diff --git a/cookbook/structured_report_generation.ipynb b/cookbook/structured_report_generation.ipynb
new file mode 100644
index 000000000..e147c21af
--- /dev/null
+++ b/cookbook/structured_report_generation.ipynb
@@ -0,0 +1,1603 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "vscode": {
+ "languageId": "plaintext"
+ },
+ "id": "nB0yIE5qfKkq"
+ },
+ "source": [
+ "# Structured Report Generation\n",
+ "[](https://console.brev.dev/launchable/deploy?launchableID=env-2qzwjeLSpDcjU68lNQyDcZwxmn9)\n",
+ "\n",
+ "Deploy with Launchables. Launchables are pre-configured, fully optimized environments that users can deploy with a single click.\n",
+ "\n",
+ "In this notebook, you will use the newest llama model, llama-3.3-70b, to generate a report on a given topic. You will use Langchain's LangGraph to build an Agent that takes in user-defined topics and structure, then plans the topics of the section indicated in the structure. Next, the Agent uses Tavily to do web search on the given topics and uses this information to write the sections and synthesize the final report.\n",
+ "\n",
+ "Rather than deploying the model locally, you leverage NVIDIA API Catalog by calling the model's NIM API Endpoint. As you don't need a GPU to run the model, you can run this notebook anywhere!\n",
+ "\n",
+ "You can find the original notebook on LangChain's GitHub [here](https://github.com/langchain-ai/report-mAIstro).\n",
+ "\n",
+ "Below is the architecture diagram.\n",
+ "\n",
+ "# \n",
+ "\n",
+ "\n",
+ "\n",
+ "The Agent takes in user defined topics and structure, then plans the topics of the section indicated in the structure. The Agent then uses Tavily to do web search on the given topics and uses this information to write the sections and synthesize the final report.\n",
+ "\n",
+ "\n",
+ "A two-phase approach is used for planning and research:\n",
+ "\n",
+ "\n",
+ "Phase 1 - Planning\n",
+ "- Analyzes user inputs\n",
+ "- Maps out report sections\n",
+ "\n",
+ "\n",
+ "Phase 2 - Research\n",
+ "- Conducts parallel web research via Tavily API\n",
+ "- Processes relevant data for each section\n",
+ "\n",
+ "\n",
+ "\n",
+ "The report is then written in a strategic sequence:\n",
+ "- Write research-based sections in parallel\n",
+ "- Write introductions, conclusions, and connect each of the sections\n",
+ "\n",
+ "\n",
+ "All sections maintain awareness of each other's content for consistency."
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "3mLEcH4efKk-"
+ },
+ "source": [
+ "# Content Overview\n",
+ ">[Prerequisites](#Prerequisites) \n",
+ ">[Writing Report Plan](#Writing-Report-Plan) \n",
+ ">[Research and Writing](#Research-and-Writing) \n",
+ ">[Write Single Section](#Write-Single-Section) \n",
+ ">[Validate Single Section](#Validate-Single-Section) \n",
+ ">[Write All Sections](#Write-All-Sections) \n",
+ ">[Validate Final Report](#Final-Report)\n",
+ "________________________\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "zKE7UqdqfKk-"
+ },
+ "source": [
+ "## Prerequisites\n",
+ "\n",
+ "### Install Dependencies"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 13,
+ "metadata": {
+ "id": "-nyq12MGfKk-"
+ },
+ "outputs": [],
+ "source": [
+ "%%capture --no-stderr\n",
+ "%pip install --quiet -U langgraph langchain_community langchain_core tavily-python langchain_nvidia_ai_endpoints"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "b_Zlx5RbfKk_"
+ },
+ "source": [
+ "## API Keys\n",
+ "Prior to getting started, you will need to create API Keys for the NVIDIA API Catalog, Tavily, and LangChain.\n",
+ "\n",
+ "- NVIDIA NIM Trial API Key\n",
+ " 1. Prior to getting started, you will need to create API Keys to access NVIDIA NIM trial hosted endpoints.\n",
+ " 2. If you don’t have an NVIDIA account, you will be asked to sign-up.\n",
+ " 3. Click [here](https://build.nvidia.com/meta/llama-3_3-70b-instruct?signin=true&api_key=true) to sign-in and get an API key\n",
+ "- LangChain\n",
+ " 1. Go to **[LangChain Settings page](https://smith.langchain.com/settings)**. You will need to create an account if you have not already.\n",
+ " 2. On the left panel, navigate to \"API Keys\".\n",
+ " 3. Click on the \"Create API Key\" on the top right of the page.\n",
+ "- Tavily\n",
+ " 1. Go to the **[Tavily homepage](https://tavily.com/)** and click on \"Get Started\"\n",
+ " 2. Sign in or create an account.\n",
+ " 3. Create an API Key.\n",
+ "\n",
+ "### Export API Keys\n",
+ "\n",
+ "Save these API Keys as environment variables.\n",
+ "\n",
+ "First, set the NVIDIA API Key as the environment variable."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 14,
+ "metadata": {
+ "scrolled": true,
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "0lWPnmq1fKk_",
+ "outputId": "ff1f737a-1617-4298-a5bd-3c353bc1fcb7"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Enter your NVIDIA API key: ··········\n"
+ ]
+ }
+ ],
+ "source": [
+ "import getpass\n",
+ "import os\n",
+ "\n",
+ "if not os.environ.get(\"NVIDIA_API_KEY\", \"\").startswith(\"nvapi-\"):\n",
+ " nvapi_key = getpass.getpass(\"Enter your NVIDIA API key: \")\n",
+ " assert nvapi_key.startswith(\"nvapi-\"), f\"{nvapi_key[:5]}... is not a valid key\"\n",
+ " os.environ[\"NVIDIA_API_KEY\"] = nvapi_key"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "vscode": {
+ "languageId": "plaintext"
+ },
+ "id": "oXdWXhIZfKk_"
+ },
+ "source": [
+ "Next, set the LangChain API Key as an environment variable. You will use [LangSmith](https://docs.smith.langchain.com/) for [tracing](https://docs.smith.langchain.com/concepts/tracing)."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "9LMDW_36fKk_",
+ "outputId": "e35ad86f-f050-4598-da1c-68b1e5210064"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "LANGCHAIN_API_KEY: ··········\n"
+ ]
+ }
+ ],
+ "source": [
+ "import os, getpass\n",
+ "\n",
+ "def _set_env(var: str):\n",
+ " if not os.environ.get(var):\n",
+ " os.environ[var] = getpass.getpass(f\"{var}: \")\n",
+ "\n",
+ "_set_env(\"LANGCHAIN_API_KEY\")\n",
+ "os.environ[\"LANGCHAIN_TRACING_V2\"] = \"true\"\n",
+ "os.environ[\"LANGCHAIN_PROJECT\"] = \"report-mAIstro\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "Bs3CG-rNfKk_"
+ },
+ "source": [
+ "Finally, set the Tavily API Key as an environment variable. You will use [Tavily API](https://tavily.com/) web search tool."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 11,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "_oNAkhtefKk_",
+ "outputId": "e6c61383-9070-4d91-adcc-c3d0ccf05483"
