{"id":2044,"date":"2025-02-25T07:03:32","date_gmt":"2025-02-25T07:03:32","guid":{"rendered":"https:\/\/mailitics.com\/index.php\/2025\/02\/25\/enhancing-rag-beyond-vanilla-approaches\/"},"modified":"2025-02-25T07:03:32","modified_gmt":"2025-02-25T07:03:32","slug":"enhancing-rag-beyond-vanilla-approaches","status":"publish","type":"post","link":"https:\/\/mailitics.com\/index.php\/2025\/02\/25\/enhancing-rag-beyond-vanilla-approaches\/","title":{"rendered":"Enhancing RAG: Beyond Vanilla Approaches"},"content":{"rendered":"

Enhancing RAG: Beyond Vanilla Approaches
\n \t

\n
<\/BR>
\n
\n \t

\n
<\/BR><\/p>\n

\n

Retrieval-Augmented Generation (RAG) is a powerful technique that enhances language models by incorporating external information retrieval mechanisms. While standard RAG implementations improve response relevance, they often struggle in complex retrieval scenarios. This article explores the limitations of a vanilla RAG setup and introduces advanced techniques to enhance its accuracy and efficiency.<\/p>\n

The Challenge with Vanilla RAG<\/strong><\/h3>\n

To illustrate RAG\u2019s limitations, consider a simple experiment where we attempt to retrieve relevant information from a set of documents. Our dataset includes:<\/p>\n

