __init__(self,
-                 model_name='sentence-transformers/all-mpnet-base-v2',
-                 dimension=768,
-                 json_file='cache.json',
-                 cosine_threshold=0.7):
+                 model_name: str = 'sentence-transformers/all-mpnet-base-v2',
+                 dimension: int = 768,
+                 json_file: str = 'cache.json'):
 
         self.model_name = model_name
         self.dimension = dimension
-        self.cosine_threshold = cosine_threshold
+        self.json_file = json_file
         self.vector_index = faiss.IndexFlatIP(self.dimension)
         self.encoder = SentenceTransformer(self.model_name)
-        self.json_file = json_file
-        self.cache = self.load_cache()
+        self.load_cache()
 
-    def load_cache(self):
+    def load_cache(self) -> None:
         """Load cache from a JSON file."""
-        local_cache = {'questions': [], 'embeddings': [], 'response_text': []}
         try:
-            if self.json_file:
-                with open(self.json_file, 'r') as file:
-                    local_cache = json.load(file)
-                    if 'embeddings' in local_cache and len(local_cache['embeddings']) > 0:
-                        for embedding in local_cache['embeddings']:
-                            _embedding = np.array(embedding, dtype=np.float32)
-                            self.vector_index.add(_embedding)
-                return local_cache
-            else:
-                return local_cache
+            with open(self.json_file, 'r') as file:
+                data = json.load(file)
+                data['embeddings'] = [np.array(e, dtype=np.float32) for e in data.get('embeddings', [])]
+                for emb in data['embeddings']:
+                    self.vector_index.add(emb)
+                self.cache = SemanticCache(**data)
         except FileNotFoundError:
-            return local_cache
+            logger.info("Cache file not found, initializing new cache.")
+        except ValidationError as e:
+            logger.error(f"Error in cache data structure: {e}")
         except Exception as e:
-            print(f"Failed to load or process cache: {e}")
-            return local_cache
+            logger.error(f"Failed to load or process cache: {e}")
+        self.cache = SemanticCache()
 
     def save_cache(self):
         """Save the current cache to a JSON file."""
+        data = self.cache.model_dump_json()
         with open(self.json_file, 'w') as file:
-            json.dump(self.cache, file)
+            json.dump(data, file)
+        logger.info("Cache saved successfully.")
 
-    def lookup(self, question: str) -> str | None:
+    def lookup(self, question: str, cosine_threshold: float = 0.7) -> str | None:
         """Check if a question is in the cache and return the cached response if it exists."""
         embedding = self.encoder.encode([question], show_progress_bar=False)
         faiss.normalize_L2(embedding)
-
-        # Search in the index
         D, I = self.vector_index.search(embedding, 1)
 
-        if D[0][0] >= self.cosine_threshold:
+        if D[0][0] >= cosine_threshold:
             row_id = I[0][0]
-            return self.cache['response_text'][row_id]
-        else:
-            return None
+            return self.cache.response_text[row_id]
+        return None
 
     def save(self, question: str, response: str):
         """Save a response to the cache."""
         embedding = self.encoder.encode([question], show_progress_bar=False)
         faiss.normalize_L2(embedding)
-
-        self.cache['questions'].append(question)
-        self.cache['embeddings'].append(embedding.tolist())
-        self.cache['response_text'].append(response)
+        self.cache.questions.append(question)
+        self.cache.embeddings.append(embedding.tolist())
+        self.cache.response_text.append(response)
         self.vector_index.add(embedding)
         self.save_cache()
+        logger.info("New response saved to cache.")
 
     def clear(self):
-        self.cache = {'questions': [], 'embeddings': [], 'response_text': []}
+        """Clear the cache."""
+        self.cache = SemanticCache()
         self.vector_index.reset()
         self.save_cache()
+        logger.info("Cache cleared.")

poetry.lock

@@ -1940,6 +1940,24 @@ llama-index-core = ">=0.11.0,<0.12.0"
 pandas = "*"
 pykx = ">=2.1.1,<3.0.0"
 