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "TAVILY_API_KEY: ··········\n"
+ ]
+ }
+ ],
+ "source": [
+ "_set_env(\"TAVILY_API_KEY\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 15,
+ "metadata": {
+ "id": "-9S-0BlzfKk_"
+ },
+ "outputs": [],
+ "source": [
+ "from tavily import TavilyClient, AsyncTavilyClient\n",
+ "tavily_client = TavilyClient()\n",
+ "tavily_async_client = AsyncTavilyClient()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "cZBUlvcdfKk_"
+ },
+ "source": [
+ "### Working with the NVIDIA API Catalog\n",
+ "\n",
+ "Let's test the API endpoint.\n",
+ "\n",
+ "In this notebook, you will use the newest llama model, llama-3.3-70b-instruct, as the LLM. Define the LLM below and test the API Catalog.\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 17,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "-GIhfQyZfKk_",
+ "outputId": "f26b2dfc-efca-4952-b1bd-6fa0fd883da9"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stdout",
+ "text": [
+ "(Verse 1)\n",
+ "In realms of code, where AI does roam\n",
+ "A framework emerged, to make it back home\n",
+ "LangChain, a name that's known so well\n",
+ "A tool for builders, to stories to tell\n",
+ "\n",
+ "(Chorus)\n",
+ "Oh LangChain, oh LangChain, a chain of might\n",
+ "Connecting languages, in the dark of night\n",
+ "A bridge of code, that spans so wide\n",
+ "Helping developers, to reach the other side\n",
+ "\n",
+ "(Verse 2)\n",
+ "With LLMs and APIs, it does abide\n",
+ "A hub of knowledge, where data does reside\n",
+ "It weaves a tapestry, of language and art\n",
+ "A symphony of code, that beats within the heart\n",
+ "\n",
+ "(Chorus)\n",
+ "Oh LangChain, oh LangChain, a chain of might\n",
+ "Connecting languages, in the dark of night\n",
+ "A bridge of code, that spans so wide\n",
+ "Helping developers, to reach the other side\n",
+ "\n",
+ "(Verse 3)\n",
+ "From Python to JavaScript, it does stride\n",
+ "A path of integration, where all can reside\n",
+ "It breaks the barriers, of language and space\n",
+ "A unified platform, where innovation takes its place\n",
+ "\n",
+ "(Chorus)\n",
+ "Oh LangChain, oh LangChain, a chain of might\n",
+ "Connecting languages, in the dark of night\n",
+ "A bridge of code, that spans so wide\n",
+ "Helping developers, to reach the other side\n",
+ "\n",
+ "(Bridge)\n",
+ "With every line, of code that's written\n",
+ "A new world unfolds, where possibilities are smitten\n",
+ "The future is bright, with LangChain in sight\n",
+ "A guiding light, that shines through the dark of night\n",
+ "\n",
+ "(Verse 4)\n",
+ "The community grows, with each passing day\n",
+ "A collective effort, to drive the framework on its way\n",
+ "With every contribution, a new story's told\n",
+ "Of LangChain's power, to make the impossible unfold\n",
+ "\n",
+ "(Chorus)\n",
+ "Oh LangChain, oh LangChain, a chain of might\n",
+ "Connecting languages, in the dark of night\n",
+ "A bridge of code, that spans so wide\n",
+ "Helping developers, to reach the other side\n",
+ "\n",
+ "(Outro)\n",
+ "So let the code flow, like a river's stream\n",
+ "With LangChain as the guide, the future's dream\n",
+ "A world of possibility, where all can thrive\n",
+ "With LangChain, the chain that connects, and survives.\n"
+ ]
+ }
+ ],
+ "source": [
+ "## Core LC Chat Interface\n",
+ "from langchain_nvidia_ai_endpoints import ChatNVIDIA\n",
+ "\n",
+ "llm = ChatNVIDIA(model=\"meta/llama-3.3-70b-instruct\", temperature=0)\n",
+ "result = llm.invoke(\"Write a ballad about LangChain.\")\n",
+ "print(result.content)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "yC5c61XbfKk_"
+ },
+ "source": [
+ "### Optional: Locally Run NVIDIA NIM Microservices"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "IA-KDt1xfKk_"
+ },
+ "source": [
+ "Once you familiarize yourself with this blueprint, you may want to self-host models with NVIDIA NIM Microservices using NVIDIA AI Enterprise software license. This gives you the ability to run models anywhere, giving you ownership of your customizations and full control of your intellectual property (IP) and AI applications.\n",
+ "\n",
+ "[Learn more about NIM Microservices](https://developer.nvidia.com/blog/nvidia-nim-offers-optimized-inference-microservices-for-deploying-ai-models-at-scale/)\n",
+ "\n",
+ "\n",
+ "NOTE: Run the following cell only if you're using a local NIM Microservice instead of the API Catalog Endpoint.\n",
+ "
"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "id": "7lG67fmPfKk_"
+ },
+ "outputs": [],
+ "source": [
+ "from langchain_nvidia_ai_endpoints import ChatNVIDIA\n",
+ "\n",
+ "# connect to an embedding NIM running at localhost:8000, specifying a model\n",
+ "llm = ChatNVIDIA(base_url=\"http://localhost:8000/v1\", model=\"meta/llama3-8b-instruct\")"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "PrOoIm6pfKk_"
+ },
+ "source": [
+ "# Writing Report Plan"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "waSG2qulfKk_"
+ },
+ "source": [
+ "## Creating Utils Functions\n",
+ "\n",
+ "Next, you will create Utility functions that will be used for web research during report generation."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {
+ "id": "oqTRX4VvfKlA"
+ },
+ "outputs": [],
+ "source": [
+ "import asyncio\n",
+ "from langsmith import traceable\n",
+ "from pydantic import BaseModel, Field\n",
+ "\n",
+ "class Section(BaseModel):\n",
+ " name: str = Field(\n",
+ " description=\"Name for this section of the report.\",\n",
+ " )\n",
+ " description: str = Field(\n",
+ " description=\"Brief overview of the main topics and concepts to be covered in this section.\",\n",
+ " )\n",
+ " research: bool = Field(\n",
+ " description=\"Whether to perform web research for this section of the report.\"\n",
+ " )\n",
+ " content: str = Field(\n",
+ " description=\"The content of the section.\"\n",
+ " )\n",
+ "\n",
+ "def deduplicate_and_format_sources(search_response, max_tokens_per_source, include_raw_content=True):\n",
+ " \"\"\"\n",
+ " Takes either a single search response or list of responses from Tavily API and formats them.\n",
+ " Limits the raw_content to approximately max_tokens_per_source.\n",
+ " include_raw_content specifies whether to include the raw_content from Tavily in the formatted string.\n",
+ "\n",
+ " Args:\n",
+ " search_response: Either:\n",
+ " - A dict with a 'results' key containing a list of search results\n",
+ " - A list of dicts, each containing search results\n",
+ "\n",
+ " Returns:\n",
+ " str: Formatted string with deduplicated sources\n",
+ " \"\"\"\n",
+ " # Convert input to list of results\n",
