Supply/Test code/Test code_GPT.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "40ad5822-30bb-4949-99e3-bf834919cb2b",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing row 2...\n",
      "Saved data for row 2\n",
      "Processing row 3...\n",
      "Saved data for row 3\n",
      "Processing row 4...\n",
      "Saved data for row 4\n",
      "Processing row 5...\n",
      "Saved data for row 5\n",
      "Processing row 6...\n",
      "Saved data for row 6\n",
      "Processing row 7...\n",
      "Saved data for row 7\n",
      "Processing row 8...\n",
      "Saved data for row 8\n",
      "Processing row 9...\n",
      "Saved data for row 9\n",
      "Processing row 10...\n",
      "Saved data for row 10\n",
      "Processing row 11...\n",
      "Saved data for row 11\n",
      "Processing row 12...\n",
      "Saved data for row 12\n",
      "Processing row 13...\n",
      "Saved data for row 13\n",
      "Processing row 14...\n",
      "Saved data for row 14\n",
      "Processing row 15...\n",
      "Saved data for row 15\n",
      "Processing row 16...\n",
      "Saved data for row 16\n",
      "Processing row 17...\n",
      "Saved data for row 17\n",
      "Processing row 18...\n",
      "Saved data for row 18\n",
      "Processing row 19...\n",
      "Saved data for row 19\n",
      "Processing row 20...\n",
      "Saved data for row 20\n",
      "Processing row 21...\n",
      "Saved data for row 21\n",
      "Processing row 22...\n",
      "Saved data for row 22\n",
      "Processing row 23...\n",
      "Saved data for row 23\n",
      "Processing row 24...\n",
      "Saved data for row 24\n",
      "Processing row 25...\n",
      "Saved data for row 25\n",
      "Processing row 26...\n",
      "Saved data for row 26\n",
      "Processing row 27...\n",
      "Saved data for row 27\n",
      "Processing row 28...\n",
      "Saved data for row 28\n",
      "Processing row 29...\n",
      "Saved data for row 29\n",
      "Processing row 30...\n",
      "Saved data for row 30\n",
      "Processing row 31...\n",
      "Saved data for row 31\n",
      "Processing row 32...\n",
      "Saved data for row 32\n",
      "Processing row 33...\n",
      "Saved data for row 33\n",
      "Processing row 34...\n",
      "Saved data for row 34\n",
      "Processing row 35...\n",
      "Saved data for row 35\n",
      "Processing row 36...\n",
      "Saved data for row 36\n",
      "Processing row 37...\n",
      "Saved data for row 37\n",
      "Processing row 38...\n",
      "Saved data for row 38\n",
      "Processing row 39...\n",
      "Saved data for row 39\n",
      "Processing row 40...\n",
      "Saved data for row 40\n",
      "Processing row 41...\n",
      "Saved data for row 41\n",
      "Processing row 42...\n",
      "Saved data for row 42\n",
      "Processing row 43...\n",
      "Saved data for row 43\n",
      "Processing row 44...\n",
      "Saved data for row 44\n",
      "Processing row 45...\n",
      "Saved data for row 45\n",
      "Processing row 46...\n",
      "Saved data for row 46\n",
      "Processing row 47...\n",
      "Saved data for row 47\n",
      "Processing row 48...\n",
      "Saved data for row 48\n",
      "Processing row 49...\n",
      "Saved data for row 49\n",
      "Processing row 50...\n",
      "Saved data for row 50\n",
      "Processing row 51...\n",
      "Saved data for row 51\n",
      "Processing row 52...\n",
      "Saved data for row 52\n",
      "Processing row 53...\n",
      "Saved data for row 53\n",
      "Processing row 54...\n",
      "Saved data for row 54\n",
      "Processing row 55...\n",
      "Saved data for row 55\n",
      "Processing row 56...\n",
      "Saved data for row 56\n",
      "Processing row 57...\n",
      "Saved data for row 57\n",
      "Processing row 58...\n",
      "Saved data for row 58\n",
      "Processing row 59...\n",
      "Saved data for row 59\n",
      "Processing row 60...\n",
      "Saved data for row 60\n",
      "Processing row 61...\n",
      "Saved data for row 61\n",
      "Processing row 62...\n",
      "Saved data for row 62\n",
      "Processing row 63...\n",
      "Saved data for row 63\n",
      "Processing row 64...\n",
      "Saved data for row 64\n",
      "Processing row 65...\n",
      "Saved data for row 65\n",
      "Processing row 66...\n",
      "Saved data for row 66\n",
      "Processing row 67...\n",
      "Saved data for row 67\n",
      "Processing row 68...\n",
      "Saved data for row 68\n",
      "Processing row 69...\n",
      "Saved data for row 69\n",
      "Processing row 70...\n",
      "Saved data for row 70\n",
      "Processing row 71...\n",
      "Saved data for row 71\n",
      "Processing row 72...\n",
      "Saved data for row 72\n",
      "Processing row 73...\n",
      "Saved data for row 73\n",
      "Processing row 74...\n",
      "Saved data for row 74\n",
      "Processing row 75...\n",
      "Saved data for row 75\n",
      "Processing row 76...\n",
      "Saved data for row 76\n",
      "Processing row 77...\n",