+[[package]]
+name = "loguru"
+version = "0.7.2"
+description = "Python logging made (stupidly) simple"
+optional = false
+python-versions = ">=3.5"
+files = [
+    {file = "loguru-0.7.2-py3-none-any.whl", hash = "sha256:003d71e3d3ed35f0f8984898359d65b79e5b21943f78af86aa5491210429b8eb"},
+    {file = "loguru-0.7.2.tar.gz", hash = "sha256:e671a53522515f34fd406340ee968cb9ecafbc4b36c679da03c18fd8d0bd51ac"},
+]
+
+[package.dependencies]
+colorama = {version = ">=0.3.4", markers = "sys_platform == \"win32\""}
+win32-setctime = {version = ">=1.0.0", markers = "sys_platform == \"win32\""}
+
+[package.extras]
+dev = ["Sphinx (==7.2.5)", "colorama (==0.4.5)", "colorama (==0.4.6)", "exceptiongroup (==1.1.3)", "freezegun (==1.1.0)", "freezegun (==1.2.2)", "mypy (==v0.910)", "mypy (==v0.971)", "mypy (==v1.4.1)", "mypy (==v1.5.1)", "pre-commit (==3.4.0)", "pytest (==6.1.2)", "pytest (==7.4.0)", "pytest-cov (==2.12.1)", "pytest-cov (==4.1.0)", "pytest-mypy-plugins (==1.9.3)", "pytest-mypy-plugins (==3.0.0)", "sphinx-autobuild (==2021.3.14)", "sphinx-rtd-theme (==1.3.0)", "tox (==3.27.1)", "tox (==4.11.0)"]
+
 [[package]]
 name = "lxml"
 version = "5.3.0"
@@ -5053,6 +5071,20 @@ files = [
     {file = "websockets-12.0.tar.gz", hash = "sha256:81df9cbcbb6c260de1e007e58c011bfebe2dafc8435107b0537f393dd38c8b1b"},
 ]
 
+[[package]]
+name = "win32-setctime"
+version = "1.1.0"
+description = "A small Python utility to set file creation time on Windows"
+optional = false
+python-versions = ">=3.5"
+files = [
+    {file = "win32_setctime-1.1.0-py3-none-any.whl", hash = "sha256:231db239e959c2fe7eb1d7dc129f11172354f98361c4fa2d6d2d7e278baa8aad"},
+    {file = "win32_setctime-1.1.0.tar.gz", hash = "sha256:15cf5750465118d6929ae4de4eb46e8edae9a5634350c01ba582df868e932cb2"},
+]
+
+[package.extras]
+dev = ["black (>=19.3b0)", "pytest (>=4.6.2)"]
+
 [[package]]
 name = "wrapt"
 version = "1.16.0"
@@ -5370,4 +5402,4 @@ multidict = ">=4.0"
 [metadata]
 lock-version = "2.0"
 python-versions = ">=3.10,<3.13"
-content-hash = "f03d8ab60928b71fb7bdb7dd2adfefc967249fdf18ff2660f9ba4182df1f0de4"
+content-hash = "e2a76129035f221f383481d3a02d57ba0a52337b56f684ce0b4e847e1262def5"

pyproject.toml

@@ -31,6 +31,7 @@ accelerate = ">=0.33.0"
 gradio = ">=4.42.0"
 pydantic = ">=2.8.2"
 kdbai-client = ">=1.2.4"
+loguru = "^0.7.2"
 
 
 