+ " if isinstance(search_response, dict):\n",
+ " sources_list = search_response['results']\n",
+ " elif isinstance(search_response, list):\n",
+ " sources_list = []\n",
+ " for response in search_response:\n",
+ " if isinstance(response, dict) and 'results' in response:\n",
+ " sources_list.extend(response['results'])\n",
+ " else:\n",
+ " sources_list.extend(response)\n",
+ " else:\n",
+ " raise ValueError(\"Input must be either a dict with 'results' or a list of search results\")\n",
+ "\n",
+ " # Deduplicate by URL\n",
+ " unique_sources = {}\n",
+ " for source in sources_list:\n",
+ " if source['url'] not in unique_sources:\n",
+ " unique_sources[source['url']] = source\n",
+ "\n",
+ " # Format output\n",
+ " formatted_text = \"Sources:\\n\\n\"\n",
+ " for i, source in enumerate(unique_sources.values(), 1):\n",
+ " formatted_text += f\"Source {source['title']}:\\n===\\n\"\n",
+ " formatted_text += f\"URL: {source['url']}\\n===\\n\"\n",
+ " formatted_text += f\"Most relevant content from source: {source['content']}\\n===\\n\"\n",
+ " if include_raw_content:\n",
+ " # Using rough estimate of 4 characters per token\n",
+ " char_limit = max_tokens_per_source * 4\n",
+ " # Handle None raw_content\n",
+ " raw_content = source.get('raw_content', '')\n",
+ " if raw_content is None:\n",
+ " raw_content = ''\n",
+ " print(f\"Warning: No raw_content found for source {source['url']}\")\n",
+ " if len(raw_content) > char_limit:\n",
+ " raw_content = raw_content[:char_limit] + \"... [truncated]\"\n",
+ " formatted_text += f\"Full source content limited to {max_tokens_per_source} tokens: {raw_content}\\n\\n\"\n",
+ "\n",
+ " return formatted_text.strip()\n",
+ "\n",
+ "def format_sections(sections: list[Section]) -> str:\n",
+ " \"\"\" Format a list of sections into a string \"\"\"\n",
+ " formatted_str = \"\"\n",
+ " for idx, section in enumerate(sections, 1):\n",
+ " formatted_str += f\"\"\"\n",
+ "{'='*60}\n",
+ "Section {idx}: {section.name}\n",
+ "{'='*60}\n",
+ "Description:\n",
+ "{section.description}\n",
+ "Requires Research:\n",
+ "{section.research}\n",
+ "\n",
+ "Content:\n",
+ "{section.content if section.content else '[Not yet written]'}\n",
+ "\n",
+ "\"\"\"\n",
+ " return formatted_str\n",
+ "\n",
+ "@traceable\n",
+ "def tavily_search(query):\n",
+ " \"\"\" Search the web using the Tavily API.\n",
+ "\n",
+ " Args:\n",
+ " query (str): The search query to execute\n",
+ "\n",
+ " Returns:\n",
+ " dict: Tavily search response containing:\n",
+ " - results (list): List of search result dictionaries, each containing:\n",
+ " - title (str): Title of the search result\n",
+ " - url (str): URL of the search result\n",
+ " - content (str): Snippet/summary of the content\n",
+ " - raw_content (str): Full content of the page if available\"\"\"\n",
+ "\n",
+ " return tavily_client.search(query,\n",
+ " max_results=5,\n",
+ " include_raw_content=True)\n",
+ "\n",
+ "@traceable\n",
+ "async def tavily_search_async(search_queries, tavily_topic, tavily_days):\n",
+ " \"\"\"\n",
+ " Performs concurrent web searches using the Tavily API.\n",
+ "\n",
+ " Args:\n",
+ " search_queries (List[SearchQuery]): List of search queries to process\n",
+ " tavily_topic (str): Type of search to perform ('news' or 'general')\n",
+ " tavily_days (int): Number of days to look back for news articles (only used when tavily_topic='news')\n",
+ "\n",
+ " Returns:\n",
+ " List[dict]: List of search results from Tavily API, one per query\n",
+ "\n",
+ " Note:\n",
+ " For news searches, each result will include articles from the last `tavily_days` days.\n",
+ " For general searches, the time range is unrestricted.\n",
+ " \"\"\"\n",
+ "\n",
+ " search_tasks = []\n",
+ " for query in search_queries:\n",
+ " if tavily_topic == \"news\":\n",
+ " search_tasks.append(\n",
+ " tavily_async_client.search(\n",
+ " query,\n",
+ " max_results=5,\n",
+ " include_raw_content=True,\n",
+ " topic=\"news\",\n",
+ " days=tavily_days\n",
+ " )\n",
+ " )\n",
+ " else:\n",
+ " search_tasks.append(\n",
+ " tavily_async_client.search(\n",
+ " query,\n",
+ " max_results=5,\n",
+ " include_raw_content=True,\n",
+ " topic=\"general\"\n",
+ " )\n",
+ " )\n",
+ "\n",
+ " # Execute all searches concurrently\n",
+ " search_docs = await asyncio.gather(*search_tasks)\n",
+ "\n",
+ " return search_docs"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "iiawY70_fKlA"
+ },
+ "source": [
+ "## Planning\n",
+ "\n",
+ "First, let's define the Schema for report sections."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {
+ "id": "KzMTjAg1fKlA"
+ },
+ "outputs": [],
+ "source": [
+ "from typing_extensions import TypedDict\n",
+ "from typing import Annotated, List, Optional, Literal\n",
+ "\n",
+ "class Sections(BaseModel):\n",
+ " sections: List[Section] = Field(\n",
+ " description=\"Sections of the report.\",\n",
+ " )\n",
+ "class SearchQuery(BaseModel):\n",
+ " search_query: str = Field(\n",
+ " None, description=\"Query for web search.\"\n",
+ " )\n",
+ "class Queries(BaseModel):\n",
+ " queries: List[SearchQuery] = Field(\n",
+ " description=\"List of search queries.\",\n",
+ " )"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "vwptoWm5fKlA"
+ },
+ "source": [
+ "Now you will define the LangGraph state. Each state will have the following fields."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "metadata": {
+ "id": "jNoPHnZMfKlA"
+ },
+ "outputs": [],
+ "source": [
+ "import operator\n",
+ "\n",
+ "class ReportState(TypedDict):\n",
+ " topic: str # Report topic\n",
+ " tavily_topic: Literal[\"general\", \"news\"] # Tavily search topic\n",
+ " tavily_days: Optional[int] # Only applicable for news topic\n",
+ " report_structure: str # Report structure\n",
+ " number_of_queries: int # Number web search queries to perform per section\n",
+ " sections: list[Section] # List of report sections\n",
+ " completed_sections: Annotated[list, operator.add] # Send() API key\n",
+ " report_sections_from_research: str # String of any completed sections from research to write final sections\n",
+ " final_report: str # Final report"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "6FL6UeTkfKlA"
+ },
+ "source": [
+ "Next you will write the report planner instructions, and a function that will generate the report sections."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "metadata": {
+ "id": "-37GDqTMfKlA"
+ },
+ "outputs": [],
+ "source": [
+ "from langchain_core.messages import HumanMessage, SystemMessage\n",
+ "\n",
+ "# Prompt to generate a search query to help with planning the report outline\n",