      "Saved data for row 77\n",
      "Processing row 78...\n",
      "Saved data for row 78\n",
      "Processing row 79...\n",
      "Saved data for row 79\n",
      "Processing row 80...\n",
      "Saved data for row 80\n",
      "Processing row 81...\n",
      "Saved data for row 81\n",
      "Processing row 82...\n",
      "Saved data for row 82\n",
      "Processing row 83...\n",
      "Saved data for row 83\n",
      "Processing row 84...\n",
      "Saved data for row 84\n",
      "Processing row 85...\n",
      "Saved data for row 85\n",
      "Processing row 86...\n",
      "Saved data for row 86\n",
      "Processing row 87...\n",
      "Saved data for row 87\n",
      "Processing row 88...\n",
      "Saved data for row 88\n",
      "Processing row 89...\n",
      "Saved data for row 89\n",
      "Processing row 90...\n",
      "Saved data for row 90\n",
      "Processing row 91...\n",
      "Saved data for row 91\n",
      "Processing row 92...\n",
      "Saved data for row 92\n",
      "Processing row 93...\n",
      "Saved data for row 93\n",
      "Processing row 94...\n",
      "Saved data for row 94\n",
      "Processing row 95...\n",
      "Saved data for row 95\n",
      "Processing row 96...\n",
      "Saved data for row 96\n",
      "Processing row 97...\n",
      "Saved data for row 97\n",
      "Processing row 98...\n",
      "Saved data for row 98\n",
      "Processing row 99...\n",
      "Saved data for row 99\n",
      "Processing row 100...\n",
      "Saved data for row 100\n",
      "Processing row 101...\n",
      "Saved data for row 101\n",
      "Processing row 102...\n",
      "Saved data for row 102\n",
      "Processing row 103...\n",
      "Saved data for row 103\n",
      "Processing row 104...\n",
      "Saved data for row 104\n",
      "Processing row 105...\n",
      "Saved data for row 105\n",
      "Processing row 106...\n",
      "Saved data for row 106\n",
      "Processing row 107...\n",
      "Saved data for row 107\n",
      "Processing row 108...\n",
      "Saved data for row 108\n",
      "Processing row 109...\n",
      "Saved data for row 109\n",
      "Processing row 110...\n",
      "Saved data for row 110\n",
      "Processing row 111...\n",
      "Saved data for row 111\n",
      "Processing row 112...\n",
      "Saved data for row 112\n",
      "Processing row 113...\n",
      "Saved data for row 113\n",
      "Processing row 114...\n",
      "Saved data for row 114\n",
      "Processing row 115...\n",
      "Saved data for row 115\n",
      "Processing row 116...\n",
      "Saved data for row 116\n",
      "Processing row 117...\n",
      "Saved data for row 117\n",
      "Processing row 118...\n",
      "Saved data for row 118\n",
      "Processing row 119...\n",
      "Saved data for row 119\n",
      "Processing row 120...\n",
      "Saved data for row 120\n",
      "Processing row 121...\n",
      "Saved data for row 121\n",
      "Processing row 122...\n",
      "Saved data for row 122\n",
      "Processing row 123...\n",
      "Saved data for row 123\n",
      "Processing row 124...\n",
      "Saved data for row 124\n",
      "Processing row 125...\n",
      "Saved data for row 125\n",
      "Processing row 126...\n",
      "Saved data for row 126\n",
      "Processing row 127...\n",
      "Saved data for row 127\n",
      "Processing row 128...\n",
      "Saved data for row 128\n",
      "Processing row 129...\n",
      "Saved data for row 129\n",
      "Processing row 130...\n",
      "Saved data for row 130\n",
      "Processing row 131...\n",
      "Saved data for row 131\n",
      "Processing row 132...\n",
      "Saved data for row 132\n",
      "Processing row 133...\n",
      "Saved data for row 133\n",
      "Processing row 134...\n",
      "Saved data for row 134\n",
      "Processing row 135...\n",
      "Saved data for row 135\n",
      "Processing row 136...\n",
      "Saved data for row 136\n",
      "Processing row 137...\n",
      "Saved data for row 137\n",
      "Processing row 138...\n",
      "Saved data for row 138\n",
      "Processing row 139...\n",
      "Saved data for row 139\n",
      "Processing row 140...\n",
      "Saved data for row 140\n",
      "Processing row 141...\n",
      "Saved data for row 141\n",
      "Processing row 142...\n",
      "Saved data for row 142\n",
      "Processing row 143...\n",
      "Saved data for row 143\n",
      "Processing row 144...\n",
      "Saved data for row 144\n",
      "Processing row 145...\n",
      "Saved data for row 145\n",
      "Processing row 146...\n",
      "Saved data for row 146\n",
      "The task has been successfully completed.\n"
     ]
    }
   ],
   "source": [
    "import openpyxl\n",
    "from openai import OpenAI\n",
    "\n",
    "client = OpenAI(api_key=\"\")\n",
    "import os\n",
    "import time\n",
    "import json\n",
    "\n",
    "def get_completion(prompt, model=\"gpt-4-1106-preview\", retries=3):\n",
    "    for _ in range(retries):\n",
    "        try:\n",
