cognee/notebooks/cognee_graphiti_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Cognee Graphiti integration demo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "vscode": {
     "languageId": "plaintext"
    }
   },
   "source": [
    "First we import the necessary libaries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cognee\n",
    "from cognee.shared.logging_utils import get_logger, ERROR\n",
    "import warnings\n",
    "from cognee.modules.pipelines import Task, run_tasks\n",
    "from cognee.tasks.temporal_awareness import build_graph_with_temporal_awareness\n",
    "from cognee.infrastructure.databases.relational import (\n",
    "    create_db_and_tables as create_relational_db_and_tables,\n",
    ")\n",
    "from cognee.tasks.temporal_awareness.index_graphiti_objects import (\n",
    "    index_and_transform_graphiti_nodes_and_edges,\n",
    ")\n",
    "from cognee.modules.retrieval.utils.brute_force_triplet_search import brute_force_triplet_search\n",
    "from cognee.modules.retrieval.graph_completion_retriever import GraphCompletionRetriever\n",
    "from cognee.infrastructure.llm.prompts import read_query_prompt, render_prompt\n",
    "from cognee.infrastructure.llm.get_llm_client import get_llm_client"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set environment variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-01-15T10:43:57.893763Z",
     "start_time": "2025-01-15T10:43:57.891332Z"
    }
   },
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "# We ignore warnigns for now\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "# API key for cognee\n",
    "if \"LLM_API_KEY\" not in os.environ:\n",
    "    os.environ[\"LLM_API_KEY\"] = \"\"\n",
    "\n",
    "# API key for graphiti\n",
    "if \"OPENAI_API_KEY\" not in os.environ:\n",
    "    os.environ[\"OPENAI_API_KEY\"] = \"\"\n",
    "\n",
    "GRAPH_DATABASE_PROVIDER = \"neo4j\"\n",
    "GRAPH_DATABASE_USERNAME = \"neo4j\"\n",
    "GRAPH_DATABASE_PASSWORD = \"pleaseletmein\"\n",
    "GRAPH_DATABASE_URL = \"bolt://localhost:7687\"\n",
    "\n",
    "os.environ[\"GRAPH_DATABASE_PROVIDER\"] = GRAPH_DATABASE_PROVIDER\n",
    "os.environ[\"GRAPH_DATABASE_USERNAME\"] = GRAPH_DATABASE_USERNAME\n",
    "os.environ[\"GRAPH_DATABASE_PASSWORD\"] = GRAPH_DATABASE_PASSWORD\n",
    "os.environ[\"GRAPH_DATABASE_URL\"] = GRAPH_DATABASE_URL\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Input texts with temporal information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-01-15T10:43:57.928664Z",
     "start_time": "2025-01-15T10:43:57.927105Z"
    }
   },
   "outputs": [],
   "source": [
    "text_list = [\n",
    "    \"Kamala Harris is the Attorney General of California. She was previously \"\n",
    "    \"the district attorney for San Francisco.\",\n",
    "    \"As AG, Harris was in office from January 3, 2011 – January 3, 2017\",\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Running graphiti + transforming its graph into cognee's core system (graph transformation + vector embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-01-15T10:44:25.008501Z",
     "start_time": "2025-01-15T10:43:57.932240Z"
    }
   },
   "outputs": [],
   "source": [
    "# 🔧 Setting Up Logging to Suppress Errors\n",
    "logger = get_logger(level=ERROR) # Keeping logs clean and focused\n",
    "\n",
    "# 🧹 Pruning Old Data and Metadata\n",
    "await cognee.prune.prune_data()  # Removing outdated data\n",
    "await cognee.prune.prune_system(metadata=True)\n",
    "\n",
    "# 🏗️ Creating Relational Database and Tables\n",
    "await create_relational_db_and_tables()\n",
    "\n",
    "# 📚 Adding Text Data to Cognee\n",
    "for text in text_list:\n",
    "    await cognee.add(text)\n",
    "\n",
    "# 🕰️ Building Temporal-Aware Graphs\n",
    "tasks = [\n",
    "    Task(build_graph_with_temporal_awareness, text_list=text_list),\n",
    "]\n",
    "\n",
    "# 🚀 Running the Task Pipeline\n",
    "pipeline = run_tasks(tasks)\n",
    "\n",
    "# 🌟 Processing Pipeline Results\n",
    "async for result in pipeline:\n",
    "    print(f\"✅ Result Processed: {result}\")\n",
    "\n",
    "# 🔄 Indexing and Transforming Graph Data\n",
    "await index_and_transform_graphiti_nodes_and_edges()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Retrieving and generating answer from graphiti graph with cognee retriever"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2025-01-15T10:44:27.844438Z",
     "start_time": "2025-01-15T10:44:25.013325Z"
    }
   },
   "outputs": [],
   "source": [
    "# Step 1: Formulating the Query 🔍\n",
    "query = \"When was Kamala Harris in office?\"\n",
    "\n",
    "# Step 2: Searching for Relevant Triplets 📊\n",
    "triplets = await brute_force_triplet_search(\n",
    "    query=query,\n",
    "    top_k=3,\n",
    "    collections=[\"graphitinode_content\", \"graphitinode_name\", \"graphitinode_summary\"],\n",
    ")\n",
    "\n",
    "# Step 3: Preparing the Context for the LLM\n",
    "retriever = GraphCompletionRetriever()\n",
    "context = await retriever.resolve_edges_to_text(triplets)\n",
    "\n",
    "args = {\"question\": query, \"context\": context}\n",
    "\n",
    "# Step 4: Generating Prompts ✍️\n",
    "user_prompt = render_prompt(\"graph_context_for_question.txt\", args)\n",
    "system_prompt = read_query_prompt(\"answer_simple_question_restricted.txt\")\n",
    "\n",
    "# Step 5: Interacting with the LLM 🤖\n",
    "llm_client = get_llm_client()\n",
    "computed_answer = await llm_client.acreate_structured_output(\n",
    "    text_input=user_prompt,  # Input prompt for the user context\n",
    "    system_prompt=system_prompt,  # System-level instructions for the model\n",
    "    response_model=str,\n",
    ")\n",
    "\n",
    "# Step 6: Displaying the Computed Answer ✨\n",
    "print(f\"💡 Answer: {computed_answer}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.11.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}