Langchain_gemini.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "84d12ad1-b135-496f-be37-aa842b50a9e5",
   "metadata": {},
   "source": [
    "<h2>LangChain & Google Gemini API</h2>\n",
    "<h5><a href=\"https://ai.google.dev/\">Google AI for Developers</a> | <a href=\"https://api.python.langchain.com/en/latest/google_genai_api_reference.html\">LangChain Google</a></h5> "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e556b9f9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# API KEY using os.environ\n",
    "'''\n",
    "import os\n",
    "\n",
    "llm_key = 'GOOGLE_API_KEY' # OPENAI_API_KEY\n",
    "os.environ[llm_key] = '##' \n",
    "api_key = os.environ[llm_key]\n",
    "'''"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "8bdcf8e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read API KEY from file\n",
    "import json\n",
    "\n",
    "file_path = 'llms_vault.json'\n",
    "llm_key = 'GOOGLE_API_KEY'\n",
    "\n",
    "try:\n",
    "    with open(file_path,'r') as f:\n",
    "        api_key = json.loads(f.read()).get(llm_key)\n",
    "        if not api_key:\n",
    "            raise ValueError(f\"Some issue with '{llm_key}' from '{file_path}'. Verify the key:'{llm_key}' if it exists!\")\n",
    "except FileNotFoundError:\n",
    "    raise FileNotFoundError(f\"File related issue encountered, verify file path and name '{file_path}'\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "eb640c2e-747a-47b0-a526-75fd3b3cc432",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "39"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(api_key) # api_key"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "32029c69",
   "metadata": {},
   "source": [
    "<h3>Using Gemini API</h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c6c15c48-c1bd-4435-92d2-85631aff98c9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://ai.google.dev/api\n",
    "import google.generativeai as genai\n",
    "\n",
    "genai.configure(api_key=api_key) # genai.configure(api_key=os.environ[llm_key]) \n",
    "\n",
    "# list model names, requirement here is for content generation.\n",
    "models = [model for model in genai.list_models()]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3d2ad15",
   "metadata": {},
   "source": [
    "<h3> Exploring genai using Pandas </h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "9e1a48ac",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd # for organizing data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "376fd8a9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Method</th>\n",
       "      <th>Description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ChatSession</td>\n",
       "      <td>Contains an ongoing conversation with the mode...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GenerationConfig</td>\n",
       "      <td>A simple dataclass used to configure the gener...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GenerativeModel</td>\n",
       "      <td>\\n    The `genai.GenerativeModel` class wraps ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>chat</td>\n",
       "      <td>Calls the API to initiate a chat with a model ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>chat_async</td>\n",
       "      <td>Calls the API to initiate a chat with a model ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>configure</td>\n",
       "      <td>Captures default client configuration.\\n\\n    ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>count_message_tokens</td>\n",
       "      <td>Calls the API to calculate the number of token...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>count_text_tokens</td>\n",
       "      <td>Calls the API to count the number of tokens in...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>create_tuned_model</td>\n",
       "      <td>Calls the API to initiate a tuning process tha...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>delete_file</td>\n",
       "      <td>Calls the API to permanently delete a specifie...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>delete_tuned_model</td>\n",
       "      <td>Calls the API to delete a specified tuned model</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>embed_content</td>\n",
       "      <td>Calls the API to create embeddings for content...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>embed_content_async</td>\n",
       "      <td>Calls the API to create async embeddings for c...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>generate_embeddings</td>\n",
       "      <td>Calls the API to create an embedding for the t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>generate_text</td>\n",
       "      <td>Calls the API to generate text based on the pr...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>get_base_model</td>\n",
       "      <td>Calls the API to fetch a base model by name.\\n...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>get_file</td>\n",
       "      <td>Calls the API to retrieve a specified file usi...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>get_model</td>\n",