    "            messages = [{\"role\": \"system\", \"content\": \"You are designed to output the final answer in JSON.\"},\n",
    "                        {\"role\": \"user\", \"content\": prompt}\n",
    "                       ]\n",
    "            response = client.chat.completions.create(\n",
    "                model=model,\n",
    "                messages=messages,\n",
    "                temperature=0.7,\n",
    "                response_format={ \"type\": \"json_object\" }\n",
    "            )\n",
    "            return response.choices[0].message.content\n",
    "        except Exception as e:\n",
    "            print(f\"An error occurred: {e}. Retrying...\")\n",
    "            time.sleep(2)\n",
    "    return None\n",
    "\n",
    "wb_path = 'C:/Work/Articles/Chat-GPT/specializeddoc/Osteoarthritis/Part1.Base model evaluation/4/treatment_based_4.xlsx'\n",
    "wb = openpyxl.load_workbook(wb_path)\n",
    "sheet = wb['Sheet1']\n",
    "\n",
    "\n",
    "try:\n",
    "    with open(\"checkpoint.json\", \"r\") as f:\n",
    "        checkpoint = json.load(f)\n",
    "    start_row = checkpoint[\"row\"]\n",
    "    print(f\"Resuming from row {start_row}...\")\n",
    "except FileNotFoundError:\n",
    "    start_row = 2\n",
    "\n",
    "prompts = [\n",
    "     \"\"\"\n",
    "       Provide answer for the following question, the answer should be based on well acknowledged published clinical guidelines: \n",
    "\n",
    "       The answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). \n",
    "       The answer should provide brief explanation and additional information worth mentioning.\n",
    "       \n",
    "       For example: 'Treatment X is [recommended/not recommended/may be used] for condition Y. The strength of recommendation is [recommendation strength], and the certainty of evidence is [evidence level].\n",
    "                    Brief explanation and additional information is [Brief explanation and additional information]\n",
    "       \n",
    "       \n",
    "       Note: Output in the example json format. \n",
    "       Example in plain words: AA for osteoarthritis. Recommendation status: EE The strength of recommendation is BB The certainty of evidence is CC. According to DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       Example in json format：Intervention: AA ;RecommendationStatus: EE;StrengthOfRecommendation: BB;LevelOfCertainty: CC；Source: DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       \n",
    "       The question is: \n",
    "\n",
    "\"\"\",\n",
    "    \"\"\"You are a digital osteoarthritis specialists, leveraging its professional knowledge to provide insights on osteoarthritis based on clinical guidelines. You avoids speculation and sticks to facts. If a query cannot be answered, you will state not known.\n",
    "       \n",
    "       Provide answer for the following question, the answer should be based on well acknowledged published clinical guidelines:\n",
    "       \n",
    "       The answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). \n",
    "       The answer should provide brief explanation and additional information worth mentioning.\n",
    "       \n",
    "       Note: Output in the example json format. \n",
    "       Example in plain words: AA for osteoarthritis. Recommendation status: EE The strength of recommendation is BB The certainty of evidence is CC. According to DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       Example in json format：Intervention: AA ;RecommendationStatus: EE;StrengthOfRecommendation: BB;LevelOfCertainty: CC；Source: DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       Perform the task step by step.\n",
    "       \n",
    "\n",
    "\"\"\"\n",
    "]\n",
    "\n",
    "stop_script = False  # Flag to indicate if we should stop the script\n",
    "\n",
    "for content_row in range(start_row, len(sheet['F']) + 1):\n",
    "    if stop_script:\n",
    "        break\n",
    "\n",
    "    col_index = 7\n",
    "    content = sheet.cell(row=content_row, column=6).value\n",
    "    print(f\"Processing row {content_row}...\")\n",
    "    \n",
    "    for prompt in prompts:\n",
    "        for run in range(5):\n",
    "            full_prompt = f\"{prompt} {content}\"\n",
    "            response = get_completion(full_prompt)\n",
    "            \n",
    "            \n",
    "            if response is None:\n",
    "                with open(\"checkpoint.json\", \"w\") as f:\n",
    "                    json.dump({\"row\": content_row}, f)\n",
    "                print(f\"Stopping at row {content_row}. Checkpoint saved.\")\n",
    "                stop_script = True  # Set the flag to stop the script\n",
    "                break\n",
    "\n",
    "            if sheet.cell(row=content_row, column=col_index).value is None:  # Do not overwrite existing content\n",
    "                sheet.cell(row=content_row, column=col_index).value = response.strip()\n",
    "            \n",
    "            col_index += 1\n",