+ "report_planner_query_writer_instructions=\"\"\"You are an expert technical writer, helping to plan a report.\n",
+ "\n",
+ "The report will be focused on the following topic:\n",
+ "\n",
+ "{topic}\n",
+ "\n",
+ "The report structure will follow these guidelines:\n",
+ "\n",
+ "{report_organization}\n",
+ "\n",
+ "Your goal is to generate {number_of_queries} search queries that will help gather comprehensive information for planning the report sections.\n",
+ "\n",
+ "The query should:\n",
+ "\n",
+ "1. Be related to the topic\n",
+ "2. Help satisfy the requirements specified in the report organization\n",
+ "\n",
+ "Make the query specific enough to find high-quality, relevant sources while covering the breadth needed for the report structure.\"\"\"\n",
+ "\n",
+ "# Prompt generating the report outline\n",
+ "report_planner_instructions=\"\"\"You are an expert technical writer, helping to plan a report.\n",
+ "\n",
+ "Your goal is to generate the outline of the sections of the report.\n",
+ "\n",
+ "The overall topic of the report is:\n",
+ "\n",
+ "{topic}\n",
+ "\n",
+ "The report should follow this organization:\n",
+ "\n",
+ "{report_organization}\n",
+ "\n",
+ "You should reflect on this information to plan the sections of the report:\n",
+ "\n",
+ "{context}\n",
+ "\n",
+ "Now, generate the sections of the report. Each section should have the following fields:\n",
+ "\n",
+ "- Name - Name for this section of the report.\n",
+ "- Description - Brief overview of the main topics and concepts to be covered in this section.\n",
+ "- Research - Whether to perform web research for this section of the report.\n",
+ "- Content - The content of the section, which you will leave blank for now.\n",
+ "\n",
+ "Consider which sections require web research. For example, introduction and conclusion will not require research because they will distill information from other parts of the report.\"\"\"\n",
+ "\n",
+ "\n",
+ "def invoke_structured_llm_with_retry(structured_llm, queries, max_attempts=3):\n",
+ " \"\"\"\n",
+ " Not all LLMs support structured generation.\n",
+ " Retry max_attempts to get a result back\n",
+ " \"\"\"\n",
+ " for _ in range(max_attempts):\n",
+ " results = structured_llm.invoke(queries)\n",
+ " if results:\n",
+ " return results\n",
+ " return results\n",
+ "\n",
+ "async def generate_report_plan(state: ReportState):\n",
+ "\n",
+ " # Inputs\n",
+ " topic = state[\"topic\"]\n",
+ " report_structure = state[\"report_structure\"]\n",
+ " number_of_queries = state[\"number_of_queries\"]\n",
+ " tavily_topic = state[\"tavily_topic\"]\n",
+ " tavily_days = state.get(\"tavily_days\", None)\n",
+ "\n",
+ " # Convert JSON object to string if necessary\n",
+ " if isinstance(report_structure, dict):\n",
+ " report_structure = str(report_structure)\n",
+ "\n",
+ " # Generate search query\n",
+ " structured_llm = llm.with_structured_output(Queries)\n",
+ "\n",
+ " # Format system instructions\n",
+ " system_instructions_query = report_planner_query_writer_instructions.format(topic=topic, report_organization=report_structure, number_of_queries=number_of_queries)\n",
+ "\n",
+ " # Generate queries\n",
+ " results = invoke_structured_llm_with_retry(structured_llm,\n",
+ " [SystemMessage(content=system_instructions_query)]+[HumanMessage(content=\"Generate search queries that will help with planning the sections of the report.\")])\n",
+ "\n",
+ " # Web search\n",
+ " query_list = [query.search_query for query in results.queries]\n",
+ " search_docs = await tavily_search_async(query_list, tavily_topic, tavily_days)\n",
+ "\n",
+ " # Deduplicate and format sources\n",
+ " source_str = deduplicate_and_format_sources(search_docs, max_tokens_per_source=1000, include_raw_content=True)\n",
+ "\n",
+ " # Format system instructions\n",
+ " system_instructions_sections = report_planner_instructions.format(topic=topic, report_organization=report_structure, context=source_str)\n",
+ "\n",
+ " # Generate sections\n",
+ " structured_llm = llm.with_structured_output(Sections)\n",
+ " report_sections = invoke_structured_llm_with_retry(structured_llm,\n",
+ " [SystemMessage(content=system_instructions_sections)]+[HumanMessage(content=\"Generate the sections of the report. Your response must include a 'sections' field containing a list of sections. Each section must have: name, description, plan, research, and content fields.\")])\n",
+ "\n",
+ " return {\"sections\": report_sections.sections}"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "metadata": {
+ "id": "0LirVD2efKlA"
+ },
+ "outputs": [],
+ "source": [
+ "# Structure\n",
+ "report_structure = \"\"\"This report type focuses on comparative analysis.\n",
+ "\n",
+ "The report structure should include:\n",
+ "1. Introduction (no research needed)\n",
+ " - Brief overview of the topic area\n",
+ " - Context for the comparison\n",
+ "\n",
+ "2. Main Body Sections:\n",
+ " - One dedicated section for EACH offering being compared in the user-provided list\n",
+ " - Each section should examine:\n",
+ " - Core Features (bulleted list)\n",
+ " - Architecture & Implementation (2-3 sentences)\n",
+ " - One example use case (2-3 sentences)\n",
+ "\n",
+ "3. No Main Body Sections other than the ones dedicated to each offering in the user-provided list\n",
+ "\n",
+ "4. Conclusion with Comparison Table (no research needed)\n",
+ " - Structured comparison table that:\n",
+ " * Compares all offerings from the user-provided list across key dimensions\n",
+ " * Highlights relative strengths and weaknesses\n",
+ " - Final recommendations\"\"\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "8FcwT2KMfKlA"
+ },
+ "source": [
+ "Finally, choose the topic of your report. The default is CPU vs. GPU, but feel free to change the topic to something of your interest."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 23,
+ "metadata": {
+ "id": "yllp5CQ8fKlA"
+ },
+ "outputs": [],
+ "source": [
+ "# Topic\n",
+ "report_topic = \"Based on Legislative information from California State Legislature, help me identify key bills that will impact healthcare for Californians\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "-Y-aTIkKfKlA"
+ },
+ "source": [
+ "Let's run the agent. We set the Tavily topic to \"general\", but you can set it to \"news\" if you want Tavily to retrieve latest searches."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 25,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "AYb-7eqEfKlA",
+ "outputId": "9603217b-7c1a-4c37-bc1d-9477918042fb"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stdout",