       "      <td>Calls the API to fetch a model by name.\\n\\n   ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>get_operation</td>\n",
       "      <td>Calls the API to get a specific operation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>get_tuned_model</td>\n",
       "      <td>Calls the API to fetch a tuned model by name.\\...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>list_files</td>\n",
       "      <td>Calls the API to list files using a supported ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>list_models</td>\n",
       "      <td>Calls the API to list all available models.\\n\\...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>list_operations</td>\n",
       "      <td>Calls the API to list all operations</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>list_tuned_models</td>\n",
       "      <td>Calls the API to list all tuned models.\\n\\n   ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>update_tuned_model</td>\n",
       "      <td>Calls the API to push updates to a specified t...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>upload_file</td>\n",
       "      <td>Calls the API to upload a file using a support...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  Method                                        Description\n",
       "0            ChatSession  Contains an ongoing conversation with the mode...\n",
       "1       GenerationConfig  A simple dataclass used to configure the gener...\n",
       "2        GenerativeModel  \\n    The `genai.GenerativeModel` class wraps ...\n",
       "3                   chat  Calls the API to initiate a chat with a model ...\n",
       "4             chat_async  Calls the API to initiate a chat with a model ...\n",
       "5              configure  Captures default client configuration.\\n\\n    ...\n",
       "6   count_message_tokens  Calls the API to calculate the number of token...\n",
       "7      count_text_tokens  Calls the API to count the number of tokens in...\n",
       "8     create_tuned_model  Calls the API to initiate a tuning process tha...\n",
       "9            delete_file  Calls the API to permanently delete a specifie...\n",
       "10    delete_tuned_model    Calls the API to delete a specified tuned model\n",
       "11         embed_content  Calls the API to create embeddings for content...\n",
       "12   embed_content_async  Calls the API to create async embeddings for c...\n",
       "13   generate_embeddings  Calls the API to create an embedding for the t...\n",
       "14         generate_text  Calls the API to generate text based on the pr...\n",
       "15        get_base_model  Calls the API to fetch a base model by name.\\n...\n",
       "16              get_file  Calls the API to retrieve a specified file usi...\n",
       "17             get_model  Calls the API to fetch a model by name.\\n\\n   ...\n",
       "18         get_operation          Calls the API to get a specific operation\n",
       "19       get_tuned_model  Calls the API to fetch a tuned model by name.\\...\n",
       "20            list_files  Calls the API to list files using a supported ...\n",
       "21           list_models  Calls the API to list all available models.\\n\\...\n",
       "22       list_operations               Calls the API to list all operations\n",
       "23     list_tuned_models  Calls the API to list all tuned models.\\n\\n   ...\n",
       "24    update_tuned_model  Calls the API to push updates to a specified t...\n",
       "25           upload_file  Calls the API to upload a file using a support..."
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Explore 'genai', listing out some methods with it's description \n",
    "# dir(genai)\n",
    "\n",
    "genai_methods = []\n",
    "genai_descriptions = []\n",
    "\n",
    "for method_name in dir(genai):\n",
    "    if callable(getattr(genai, method_name)): # having doc\n",
    "        genai_methods.append(method_name)\n",
    "        method = getattr(genai, method_name)\n",
    "        genai_descriptions.append(method.__doc__)\n",
    "\n",
    "\n",
    "df_genai = pd.DataFrame({'Method': genai_methods, 'Description': genai_descriptions})\n",
    "\n",
    "df_genai\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "673d241d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Convert genai - models to pandas dataframe.\n",
    "\n",
    "df = pd.DataFrame(models, columns=['name', 'description', 'supported_generation_methods', 'display_name'])\n",
    "df.rename(columns={'name': 'model_name'}, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "id": "c9bb1015",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model_name</th>\n",
       "      <th>description</th>\n",
       "      <th>supported_generation_methods</th>\n",
       "      <th>display_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>models/chat-bison-001</td>\n",
       "      <td>A legacy text-only model optimized for chat co...</td>\n",
       "      <td>[generateMessage, countMessageTokens]</td>\n",
