ai_chatbot 0.1.6.4 → 0.1.6.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 802b2da58aa8ef78b7b9813faf48e789db953d4c6e62a437002b1dc6017e4167
-  data.tar.gz: b3525c33bd0965aeccc2b13d67380e67df0ec6f3c502b0afe24968d694833d0f
+  metadata.gz: 65a8e9d42a6f0a36cff9a13f7428975232912b4a2decd9509772b38d849f6340
+  data.tar.gz: f54c3642f29fb1f12f49af978f4ae34bd9dcdb85c7d22b48d07b195384354832
 SHA512:
-  metadata.gz: eff5b8d9885206a83a4f6a05ca275c925717299360ae62df2af451009a07e5d244de503031d0f8a091a4cd1ef067af85a18cadf6ded9ef994899b6d5639e4747
-  data.tar.gz: b05cd753b9e3d4db2d9e6fb86aa43b2e0e355d8cf723e3d852bfb43a23897aade4d5849232179c7338babc0f19568d05d5c5eafdda781337d00daca83b6594fc
+  metadata.gz: 6a7d4c388b24487a16a51dac935eb54d6dfc384f8d478dd80667f5de9a3dc2d8140e2785a0721db5d93f79b29285ad5deb3462042238536925c5b8e66b481939
+  data.tar.gz: e691cb235c3cae3c9b754c30a20cec05ca17bc684e0e7c727528ccf0b8f06e4a100e4473d050793a62029bf3352425f1b1c00540e709f04f868366bb2a0f5559

data/lib/ai_chatbot/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module AiChatbot
-  VERSION = "0.1.6.4"
+  VERSION = "0.1.6.5"
 end

data/lib/ml_model.py

@@ -1,162 +1,119 @@
 import sys
+import psycopg2
+import os
+from dotenv import load_dotenv
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.naive_bayes import MultinomialNB
 from sklearn.pipeline import make_pipeline
 from sklearn.metrics.pairwise import cosine_similarity
-import pickle
-import os
 
-# Cache to store previously asked questions
-cache = {}
-
-# Load or initialize the dataset
-if os.path.exists("qa_model.pkl"):
-    with open("qa_model.pkl", "rb") as f:
-        model_data = pickle.load(f)
-        questions = model_data.get("questions", [])
-        answers = model_data.get("answers", [])
-else:
-    questions = [
-        "How to create a new model in Rails?",
-        "What is migration?",
-        "How to add a route?"
-    ]
-    answers = [
-        "You can create a model using 'rails generate model'.",
-        "Migration is a database schema change.",
-        "You can add a route in the config/routes.rb file."
-    ]
-
-# Create a pipeline (TF-IDF + MultinomialNB)
-model = make_pipeline(TfidfVectorizer(), MultinomialNB())
-model.fit(questions, answers)
-
-# Populate cache with known questions and answers
-for q, a in zip(questions, answers):
-    cache[q.lower()] = a
-
-# Function to predict or retrain the model
-def main(action, query=None, new_answer=None):
-    if action == "predict":
-        return get_prediction(query)
-    elif action == "train_model":
-        return train_model(query, new_answer)
-    elif action == "update_answer":
-        return update_answer(query, new_answer)
-    elif action == "update_or_delete_question":
-        return update_or_delete_question(query, new_answer)
-    elif action == "list_questions":
-        return list_questions()
-    elif action == "list_answers":
-        return list_answers()
+# Connect to PostgreSQL
+conn = psycopg2.connect(
+    dbname= os.getenv("DB_NAME"),
+    user=os.getenv("DB_USERNAME"),
+    password=os.getenv("DB_PASSWORD"),
+    host=os.getenv("DB_HOST"), 
+    port=os.getenv("DB_PORT"),
+)
+cursor = conn.cursor()
 
-# Function to predict the response with caching
-def get_prediction(query):
-    query_lower = query.lower()
+db_name = os.getenv("DB_NAME")
+
+cursor.execute("SELECT question, answer FROM qa_data")
+rows = cursor.fetchall()
+questions = [row[0] for row in rows]
+answers = [row[1] for row in rows]
+
+vectorizer = TfidfVectorizer()
+question_vecs = vectorizer.fit_transform(questions) if questions else None
 
-    # **Check cache first**
-    if query_lower in cache:
-        return cache[query_lower]
 
-    query_vec = model.named_steps['tfidfvectorizer'].transform([query])
-    question_vecs = model.named_steps['tfidfvectorizer'].transform(questions)
+def get_prediction(query):
+    if not questions:
+        return "No questions available in the database."
+    
+    query_vec = vectorizer.transform([query])
+    similarities = cosine_similarity(query_vec, question_vecs).flatten()
 
-    # Calculate cosine similarity
-    similarities = cosine_similarity(query_vec, question_vecs)
-    max_similarity = similarities.max()
+    max_sim_index = similarities.argmax()
+    max_similarity = similarities[max_sim_index]
 
     threshold = 0.65
     if max_similarity < threshold:
         return "No good match found. Please provide the correct answer."
     else:
-        prediction = model.predict([query])[0]
-        
-        # **Store in cache for faster future retrieval**
-        cache[query_lower] = prediction
-        
-        return prediction
-
-# Function to train the model with a new question and answer
+        return answers[max_sim_index]
+
+# Function to train the model with new data
 def train_model(new_question, new_answer):
     global questions, answers
 