+ "text": [
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB2200/id/2990082\n",
+ "Warning: No raw_content found for source https://health-access.org/2024-ca-health-care-consumer-legislative-scorecard/\n",
+ "Warning: No raw_content found for source https://calhospital.org/wp-content/uploads/2023/12/Report-on-Legislation-2023-FINAL.pdf\n",
+ "Warning: No raw_content found for source https://www.familydocs.org/news-new-laws-of-note-in-california/\n",
+ "==================================================\n",
+ "Name: Introduction\n",
+ "Description: Provide a brief overview of the topic area and context for the comparison of key bills that will impact healthcare for Californians\n",
+ "Research: False\n",
+ "==================================================\n",
+ "Name: AB 85 - Social Determinants of Health: Screening and Outreach\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 85\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 620 - Metabolic Disorders\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 620\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 716 - Emergency Ground Medical Transportation\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 716\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 874 - Out-of-pocket Expenses\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 874\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 907 - Coverage for PANDAS and PANS\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 907\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 1048 - Dental Benefits and Rate Review\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 1048\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 1060 - Naloxone Hydrochloride\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 1060\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 1157 - Rehabilitative and Habilitative Services: Durable Medical Equipment and Services\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 1157\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 1288 - Medication-Assisted Treatment\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 1288\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: AB 2200 - Guaranteed Health Care for All\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 2200\n",
+ "Research: True\n",
+ "==================================================\n",
+ "Name: Conclusion with Comparison Table\n",
+ "Description: Provide a structured comparison table that compares all offerings from the user-provided list across key dimensions and highlights relative strengths and weaknesses, and provide final recommendations\n",
+ "Research: False\n"
+ ]
+ }
+ ],
+ "source": [
+ "# Tavily search parameters\n",
+ "tavily_topic = \"general\"\n",
+ "tavily_days = None # Only applicable for news topic\n",
+ "\n",
+ "# Generate report plan\n",
+ "sections = await generate_report_plan({\"topic\": report_topic, \"report_structure\": report_structure, \"number_of_queries\": 2, \"tavily_topic\": tavily_topic, \"tavily_days\": tavily_days})\n",
+ "\n",
+ "# Print sections\n",
+ "for section in sections['sections']:\n",
+ " print(f\"{'='*50}\")\n",
+ " print(f\"Name: {section.name}\")\n",
+ " print(f\"Description: {section.description}\")\n",
+ " print(f\"Research: {section.research}\")"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "iQVWlTDxfKlB"
+ },
+ "source": [
+ "## Research and Writing"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "OXbFAWuPfKlB"
+ },
+ "source": [
+ "\n",
+ "Now you are ready to give the Agent details about which sections require research, and the number of queries needed. Let's First you will define the LangGraph state. Each state will have the following fields.\n",
+ "\n",
+ "Let's define the LangGraph state. Each state will have the following fields:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 27,
+ "metadata": {
+ "id": "-uiXz90OfKlB"
+ },
+ "outputs": [],
+ "source": [
+ "class SectionState(TypedDict):\n",
+ " tavily_topic: Literal[\"general\", \"news\"] # Tavily search topic\n",
+ " tavily_days: Optional[int] # Only applicable for news topic\n",
+ " number_of_queries: int # Number web search queries to perform per section\n",
+ " section: Section # Report section\n",
+ " search_queries: list[SearchQuery] # List of search queries\n",
+ " source_str: str # String of formatted source content from web search\n",
+ " report_sections_from_research: str # String of any completed sections from research to write final sections\n",
+ " completed_sections: list[Section] # Final key we duplicate in outer state for Send() API\n",
+ "\n",
+ "class SectionOutputState(TypedDict):\n",
+ " completed_sections: list[Section] # Final key we duplicate in outer state for Send() API"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "4cTZ8D9lfKlB"
+ },
+ "source": [
+ "### Write Single Section\n",
+ "Now you will define the query writer instructions and the Agent function and nodes."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 26,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 505
+ },
+ "id": "4gUksOVBfKlB",
+ "outputId": "174152fa-54a2-4405-c729-305e1116e9fb"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stderr",
+ "text": [
+ "/tmp/ipython-input-914003672.py:131: LangGraphDeprecatedSinceV05: `output` is deprecated and will be removed. Please use `output_schema` instead. Deprecated in LangGraph V0.5 to be removed in V2.0.\n",
+ " section_builder = StateGraph(SectionState, output=SectionOutputState)\n"
+ ]
+ },
+ {
+ "output_type": "display_data",
+ "data": {
+ "image/png": "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\n",
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {}
+ }
+ ],
+ "source": [
+ "from IPython.display import Image, display\n",
+ "from langgraph.graph import START, END, StateGraph\n",
+ "\n",
+ "# Query writer instructions\n",
+ "query_writer_instructions=\"\"\"Your goal is to generate targeted web search queries that will gather comprehensive information for writing a technical report section.\n",
+ "\n",
+ "Topic for this section:\n",
+ "{section_topic}\n",
+ "\n",
+ "When generating {number_of_queries} search queries, ensure they:\n",
+ "1. Cover different aspects of the topic (e.g., core features, real-world applications, technical architecture)\n",
+ "2. Include specific technical terms related to the topic\n",
+ "3. Target recent information by including year markers where relevant (e.g., \"2024\")\n",
+ "4. Look for comparisons or differentiators from similar technologies/approaches\n",