       "      <td>PaLM 2 Chat (Legacy)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>models/text-bison-001</td>\n",
       "      <td>A legacy model that understands text and gener...</td>\n",
       "      <td>[generateText, countTextTokens, createTunedTex...</td>\n",
       "      <td>PaLM 2 (Legacy)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>models/embedding-gecko-001</td>\n",
       "      <td>Obtain a distributed representation of a text.</td>\n",
       "      <td>[embedText, countTextTokens]</td>\n",
       "      <td>Embedding Gecko</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>models/gemini-1.0-pro</td>\n",
       "      <td>The best model for scaling across a wide range...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>models/gemini-1.0-pro-001</td>\n",
       "      <td>The best model for scaling across a wide range...</td>\n",
       "      <td>[generateContent, countTokens, createTunedModel]</td>\n",
       "      <td>Gemini 1.0 Pro 001 (Tuning)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>models/gemini-1.0-pro-latest</td>\n",
       "      <td>The best model for scaling across a wide range...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>models/gemini-1.0-pro-vision-latest</td>\n",
       "      <td>The best image understanding model to handle a...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro Vision</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>models/gemini-1.5-flash</td>\n",
       "      <td>Fast and versatile multimodal model for scalin...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.5 Flash</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>models/gemini-1.5-flash-001</td>\n",
       "      <td>Fast and versatile multimodal model for scalin...</td>\n",
       "      <td>[generateContent, countTokens, createCachedCon...</td>\n",
       "      <td>Gemini 1.5 Flash 001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>models/gemini-1.5-flash-latest</td>\n",
       "      <td>Fast and versatile multimodal model for scalin...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.5 Flash Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>models/gemini-1.5-pro</td>\n",
       "      <td>Mid-size multimodal model that supports up to ...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.5 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>models/gemini-1.5-pro-001</td>\n",
       "      <td>Mid-size multimodal model that supports up to ...</td>\n",
       "      <td>[generateContent, countTokens, createCachedCon...</td>\n",
       "      <td>Gemini 1.5 Pro 001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>models/gemini-1.5-pro-latest</td>\n",
       "      <td>Mid-size multimodal model that supports up to ...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.5 Pro Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>models/gemini-pro</td>\n",
       "      <td>The best model for scaling across a wide range...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>models/gemini-pro-vision</td>\n",
       "      <td>The best image understanding model to handle a...</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro Vision</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>models/embedding-001</td>\n",
       "      <td>Obtain a distributed representation of a text.</td>\n",
       "      <td>[embedContent]</td>\n",
       "      <td>Embedding 001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>models/text-embedding-004</td>\n",
       "      <td>Obtain a distributed representation of a text.</td>\n",
       "      <td>[embedContent]</td>\n",
       "      <td>Text Embedding 004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>models/aqa</td>\n",
       "      <td>Model trained to return answers to questions t...</td>\n",
       "      <td>[generateAnswer]</td>\n",
       "      <td>Model that performs Attributed Question Answer...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                             model_name  \\\n",
       "0                 models/chat-bison-001   \n",
       "1                 models/text-bison-001   \n",
       "2            models/embedding-gecko-001   \n",
       "3                 models/gemini-1.0-pro   \n",
       "4             models/gemini-1.0-pro-001   \n",
       "5          models/gemini-1.0-pro-latest   \n",
       "6   models/gemini-1.0-pro-vision-latest   \n",
       "7               models/gemini-1.5-flash   \n",
       "8           models/gemini-1.5-flash-001   \n",
       "9        models/gemini-1.5-flash-latest   \n",
       "10                models/gemini-1.5-pro   \n",
       "11            models/gemini-1.5-pro-001   \n",
       "12         models/gemini-1.5-pro-latest   \n",
       "13                    models/gemini-pro   \n",
       "14             models/gemini-pro-vision   \n",
       "15                 models/embedding-001   \n",
       "16            models/text-embedding-004   \n",
       "17                           models/aqa   \n",
       "\n",
       "                                          description  \\\n",