-    # Append new question-answer pair
+    # Store in database
+    
+    cursor.execute("INSERT INTO qa_data (question, answer, created_at,updated_at) VALUES (%s, %s, NOW(),NOW()) ON CONFLICT (question) DO NOTHING",
+    (new_question, new_answer))
+    conn.commit()
+
+    # Update lists and retrain model
     questions.append(new_question)
     answers.append(new_answer)
-
-    # Retrain the model
     model.fit(questions, answers)
 
-    # **Update cache**
-    cache[new_question.lower()] = new_answer
-
-    # Save the updated model
-    with open("qa_model.pkl", "wb") as f:
-        pickle.dump({"questions": questions, "answers": answers}, f)
-
-    return f"Model retrained with: '{new_question}' -> '{new_answer}'"
+    return f"Added: '{new_question}' -> '{new_answer}'"
 
 # Function to update an answer
 def update_answer(existing_question, new_answer):
-    global questions, answers
+    cursor.execute("UPDATE qa_data SET answer = %s WHERE question = %s", (new_answer, existing_question))
+    conn.commit()
 
-    if existing_question in questions:
-        index = questions.index(existing_question)
-        answers[index] = new_answer
-
-        # Retrain the model
-        model.fit(questions, answers)
-
-        # **Update cache**
-        cache[existing_question.lower()] = new_answer
+    # Update lists and retrain model
+    index = questions.index(existing_question)
+    answers[index] = new_answer
+    model.fit(questions, answers)
 
-        # Save the model
-        with open("qa_model.pkl", "wb") as f:
-            pickle.dump({"questions": questions, "answers": answers}, f)
-        
-        return f"Answer updated for: '{existing_question}'"
-    
-    return "Question not found."
+    return f"Updated: '{existing_question}' -> '{new_answer}'"
 
-# Function to update or delete a question
-def update_or_delete_question(existing_question, new_question):
-    global questions, answers
+# Function to delete a question
+def delete_question(existing_question):
+    cursor.execute("DELETE FROM qa_data WHERE question = %s", (existing_question,))
+    conn.commit()
 
     if existing_question in questions:
         index = questions.index(existing_question)
-
-        if new_question:
-            questions[index] = new_question
-            # **Update cache**
-            cache[new_question.lower()] = answers[index]
-        else:
-            # Delete the question
-            del questions[index]
-            del answers[index]
-
-        # Retrain the model
+        del questions[index]
+        del answers[index]
         model.fit(questions, answers)
 
-        # **Remove from cache if deleted**
-        if not new_question:
-            cache.pop(existing_question.lower(), None)
-
-        # Save the model
-        with open("qa_model.pkl", "wb") as f:
-            pickle.dump({"questions": questions, "answers": answers}, f)
-
-        return f"Updated question: '{existing_question}' -> '{new_question}'" if new_question else f"Deleted: '{existing_question}'"
-
-    return "Question not found."
+    return f"Deleted: '{existing_question}'"
 
+# Function to list questions
 def list_questions():
-    return questions
+    cursor.execute("SELECT question FROM qa_data")
+    return [row[0] for row in cursor.fetchall()]
 
+# Function to list answers
 def list_answers():
-    return answers
+    cursor.execute("SELECT answer FROM qa_data")
+    return [row[0] for row in cursor.fetchall()]
 
 if __name__ == "__main__":
     action = sys.argv[1]
     question = sys.argv[2] if len(sys.argv) > 2 else None
     answer = sys.argv[3] if len(sys.argv) > 3 else None
-    print(main(action, question, answer))
+
+    if action == "predict":
+        print(get_prediction(question))
+    elif action == "train_model":
+        print(train_model(question, answer))
+    elif action == "update_answer":
+        print(update_answer(question, answer))
+    elif action == "delete_question":
+        print(delete_question(question))
+    elif action == "list_questions":
+        print(list_questions())
+    elif action == "list_answers":
+        print(list_answers())
+
+# Close DB connection
+cursor.close()
+conn.close()

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ai_chatbot
 version: !ruby/object:Gem::Version
-  version: 0.1.6.4
+  version: 0.1.6.5
 platform: ruby
 authors:
 - Sanket
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2025-02-26 00:00:00.000000000 Z
+date: 2025-03-04 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: open3
@@ -24,8 +24,8 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
-description: Integrates a chatbot using Python for predictions and training in a Rails
-  application.
+description: Added caching to avoid redundant DB queries, improving chatbot response
+  time. Also updated unanswered question logging.
 email:
 - sanket.tikhande@gmail.com
 executables: []
@@ -58,5 +58,5 @@ requirements: []
 rubygems_version: 3.3.7
 signing_key: 
 specification_version: 4
-summary: 'Fix: Improved response caching in ChatbotService'
+summary: 'Fix: Added postgres integration'
 test_files: []