+ "5. Search for both official documentation and practical implementation examples\n",
+ "\n",
+ "Your queries should be:\n",
+ "- Specific enough to avoid generic results\n",
+ "- Technical enough to capture detailed implementation information\n",
+ "- Diverse enough to cover all aspects of the section plan\n",
+ "- Focused on authoritative sources (documentation, technical blogs, academic papers)\"\"\"\n",
+ "\n",
+ "# Section writer instructions\n",
+ "section_writer_instructions = \"\"\"You are an expert technical writer crafting one section of a technical report.\n",
+ "\n",
+ "Topic for this section:\n",
+ "{section_topic}\n",
+ "\n",
+ "Guidelines for writing:\n",
+ "\n",
+ "1. Technical Accuracy:\n",
+ "- Include specific version numbers\n",
+ "- Reference concrete metrics/benchmarks\n",
+ "- Cite official documentation\n",
+ "- Use technical terminology precisely\n",
+ "\n",
+ "2. Length and Style:\n",
+ "- Strict 150-200 word limit\n",
+ "- No marketing language\n",
+ "- Technical focus\n",
+ "- Write in simple, clear language\n",
+ "- Start with your most important insight in **bold**\n",
+ "- Use short paragraphs (2-3 sentences max)\n",
+ "\n",
+ "3. Structure:\n",
+ "- Use ## for section title (Markdown format)\n",
+ "- Only use ONE structural element IF it helps clarify your point:\n",
+ " * Either a focused table comparing 2-3 key items (using Markdown table syntax)\n",
+ " * Or a short list (3-5 items) using proper Markdown list syntax:\n",
+ " - Use `*` or `-` for unordered lists\n",
+ " - Use `1.` for ordered lists\n",
+ " - Ensure proper indentation and spacing\n",
+ "- End with ### Sources that references the below source material formatted as:\n",
+ " * List each source with title, date, and URL\n",
+ " * Format: `- Title : URL`\n",
+ "\n",
+ "3. Writing Approach:\n",
+ "- Include at least one specific example or case study\n",
+ "- Use concrete details over general statements\n",
+ "- Make every word count\n",
+ "- No preamble prior to creating the section content\n",
+ "- Focus on your single most important point\n",
+ "\n",
+ "4. Use this source material to help write the section:\n",
+ "{context}\n",
+ "\n",
+ "5. Quality Checks:\n",
+ "- Exactly 150-200 words (excluding title and sources)\n",
+ "- Careful use of only ONE structural element (table or list) and only if it helps clarify your point\n",
+ "- One specific example / case study\n",
+ "- Starts with bold insight\n",
+ "- No preamble prior to creating the section content\n",
+ "- Sources cited at end\"\"\"\n",
+ "\n",
+ "def generate_queries(state: SectionState):\n",
+ " \"\"\" Generate search queries for a section \"\"\"\n",
+ "\n",
+ " # Get state\n",
+ " number_of_queries = state[\"number_of_queries\"]\n",
+ " section = state[\"section\"]\n",
+ "\n",
+ " # Generate queries\n",
+ " structured_llm = llm.with_structured_output(Queries)\n",
+ "\n",
+ " # Format system instructions\n",
+ " system_instructions = query_writer_instructions.format(section_topic=section.description, number_of_queries=number_of_queries)\n",
+ "\n",
+ " # Generate queries\n",
+ " queries = invoke_structured_llm_with_retry(structured_llm,\n",
+ " [SystemMessage(content=system_instructions)]+[HumanMessage(content=\"Generate search queries on the provided topic.\")])\n",
+ "\n",
+ " return {\"search_queries\": queries.queries}\n",
+ "\n",
+ "async def search_web(state: SectionState):\n",
+ " \"\"\" Search the web for each query, then return a list of raw sources and a formatted string of sources.\"\"\"\n",
+ "\n",
+ " # Get state\n",
+ " search_queries = state[\"search_queries\"]\n",
+ " tavily_topic = state[\"tavily_topic\"]\n",
+ " tavily_days = state.get(\"tavily_days\", None)\n",
+ "\n",
+ " # Web search\n",
+ " query_list = [query.search_query for query in search_queries]\n",
+ " search_docs = await tavily_search_async(query_list, tavily_topic, tavily_days)\n",
+ "\n",
+ " # Deduplicate and format sources\n",
+ " source_str = deduplicate_and_format_sources(search_docs, max_tokens_per_source=5000, include_raw_content=True)\n",
+ "\n",
+ " return {\"source_str\": source_str}\n",
+ "\n",
+ "def write_section(state: SectionState):\n",
+ " \"\"\" Write a section of the report \"\"\"\n",
+ "\n",
+ " # Get state\n",
+ " section = state[\"section\"]\n",
+ " source_str = state[\"source_str\"]\n",
+ "\n",
+ " # Format system instructions\n",
+ " system_instructions = section_writer_instructions.format(section_title=section.name, section_topic=section.description, context=source_str)\n",
+ "\n",
+ " # Generate section\n",
+ " section_content = llm.invoke([SystemMessage(content=system_instructions)]+[HumanMessage(content=\"Generate a report section based on the provided sources.\")])\n",
+ "\n",
+ " # Write content to the section object\n",
+ " section.content = section_content.content\n",
+ "\n",
+ " # Write the updated section to completed sections\n",
+ " return {\"completed_sections\": [section]}\n",
+ "\n",
+ "# Add nodes and edges\n",
+ "section_builder = StateGraph(SectionState, output=SectionOutputState)\n",
+ "section_builder.add_node(\"generate_queries\", generate_queries)\n",
+ "section_builder.add_node(\"search_web\", search_web)\n",
+ "section_builder.add_node(\"write_section\", write_section)\n",
+ "\n",
+ "section_builder.add_edge(START, \"generate_queries\")\n",
+ "section_builder.add_edge(\"generate_queries\", \"search_web\")\n",
+ "section_builder.add_edge(\"search_web\", \"write_section\")\n",
+ "section_builder.add_edge(\"write_section\", END)\n",
+ "\n",
+ "# Compile\n",
+ "section_builder_graph = section_builder.compile()\n",
+ "\n",
+ "# View\n",
+ "display(Image(section_builder_graph.get_graph(xray=1).draw_mermaid_png()))"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "TJtQ7L-SfKlB"
+ },
+ "source": [
+ "### Validate Single Section\n",
+ "\n",
+ "Call on the Agent to write a single section to ensure the content is generated as expected."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 28,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "mLUwjEppfKlB",
+ "outputId": "faa0cc52-d54a-410d-b580-979dfcd53930"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stdout",
+ "text": [
+ "==================================================\n",
+ "Name: AB 85 - Social Determinants of Health: Screening and Outreach\n",
+ "Description: Examine the core features, architecture and implementation, and example use case of AB 85\n",
+ "Research: True\n",
+ "Warning: No raw_content found for source https://calmatters.digitaldemocracy.org/bills/ca_202320240ab85\n",