       "0   A legacy text-only model optimized for chat co...   \n",
       "1   A legacy model that understands text and gener...   \n",
       "2      Obtain a distributed representation of a text.   \n",
       "3   The best model for scaling across a wide range...   \n",
       "4   The best model for scaling across a wide range...   \n",
       "5   The best model for scaling across a wide range...   \n",
       "6   The best image understanding model to handle a...   \n",
       "7   Fast and versatile multimodal model for scalin...   \n",
       "8   Fast and versatile multimodal model for scalin...   \n",
       "9   Fast and versatile multimodal model for scalin...   \n",
       "10  Mid-size multimodal model that supports up to ...   \n",
       "11  Mid-size multimodal model that supports up to ...   \n",
       "12  Mid-size multimodal model that supports up to ...   \n",
       "13  The best model for scaling across a wide range...   \n",
       "14  The best image understanding model to handle a...   \n",
       "15     Obtain a distributed representation of a text.   \n",
       "16     Obtain a distributed representation of a text.   \n",
       "17  Model trained to return answers to questions t...   \n",
       "\n",
       "                         supported_generation_methods  \\\n",
       "0               [generateMessage, countMessageTokens]   \n",
       "1   [generateText, countTextTokens, createTunedTex...   \n",
       "2                        [embedText, countTextTokens]   \n",
       "3                      [generateContent, countTokens]   \n",
       "4    [generateContent, countTokens, createTunedModel]   \n",
       "5                      [generateContent, countTokens]   \n",
       "6                      [generateContent, countTokens]   \n",
       "7                      [generateContent, countTokens]   \n",
       "8   [generateContent, countTokens, createCachedCon...   \n",
       "9                      [generateContent, countTokens]   \n",
       "10                     [generateContent, countTokens]   \n",
       "11  [generateContent, countTokens, createCachedCon...   \n",
       "12                     [generateContent, countTokens]   \n",
       "13                     [generateContent, countTokens]   \n",
       "14                     [generateContent, countTokens]   \n",
       "15                                     [embedContent]   \n",
       "16                                     [embedContent]   \n",
       "17                                   [generateAnswer]   \n",
       "\n",
       "                                         display_name  \n",
       "0                                PaLM 2 Chat (Legacy)  \n",
       "1                                     PaLM 2 (Legacy)  \n",
       "2                                     Embedding Gecko  \n",
       "3                                      Gemini 1.0 Pro  \n",
       "4                         Gemini 1.0 Pro 001 (Tuning)  \n",
       "5                               Gemini 1.0 Pro Latest  \n",
       "6                               Gemini 1.0 Pro Vision  \n",
       "7                                    Gemini 1.5 Flash  \n",
       "8                                Gemini 1.5 Flash 001  \n",
       "9                             Gemini 1.5 Flash Latest  \n",
       "10                                     Gemini 1.5 Pro  \n",
       "11                                 Gemini 1.5 Pro 001  \n",
       "12                              Gemini 1.5 Pro Latest  \n",
       "13                                     Gemini 1.0 Pro  \n",
       "14                              Gemini 1.0 Pro Vision  \n",
       "15                                      Embedding 001  \n",
       "16                                 Text Embedding 004  \n",
       "17  Model that performs Attributed Question Answer...  "
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Exploring the Model DataFrame\n",
    "\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "374d4992",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model_name</th>\n",
       "      <th>description</th>\n",
       "      <th>supported_generation_methods</th>\n",
       "      <th>display_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>18</td>\n",
       "      <td>18</td>\n",
       "      <td>18</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unique</th>\n",
       "      <td>18</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>top</th>\n",
       "      <td>models/chat-bison-001</td>\n",
       "      <td>Obtain a distributed representation of a text.</td>\n",
       "      <td>[generateContent, countTokens]</td>\n",
       "      <td>Gemini 1.0 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>freq</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>9</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   model_name                                     description  \\\n",
       "count                      18                                              18   \n",
       "unique                     18                                              10   \n",