    \n
  • A primary document discussing best practices for staying healthy, productive, and in good shape.<\/li>\n
  • Two additional documents on unrelated topics, but contain some similar words used in different contexts.<\/li>\n<\/ul>\n
    main_document_text = \"\"\"\nMorning Routine (5:30 AM - 9:00 AM)\n\"\u2705\" Wake Up Early - Aim for 6-8 hours of sleep to feel well-rested.\n\"\u2705\" Hydrate First - Drink a glass of water to rehydrate your body.\n\"\u2705\" Morning Stretch or Light Exercise - Do 5-10 minutes of stretching or a short workout to activate your body.\n\"\u2705\" Mindfulness or Meditation - Spend 5-10 minutes practicing mindfulness or deep breathing.\n\"\u2705\" Healthy Breakfast - Eat a balanced meal with protein, healthy fats, and fiber.\n\"\u2705\" Plan Your Day - Set goals, review your schedule, and prioritize tasks.\n...\n\"\"\"<\/code><\/pre>\n
    Using a standard RAG setup, we query the system with:<\/p>\n
      \n
    1. What should I do to stay healthy and productive?<\/em><\/li>\n
    2. What are the best practices to stay healthy and productive?<\/em><\/li>\n<\/ol>\n
      Helper Functions<\/strong><\/h3>\n
      To enhance retrieval accuracy and streamline query processing, we implement a set of essential helper functions. These functions serve various purposes, from querying the ChatGPT API to computing document embeddings and similarity scores. By leveraging these functions, we create a more efficient RAG pipeline that effectively retrieves the most relevant information for user queries.<\/p>\n
      To support our RAG improvements, we define the following helper functions:<\/p>\n
      # **Imports**\nimport os\nimport json\nimport openai\nimport numpy as np\nfrom scipy.spatial.distance import cosine\nfrom google.colab import userdata\n\n# Set up OpenAI API key\nos.environ[\"OPENAI_API_KEY\"] = userdata.get('AiTeam')<\/code><\/pre>\n
      def query_chatgpt(prompt, model=\"gpt-4o\", response_format=openai.NOT_GIVEN):\n\u00a0 \u00a0 try:\n\u00a0 \u00a0 \u00a0 \u00a0 response = client.chat.completions.create(\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 model=model,\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 messages=[{\"role\": \"user\", \"content\": prompt}],\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 temperature=0.0 , # Adjust for more or less creativity\n\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 response_format=response_format\n\u00a0 \u00a0 \u00a0 \u00a0 )\n\u00a0 \u00a0 \u00a0 \u00a0 return response.choices[0].message.content.strip()\n\u00a0 \u00a0 except Exception as e:\n\u00a0 \u00a0 \u00a0 \u00a0 return f\"Error: {e}\"<\/code><\/pre>\n
      def get_embedding(text, model=\"text-embedding-3-large\"): #\"text-embedding-ada-002\"\n    \"\"\"Fetches the embedding for a given text using OpenAI's API.\"\"\"\n    response = client.embeddings.create(\n        input=[text],\n        model=model\n    )\n    return response.data[0].embedding<\/code><\/pre>\n
      def compute_similarity_metrics(embed1, embed2):\n    \"\"\"Computes different similarity\/distance metrics between two embeddings.\"\"\"\n    cosine_sim = 1- cosine(embed1, embed2)  # Cosine similarity\n\n    return cosine_sim<\/code><\/pre>\n
      def fetch_similar_docs(query, docs, threshold = .55, top=1):\n  query_em = get_embedding(query)\n  data = []\n  for d in docs:\n    # Compute and print similarity metrics\n    similarity_results = compute_similarity_metrics(d[\"embedding\"], query_em)\n    if(similarity_results >= threshold):\n      data.append({\"id\":d[\"id\"], \"ref_doc\":d.get(\"ref_doc\", \"\"), \"score\":similarity_results})\n\n  # Sorting by value (second element in each tuple)\n  sorted_data = sorted(data, key=lambda x: x[\"score\"], reverse=True)  # Ascending order\n  sorted_data = sorted_data[:min(top, len(sorted_data))]\n  return sorted_data<\/code><\/pre>\n
      Evaluating the Vanilla RAG<\/strong><\/h2>\n
      To evaluate the effectiveness of a vanilla RAG setup, we conduct a simple test using predefined queries. Our goal is to determine whether the system retrieves the most relevant document based on semantic similarity. We then analyze the limitations and explore possible improvements.<\/p>\n
      \"\"\"# **Testing Vanilla RAG**\"\"\"\n\nquery = \"what should I do to stay healthy and productive?\"\nr = fetch_similar_docs(query, docs)\nprint(\"query = \", query)\nprint(\"documents = \", r)\n\nquery = \"what are the best practices to stay healthy and productive ?\"\nr = fetch_similar_docs(query, docs)\nprint(\"query = \", query)\nprint(\"documents = \", r)<\/code><\/pre>\n
      Advanced Techniques for Improved RAG<\/strong><\/h2>\n
      To further refine the retrieval process, we introduce advanced functions that enhance the capabilities of our RAG system. These functions generate structured information that aids in document retrieval and query processing, making our system more robust and context-aware.<\/p>\n
      To address these challenges, we implement three key enhancements:<\/p>\n
      1. Generating FAQs<\/strong><\/h4>\n
      By automatically creating a list of frequently asked questions related to a document, we expand the range of potential queries the model can match. These FAQs are generated once and stored alongside the document, providing a richer search space without incurring ongoing costs.<\/p>\n
      def generate_faq(text):\n  prompt = f'''\n  given the following text: \"\"\"{text}\"\"\"\n  Ask relevant simple atomic questions ONLY (don't answer them) to cover all subjects covered by the text. Return the result as a json list example [q1, q2, q3...]\n  '''\n  return query_chatgpt(prompt, response_format={ \"type\": \"json_object\" })<\/code><\/pre>\n
      2. Creating an Overview<\/strong><\/h4>\n
      A high-level summary of the document helps capture its core ideas, making retrieval more effective. By embedding the overview alongside the document, we provide additional entry points for relevant queries, improving match rates.<\/p>\n
      def generate_overview(text):\n\u00a0 prompt = f'''\n\u00a0 given the following text: \"\"\"{text}\"\"\"\n\u00a0 Generate an abstract for it that tells in maximum 3 lines what is it about and use high level terms that will capture the main points,\n\u00a0 Use terms and words that will be most likely used by average person.\n\u00a0 '''\n\u00a0 return query_chatgpt(prompt)<\/code><\/pre>\n
      3. Query Decomposition<\/strong><\/h4>\n
      Instead of searching with broad user queries, we break them down into smaller, more precise sub-queries. Each sub-query is then compared against our enhanced document collection, which now includes:<\/p>\n