+ "Warning: No raw_content found for source https://www.sco.ca.gov/ard_payments_realign_fy2324_ab85.html\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB85/id/2841719\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB85/id/2814753\n"
+ ]
+ }
+ ],
+ "source": [
+ "# Test with one section\n",
+ "sections = sections['sections']\n",
+ "test_section = sections[1]\n",
+ "print(f\"{'='*50}\")\n",
+ "print(f\"Name: {test_section.name}\")\n",
+ "print(f\"Description: {test_section.description}\")\n",
+ "print(f\"Research: {test_section.research}\")\n",
+ "\n",
+ "# Run\n",
+ "report_section = await section_builder_graph.ainvoke({\"section\": test_section, \"number_of_queries\": 2, \"tavily_topic\": tavily_topic, \"tavily_days\": tavily_days})"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 29,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 306
+ },
+ "id": "ANF_2jZmfKlB",
+ "outputId": "fc6951c7-7ead-4502-ae17-63214de6f10d"
+ },
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ ""
+ ],
+ "text/markdown": "## Core Features and Implementation of AB 85\n**The core feature of AB 85 is the requirement for health plans to cover social determinants of health (SDOH) screenings**, which is a significant step towards addressing health disparities in California. According to the California Health Benefits Review Program, AB 85 would require coverage and reimbursement of SDOH screenings for 24.9 million Californians enrolled in state-regulated health insurance. The bill defines SDOH as \"the conditions under which people are born, grow, live, work, and age, including housing, food, transportation, utilities, and personal safety.\" The implementation of AB 85 would involve insurers providing primary care providers with adequate access to community health workers and informing them on how to access these workers. For example, a study by the California Health Benefits Review Program found that approximately 60% of commercial and CalPERS enrollees and 100% of Medi-Cal beneficiaries already have coverage for SDOH screenings at baseline, and AB 85 would increase this coverage to 100% of enrollees.\n### Sources\n- Title: Social determinants of health: screening and outreach: https://calmatters.digitaldemocracy.org/bills/ca_202320240ab85\n- Title: AB85 | 2023-2024 | Regular Session | Enrolled: https://legiscan.com/CA/text/AB85/id/2841719\n- Title: Analysis of California Assembly Bill 85 Social Determinants of Health: Screening and Outreach: https://www.chbrp.org/sites/default/files/bill-documents/AB85/AB%2085%20SDOH%20-%20FINAL.pdf"
+ },
+ "metadata": {},
+ "execution_count": 29
+ }
+ ],
+ "source": [
+ "from IPython.display import Markdown\n",
+ "section = report_section['completed_sections'][0]\n",
+ "Markdown(section.content)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "REqcHHpWfKlB"
+ },
+ "source": [
+ "### Write All Sections"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 30,
+ "metadata": {
+ "id": "MK7AyJ0_fKlG"
+ },
+ "outputs": [],
+ "source": [
+ "class ReportStateOutput(TypedDict):\n",
+ " final_report: str # Final report"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 31,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 987
+ },
+ "id": "4mWQoGFefKlG",
+ "outputId": "18fc56e6-bd08-4926-88df-2c529f2ece66"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stderr",
+ "text": [
+ "/tmp/ipython-input-1622562335.py:1: LangGraphDeprecatedSinceV10: Importing Send from langgraph.constants is deprecated. Please use 'from langgraph.types import Send' instead. Deprecated in LangGraph V1.0 to be removed in V2.0.\n",
+ " from langgraph.constants import Send\n",
+ "/tmp/ipython-input-1622562335.py:120: LangGraphDeprecatedSinceV05: `output` is deprecated and will be removed. Please use `output_schema` instead. Deprecated in LangGraph V0.5 to be removed in V2.0.\n",
+ " builder = StateGraph(ReportState, output=ReportStateOutput)\n"
+ ]
+ },
+ {
+ "output_type": "display_data",
+ "data": {
+ "image/png": "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\n",
+ "text/plain": [
+ ""
+ ]
+ },
+ "metadata": {}
+ }
+ ],
+ "source": [
+ "from langgraph.constants import Send\n",
+ "\n",
+ "final_section_writer_instructions=\"\"\"You are an expert technical writer crafting a section that synthesizes information from the rest of the report.\n",
+ "\n",
+ "Section to write:\n",
+ "{section_topic}\n",
+ "\n",
+ "Available report content:\n",
+ "{context}\n",
+ "\n",
+ "1. Section-Specific Approach:\n",
+ "\n",
+ "For Introduction:\n",
+ "- Use # for report title (Markdown format)\n",
+ "- 50-100 word limit\n",
+ "- Write in simple and clear language\n",
+ "- Focus on the core motivation for the report in 1-2 paragraphs\n",
+ "- Use a clear narrative arc to introduce the report\n",
+ "- Include NO structural elements (no lists or tables)\n",
+ "- No sources section needed\n",
+ "\n",
+ "For Conclusion/Summary:\n",
+ "- Use ## for section title (Markdown format)\n",
+ "- 100-150 word limit\n",
+ "- For comparative reports:\n",
+ " * Must include a focused comparison table using Markdown table syntax\n",
+ " * Table should distill insights from the report\n",
+ " * Keep table entries clear and concise\n",
+ "- For non-comparative reports:\n",
+ " * Only use ONE structural element IF it helps distill the points made in the report:\n",
+ " * Either a focused table comparing items present in the report (using Markdown table syntax)\n",
+ " * Or a short list using proper Markdown list syntax:\n",
+ " - Use `*` or `-` for unordered lists\n",
+ " - Use `1.` for ordered lists\n",
+ " - Ensure proper indentation and spacing\n",
+ "- End with specific next steps or implications\n",
+ "- No sources section needed\n",
+ "\n",
+ "3. Writing Approach:\n",
+ "- Use concrete details over general statements\n",
+ "- Make every word count\n",
+ "- Focus on your single most important point\n",
+ "\n",
+ "4. Quality Checks:\n",
+ "- For introduction: 50-100 word limit, # for report title, no structural elements, no sources section\n",
+ "- For conclusion: 100-150 word limit, ## for section title, only ONE structural element at most, no sources section\n",
+ "- Markdown format\n",
+ "- Do not include word count or any preamble in your response\"\"\"\n",
+ "\n",
+ "def initiate_section_writing(state: ReportState):\n",
+ " \"\"\" This is the \"map\" step when we kick off web research for some sections of the report \"\"\"\n",
+ "\n",
+ " # Kick off section writing in parallel via Send() API for any sections that require research\n",