       "top     models/chat-bison-001  Obtain a distributed representation of a text.   \n",
       "freq                        1                                               3   \n",
       "\n",
       "          supported_generation_methods    display_name  \n",
       "count                               18              18  \n",
       "unique                               8              16  \n",
       "top     [generateContent, countTokens]  Gemini 1.0 Pro  \n",
       "freq                                 9               2  "
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "id": "4ad6405d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total model: 18, No of unique supported_methods: 12\n"
     ]
    }
   ],
   "source": [
    "count_models = df.model_name.count()\n",
    "unique_supported_methods = df.supported_generation_methods.explode().nunique()\n",
    "print(f\"Total model: {count_models}, No of unique supported_methods: {unique_supported_methods}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "id": "096c26c8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "generateContent         12\n",
       "countTokens             12\n",
       "countTextTokens          2\n",
       "createCachedContent      2\n",
       "embedContent             2\n",
       "generateMessage          1\n",
       "countMessageTokens       1\n",
       "generateText             1\n",
       "createTunedTextModel     1\n",
       "embedText                1\n",
       "createTunedModel         1\n",
       "generateAnswer           1\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "supported_methods = [method for methods in df['supported_generation_methods'] for method in methods]\n",
    "count_of_methods = pd.Series(supported_methods).value_counts()\n",
    "count_of_methods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "id": "9784b905",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting pie chart\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(8, 6))\n",
    "count_of_methods.plot(kind='barh')\n",
    "plt.title('genai: Supported Generation Methods')\n",
    "\n",
    "plt.xlabel('Supported Generation Methods')\n",
    "plt.ylabel('Frequency')\n",
    "plt.grid(axis='x', alpha=0.3)  # Adding faint grid lines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "00ed2188",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>model_name</th>\n",
       "      <th>display_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>models/gemini-1.0-pro</td>\n",
       "      <td>Gemini 1.0 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>models/gemini-1.0-pro-001</td>\n",
       "      <td>Gemini 1.0 Pro 001 (Tuning)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>models/gemini-1.0-pro-latest</td>\n",
       "      <td>Gemini 1.0 Pro Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>models/gemini-1.0-pro-vision-latest</td>\n",
       "      <td>Gemini 1.0 Pro Vision</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>models/gemini-1.5-flash</td>\n",
       "      <td>Gemini 1.5 Flash</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>models/gemini-1.5-flash-001</td>\n",
       "      <td>Gemini 1.5 Flash 001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>models/gemini-1.5-flash-latest</td>\n",
       "      <td>Gemini 1.5 Flash Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>models/gemini-1.5-pro</td>\n",
       "      <td>Gemini 1.5 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>models/gemini-1.5-pro-001</td>\n",
       "      <td>Gemini 1.5 Pro 001</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>models/gemini-1.5-pro-latest</td>\n",
       "      <td>Gemini 1.5 Pro Latest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>models/gemini-pro</td>\n",
       "      <td>Gemini 1.0 Pro</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>models/gemini-pro-vision</td>\n",
       "      <td>Gemini 1.0 Pro Vision</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                             model_name                 display_name\n",
       "3                 models/gemini-1.0-pro               Gemini 1.0 Pro\n",
       "4             models/gemini-1.0-pro-001  Gemini 1.0 Pro 001 (Tuning)\n",
       "5          models/gemini-1.0-pro-latest        Gemini 1.0 Pro Latest\n",
       "6   models/gemini-1.0-pro-vision-latest        Gemini 1.0 Pro Vision\n",
       "7               models/gemini-1.5-flash             Gemini 1.5 Flash\n",
       "8           models/gemini-1.5-flash-001         Gemini 1.5 Flash 001\n",
       "9        models/gemini-1.5-flash-latest      Gemini 1.5 Flash Latest\n",
       "10                models/gemini-1.5-pro               Gemini 1.5 Pro\n",
       "11            models/gemini-1.5-pro-001           Gemini 1.5 Pro 001\n",
       "12         models/gemini-1.5-pro-latest        Gemini 1.5 Pro Latest\n",
       "13                    models/gemini-pro               Gemini 1.0 Pro\n",
       "14             models/gemini-pro-vision        Gemini 1.0 Pro Vision"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# filter out the 'supported_generation_methods' that contains 'generateContent' only, to be used with 'GenerativeModel'.\n",