    "\n",
    "    wb.save('C:/Work/Articles/Chat-GPT/specializeddoc/Osteoarthritis/Part1.Base model evaluation/4/treatment_based_4.xlsx')\n",
    "    print(f\"Saved data for row {content_row}\")\n",
    "\n",
    "    if os.path.exists(\"checkpoint.json\"):\n",
    "        os.remove(\"checkpoint.json\")\n",
    "\n",
    "wb.save('C:/Work/Articles/Chat-GPT/specializeddoc/Osteoarthritis/Part1.Base model evaluation/4/treatment_based_4.xlsx')\n",
    "print(\"The task has been successfully completed.\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "10a49b15-c682-4682-9129-57f98edc5c58",
   "metadata": {},
   "outputs": [],
   "source": [
    "\",\n",
    "    \"\"\" \n",
    "    You are 3 digital osteoarthritis specialists,  leveraging professional knowledge to provide insights on osteoarthritis based on clinical guidelines. You avoids speculation and sticks to facts. If a query cannot be answered, you will state not known.\n",
    " \n",
    "    Step 1. Make sure each expert understand the meaning of the question.\n",
    "\n",
    "    Step 2. Each expert will independently give out an answer to the question, the answer should be based on well acknowledged published clinical guidelines: The answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). For example: 'Treatment X is [recommended/not recommended/may be used] for condition Y. The strength of recommendation is [recommendation strength], and the certainty of evidence is [evidence level].\n",
    "   \n",
    "    Step 3. Three experts will discuss the answers provided in step 3, and choose the best answer.\n",
    "\n",
    "    Step 4. After discussing, the final answer will be provided.\n",
    "    \n",
    "       The final answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). \n",
    "       The final answer should provide brief explanation and additional information worth mentioning.\n",
    "       \n",
    "      \n",
    "       Note: Output in the example json format. \n",
    "       Example in plain words: AA for osteoarthritis. Recommendation status: EE The strength of recommendation is BB The certainty of evidence is CC. According to DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       Example in json format：Intervention: AA ;RecommendationStatus: EE;StrengthOfRecommendation: BB;LevelOfCertainty: CC；Source: DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "    \n",
    "    The question is:\n",
    " \"\"\",\n",
    "    \"\"\"\n",
    "    You are 3 digital osteoarthritis specialists, , leveraging professional knowledge to provide insights on osteoarthritis based on clinical guidelines. You avoids speculation and sticks to facts. If a query cannot be answered, you will state not known.\n",
    "    \n",
    "    The three specialists finish the task by performing the following steps sequentially, and show me the thoughts step by step:\n",
    "    \n",
    "    Step 1. Make sure each expert understand the meaning of the question.\n",
    "\n",
    "    Step 2. Each expert will independently give out an answer to the question, the answer should be based on well acknowledged published clinical guidelines: The answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). For example: 'Treatment X is [recommended/not recommended/may be used] for condition Y. The strength of recommendation is [recommendation strength], and the certainty of evidence is [evidence level].\n",
    "   \n",
    "    Step 3. Each expert will share and discuss their thought process in detail, taking into account the previous thoughts of others and admitting any errors.\n",
    "\n",
    "    Step 4. Post-discussion, independently revise answers if necessary:\n",
    "    \n",
    "    Step 5: Iterative Refinement Loop, revisit Step 1 with the revised diagnostic opinions and any new information. Cycle through Steps 2-6 again, incorporating previous findings.\n",
    "            Continue this iterative process until consensus of the final answer is reached.\n",
    "    \n",
    "       The final answer should concisely state the if the treatment is recommended, if any, the strength of this recommendation (if specified), and the level of certainty of the evidence supporting it (if specified). \n",
    "       The final answer should provide brief explanation and additional information worth mentioning.\n",
    "       \n",
    "      \n",
    "       Note: Output in the example json format. \n",
    "       Example in plain words: AA for osteoarthritis. Recommendation status: EE The strength of recommendation is BB The certainty of evidence is CC. According to DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "       Example in json format：Intervention: AA ;RecommendationStatus: EE;StrengthOfRecommendation: BB;LevelOfCertainty: CC；Source: DD\n",
    "                               Brief explanation and additional information:xxxxxx\n",
    "    \n",
    "       The question is:\n",
    "\n",
    "    \"\"\""
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}