requirements.txt

@@ -11,7 +11,7 @@ async-timeout==4.0.3 ; python_version >= "3.10" and python_version < "3.11"
 attrs==24.2.0 ; python_version >= "3.10" and python_version < "3.13"
 backoff==2.2.1 ; python_version >= "3.10" and python_version < "3.13"
 beautifulsoup4==4.12.3 ; python_version >= "3.10" and python_version < "3.13"
-certifi==2024.7.4 ; python_version >= "3.10" and python_version < "3.13"
+certifi==2024.8.30 ; python_version >= "3.10" and python_version < "3.13"
 charset-normalizer==3.3.2 ; python_version >= "3.10" and python_version < "3.13"
 click==8.1.7 ; python_version >= "3.10" and python_version < "3.13"
 colorama==0.4.6 ; python_version >= "3.10" and python_version < "3.13" and (platform_system == "Windows" or sys_platform == "win32")
@@ -46,7 +46,7 @@ huggingface-hub==0.24.6 ; python_version >= "3.10" and python_version < "3.13"
 huggingface-hub[inference]==0.24.6 ; python_version >= "3.10" and python_version < "3.13"
 idna==3.8 ; python_version >= "3.10" and python_version < "3.13"
 importlib-resources==6.4.4 ; python_version >= "3.10" and python_version < "3.13"
-ipython==8.26.0 ; python_version >= "3.10" and python_version < "3.13"
+ipython==8.27.0 ; python_version >= "3.10" and python_version < "3.13"
 jedi==0.19.1 ; python_version >= "3.10" and python_version < "3.13"
 jinja2==3.1.4 ; python_version >= "3.10" and python_version < "3.13"
 jiter==0.5.0 ; python_version >= "3.10" and python_version < "3.13"
@@ -59,9 +59,9 @@ kiwisolver==1.4.5 ; python_version >= "3.10" and python_version < "3.13"
 langchain-core==0.2.36 ; python_version >= "3.10" and python_version < "3.13"
 langchain-text-splitters==0.2.2 ; python_version >= "3.10" and python_version < "3.13"
 langchain==0.2.15 ; python_version >= "3.10" and python_version < "3.13"
-langsmith==0.1.106 ; python_version >= "3.10" and python_version < "3.13"
+langsmith==0.1.107 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-agent-openai==0.3.0 ; python_version >= "3.10" and python_version < "3.13"
-llama-index-core==0.11.2 ; python_version >= "3.10" and python_version < "3.13"
+llama-index-core==0.11.3 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-embeddings-huggingface==0.3.1 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-embeddings-openai==0.2.3 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-llms-openai==0.2.0 ; python_version >= "3.10" and python_version < "3.13"
@@ -69,6 +69,7 @@ llama-index-readers-file==0.2.0 ; python_version >= "3.10" and python_version <
 llama-index-utils-workflow==0.2.0 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-vector-stores-faiss==0.2.1 ; python_version >= "3.10" and python_version < "3.13"
 llama-index-vector-stores-kdbai==0.3.1 ; python_version >= "3.10" and python_version < "3.13"
+loguru==0.7.2 ; python_version >= "3.10" and python_version < "3.13"
 lxml==5.3.0 ; python_version >= "3.10" and python_version < "3.13"
 mako==1.3.5 ; python_version >= "3.10" and python_version < "3.13"
 markdown-it-py==3.0.0 ; python_version >= "3.10" and python_version < "3.13" and sys_platform != "emscripten"
@@ -164,6 +165,7 @@ urllib3==2.2.2 ; python_version >= "3.10" and python_version < "3.13"
 uvicorn==0.30.6 ; python_version >= "3.10" and python_version < "3.13" and sys_platform != "emscripten"
 wcwidth==0.2.13 ; python_version >= "3.10" and python_version < "3.13"
 websockets==12.0 ; python_version >= "3.10" and python_version < "3.13"
+win32-setctime==1.1.0 ; python_version >= "3.10" and python_version < "3.13" and sys_platform == "win32"
 wrapt==1.16.0 ; python_version >= "3.10" and python_version < "3.13"
 xxhash==3.5.0 ; python_version >= "3.10" and python_version < "3.13"
 yarl==1.9.4 ; python_version >= "3.10" and python_version < "3.13"

tests/rag/test_rag.py

@@ -9,7 +9,7 @@ load_dotenv()  # take environment variables from .env.
 
 
 def ask_med_question(sm, rag, query):
-    response = sm.lookup(question=query)
+    response = sm.lookup(question=query, cosine_threshold=0.9)
     if not response:
         response = rag(query).answer
         sm.save(query, response)