    "\n",
    "content_df = df[df['supported_generation_methods'].apply(lambda x: 'generateContent' in x)]\n",
    "content_df[['model_name','display_name']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "8637035b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['models/gemini-1.0-pro',\n",
       " 'models/gemini-1.5-flash',\n",
       " 'models/gemini-1.5-pro',\n",
       " 'models/gemini-1.5-pro-latest']"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# selecting few random models name. \n",
    "selected_models = []\n",
    "selected_models.extend(content_df.loc[[3,7,10,12]]['model_name']) \n",
    "selected_models"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2391094a",
   "metadata": {},
   "source": [
    "<h3> Check the Input/Output token context for selected models.</h3>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ae4046d1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: models/gemini-1.0-pro, Limits: input-30720 output-2048\n",
      "Model: models/gemini-1.5-flash, Limits: input-1048576 output-8192\n",
      "Model: models/gemini-1.5-pro, Limits: input-2097152 output-8192\n",
      "Model: models/gemini-1.5-pro-latest, Limits: input-2097152 output-8192\n"
     ]
    }
   ],
   "source": [
    "for selected_model in selected_models:\n",
    "    info = genai.get_model(selected_model)\n",
    "    input_limit = info.input_token_limit\n",
    "    output_limit = info.output_token_limit\n",
    "    print(f\"Model: {selected_model}, Limits: input-{input_limit}, output-{output_limit}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2a8d142",
   "metadata": {},
   "source": [
    "<h3>Selecting model for generating content and preparing prompt</h3>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "id": "cfcac630",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = genai.GenerativeModel(model_name=selected_models[0]) \n",
    "\n",
    "# choosing model, we can also remove the text 'models/' from model text.\n",
    "# model = genai.GenerativeModel(model_name=selected_models[0].replace('models/',''))\n",
    "\n",
    "prompts = [\"list the colors in the rainbow with words in capital\"] # issuing prompt for model chosen. String can also be provided"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "id": "ec52531e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "total_tokens: 10"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Count the tokens, number of tokens for Request!\n",
    "\n",
    "tokens = model.count_tokens(prompts)\n",
    "tokens"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "440d86ee",
   "metadata": {},
   "source": [
    "<h3>Embedding: list of floating point numbers that represent the meaning of a sentence.</h3> "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "55b0701a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# embed = genai.embed_content(model=selected_models[0], content=prompts) \n",
    "\n",
    "# Code above, will result with \"BatchEmbedContentsRequest.requests[0].model: unexpected model name format\" ?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "586724e6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2     models/embedding-gecko-001\n",
       "15          models/embedding-001\n",
       "16     models/text-embedding-004\n",
       "Name: model_name, dtype: object"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We need the model that supports embedding. \n",
    "\n",
    "embed_df = df[df['supported_generation_methods'].apply(lambda e: \"embed\" in ''.join(e))] # df['supported_generation_methods'] is a []\n",
    "embed_df['model_name']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "id": "9ed50004",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Embed model: models/embedding-001      \n",
      " Embedding: [0.027544947, -0.045868676, -0.031625714, -0.05262981, 0.053381477, -0.021768298, 0.011543624, -0.019891426, 0.01956483, 0.03539606] -(partial),      \n",
      " Dimension: 768\n"
     ]
    }
   ],
   "source": [
    "# embed_df['model_name'].loc[15]\n",
    "embed_model = embed_df['model_name'].loc[15]\n",
    "embed = genai.embed_content(model=embed_model, content=prompts) \n",
    "\n",
    "print(f\"Embed model: {embed_model}\\\n",
    "      \\n Embedding: {str(embed['embedding'][0][:10])} -(partial),\\\n",
    "      \\n Dimension: {len(embed['embedding'][0])}\"\n",
    "      )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "b9677431",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "response:\n",
       "GenerateContentResponse(\n",
       "    done=True,\n",
       "    iterator=None,\n",
       "    result=protos.GenerateContentResponse({\n",
       "      \"candidates\": [\n",
       "        {\n",
       "          \"content\": {\n",
       "            \"parts\": [\n",
       "              {\n",
       "                \"text\": \"ROYAL RED\\nORANGE OPULENCE\\nYELLOW YELL\\nGREEN GRANDEUR\\nBLUE BRILLIANCE\\nINDIGO ILLUMINATION\\nVIOLET VIBRANCY\"\n",
       "              }\n",
       "            ],\n",
       "            \"role\": \"model\"\n",
       "          },\n",
       "          \"finish_reason\": \"STOP\",\n",
       "          \"index\": 0,\n",
       "          \"safety_ratings\": [\n",
       "            {\n",
       "              \"category\": \"HARM_CATEGORY_SEXUALLY_EXPLICIT\",\n",
       "              \"probability\": \"NEGLIGIBLE\"\n",
       "            },\n",
       "            {\n",
       "              \"category\": \"HARM_CATEGORY_HATE_SPEECH\",\n",
       "              \"probability\": \"NEGLIGIBLE\"\n",
       "            },\n",
       "            {\n",
       "              \"category\": \"HARM_CATEGORY_HARASSMENT\",\n",
       "              \"probability\": \"NEGLIGIBLE\"\n",
       "            },\n",
       "            {\n",
       "              \"category\": \"HARM_CATEGORY_DANGEROUS_CONTENT\",\n",
       "              \"probability\": \"NEGLIGIBLE\"\n",
       "            }\n",
       "          ]\n",
       "        }\n",
       "      ],\n",
       "      \"usage_metadata\": {\n",
       "        \"prompt_token_count\": 11,\n",
       "        \"candidates_token_count\": 33,\n",
       "        \"total_token_count\": 44\n",
       "      }\n",
       "    }),\n",
       ")"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# https://ai.google.dev/api/python/google/generativeai/GenerativeModel#generate_content\n",
    "response = model.generate_content(prompts) \n",
    "response"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "27477c35",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ROYAL RED\n",
      "ORANGE OPULENCE\n",
      "YELLOW YELL\n",
      "GREEN GRANDEUR\n",
      "BLUE BRILLIANCE\n",
      "INDIGO ILLUMINATION\n",
      "VIOLET VIBRANCY\n"
     ]
    }
   ],
   "source": [
    "print(response.text)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6fbd0e5d",
   "metadata": {},
   "source": [
    "<hr/>\n",
    "<h2>Using LangChain</h2>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "id": "d56744aa",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain_google_genai import ChatGoogleGenerativeAI\n",
    "from langchain_google_genai import GoogleGenerativeAIEmbeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "1cbb1485",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.environ[\"GOOGLE_API_KEY\"] = api_key"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "id": "d7805c97",
   "metadata": {},
   "outputs": [],
   "source": [
    "llm = ChatGoogleGenerativeAI(\n",
    "    google_api_key = api_key,\n",
    "    model=selected_models[0],\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "id": "4f33a485",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Embed model: models/embedding-001      \n",
      " Embedding: [0.027544947, -0.045868676, -0.031625714, -0.05262981, 0.053381477, -0.021768298, 0.011543624, -0.019891426, 0.01956483, 0.03539606] -(partial),      \n",
      " Dimension: 768\n"
     ]
    }
   ],
   "source": [
    "# Embedding\n",
    "\n",
    "embeddings = GoogleGenerativeAIEmbeddings(model=embed_model)\n",
    "embeddings.embed_query(prompts[0])\n",
    "\n",
    "print(f\"Embed model: {embed_model}\\\n",
    "      \\n Embedding: {str(embed['embedding'][0][:10])} -(partial),\\\n",
    "      \\n Dimension: {len(embed['embedding'][0])}\"\n",
    "      )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "id": "653bd925",
   "metadata": {},
   "outputs": [],
   "source": [
    "gem_response = llm.invoke(prompts[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "c01bd66b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "AIMessage(content='RED\\nORANGE\\nYELLOW\\nGREEN\\nBLUE\\nINDIGO\\nVIOLET \\n', response_metadata={'prompt_feedback': {'block_reason': 0, 'safety_ratings': []}, 'finish_reason': 'STOP', 'safety_ratings': [{'category': 'HARM_CATEGORY_SEXUALLY_EXPLICIT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HATE_SPEECH', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_HARASSMENT', 'probability': 'NEGLIGIBLE', 'blocked': False}, {'category': 'HARM_CATEGORY_DANGEROUS_CONTENT', 'probability': 'NEGLIGIBLE', 'blocked': False}]}, id='run-a1b5ba10-51e1-481c-9c41-703bded29c42-0', usage_metadata={'input_tokens': 12, 'output_tokens': 15, 'total_tokens': 27})"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gem_response"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "id": "82844b26",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RED\n",
      "ORANGE\n",
      "YELLOW\n",
      "GREEN\n",
      "BLUE\n",
      "INDIGO\n",
      "VIOLET \n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(gem_response.content)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "0a815f58",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'input_tokens': 12, 'output_tokens': 15, 'total_tokens': 27}\n"
     ]
    }
   ],
   "source": [
    "print(gem_response.usage_metadata)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.12",
   "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": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}