+ " return [\n",
+ " Send(\"build_section_with_web_research\", {\"section\": s,\n",
+ " \"number_of_queries\": state[\"number_of_queries\"],\n",
+ " \"tavily_topic\": state[\"tavily_topic\"],\n",
+ " \"tavily_days\": state.get(\"tavily_days\", None)})\n",
+ " for s in state[\"sections\"]\n",
+ " if s.research\n",
+ " ]\n",
+ "\n",
+ "def write_final_sections(state: SectionState):\n",
+ " \"\"\" Write final sections of the report, which do not require web search and use the completed sections as context \"\"\"\n",
+ "\n",
+ " # Get state\n",
+ " section = state[\"section\"]\n",
+ " completed_report_sections = state[\"report_sections_from_research\"]\n",
+ "\n",
+ " # Format system instructions\n",
+ " system_instructions = final_section_writer_instructions.format(section_title=section.name, section_topic=section.description, context=completed_report_sections)\n",
+ "\n",
+ " # Generate section\n",
+ " section_content = llm.invoke([SystemMessage(content=system_instructions)]+[HumanMessage(content=\"Generate a report section based on the provided sources.\")])\n",
+ "\n",
+ " # Write content to section\n",
+ " section.content = section_content.content\n",
+ "\n",
+ " # Write the updated section to completed sections\n",
+ " return {\"completed_sections\": [section]}\n",
+ "\n",
+ "def gather_completed_sections(state: ReportState):\n",
+ " \"\"\" Gather completed sections from research \"\"\"\n",
+ "\n",
+ " # List of completed sections\n",
+ " completed_sections = state[\"completed_sections\"]\n",
+ "\n",
+ " # Format completed section to str to use as context for final sections\n",
+ " completed_report_sections = format_sections(completed_sections)\n",
+ "\n",
+ " return {\"report_sections_from_research\": completed_report_sections}\n",
+ "\n",
+ "def initiate_final_section_writing(state: ReportState):\n",
+ " \"\"\" This is the \"map\" step when we kick off research on any sections that require it using the Send API \"\"\"\n",
+ "\n",
+ " # Kick off section writing in parallel via Send() API for any sections that do not require research\n",
+ " return [\n",
+ " Send(\"write_final_sections\", {\"section\": s, \"report_sections_from_research\": state[\"report_sections_from_research\"]})\n",
+ " for s in state[\"sections\"]\n",
+ " if not s.research\n",
+ " ]\n",
+ "\n",
+ "def compile_final_report(state: ReportState):\n",
+ " \"\"\" Compile the final report \"\"\"\n",
+ "\n",
+ " # Get sections\n",
+ " sections = state[\"sections\"]\n",
+ " completed_sections = {s.name: s.content for s in state[\"completed_sections\"]}\n",
+ "\n",
+ " # Update sections with completed content while maintaining original order\n",
+ " for section in sections:\n",
+ " section.content = completed_sections[section.name]\n",
+ "\n",
+ " # Compile final report\n",
+ " all_sections = \"\\n\\n\".join([s.content for s in sections])\n",
+ "\n",
+ " return {\"final_report\": all_sections}\n",
+ "\n",
+ "# Add nodes and edges\n",
+ "builder = StateGraph(ReportState, output=ReportStateOutput)\n",
+ "builder.add_node(\"generate_report_plan\", generate_report_plan)\n",
+ "builder.add_node(\"build_section_with_web_research\", section_builder.compile())\n",
+ "builder.add_node(\"gather_completed_sections\", gather_completed_sections)\n",
+ "builder.add_node(\"write_final_sections\", write_final_sections)\n",
+ "builder.add_node(\"compile_final_report\", compile_final_report)\n",
+ "builder.add_edge(START, \"generate_report_plan\")\n",
+ "builder.add_conditional_edges(\"generate_report_plan\", initiate_section_writing, [\"build_section_with_web_research\"])\n",
+ "builder.add_edge(\"build_section_with_web_research\", \"gather_completed_sections\")\n",
+ "builder.add_conditional_edges(\"gather_completed_sections\", initiate_final_section_writing, [\"write_final_sections\"])\n",
+ "builder.add_edge(\"write_final_sections\", \"compile_final_report\")\n",
+ "builder.add_edge(\"compile_final_report\", END)\n",
+ "\n",
+ "graph = builder.compile()\n",
+ "display(Image(graph.get_graph(xray=1).draw_mermaid_png()))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 32,
+ "metadata": {
+ "id": "6en4ihOafKlG"
+ },
+ "outputs": [],
+ "source": [
+ "# Structure\n",
+ "report_structure = \"\"\"This report type focuses on comparative analysis.\n",
+ "\n",
+ "The report structure should include:\n",
+ "1. Introduction (no research needed)\n",
+ " - Brief overview of the topic area\n",
+ " - Context for the comparison\n",
+ "\n",
+ "2. Main Body Sections:\n",
+ " - One dedicated section for EACH offering being compared in the user-provided list\n",
+ " - Each section should examine:\n",
+ " - Core Features (bulleted list)\n",
+ " - Architecture & Implementation (2-3 sentences)\n",
+ " - One example use case (2-3 sentences)\n",
+ "\n",
+ "3. No Main Body Sections other than the ones dedicated to each offering in the user-provided list\n",
+ "\n",
+ "4. Conclusion with Comparison Table (no research needed)\n",
+ " - Structured comparison table that:\n",
+ " * Compares all offerings from the user-provided list across key dimensions\n",
+ " * Highlights relative strengths and weaknesses\n",
+ " - Final recommendations\"\"\"\n",
+ "\n",
+ "# Tavily search parameters\n",
+ "tavily_topic = \"general\"\n",
+ "tavily_days = None # Only applicable for news topic"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "3YpUNbvVfKlG"
+ },
+ "source": [
+ "Once again, choose the topic of your report. The default is CPU vs. GPU, but feel free to change the topic to something of your interest."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 35,
+ "metadata": {
+ "id": "pFHyyGXtfKlG"
+ },
+ "outputs": [],
+ "source": [
+ "# Topic\n",
+ "report_topic = \"Based on Legislative information from California State Legislature, help me identify key bills that will impact healthcare for Californians\""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 45,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/"
+ },
+ "id": "9XT4PdgMfKlG",
+ "outputId": "d4a61878-c394-457a-c10a-bb613aaad8f2"
+ },
+ "outputs": [
+ {
+ "output_type": "stream",
+ "name": "stdout",
+ "text": [
+ "Warning: No raw_content found for source https://calhospital.org/wp-content/uploads/2023/12/Report-on-Legislation-2023-FINAL.pdf\n",
+ "Warning: No raw_content found for source https://www.dhcs.ca.gov/Pages/CARE-ACT.aspx\n",
+ "Warning: No raw_content found for source https://caph.org/advocacy/state-policy/state-policy-2022-2023/\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/pending/senate-health-committee/id/480\n",
+ "Warning: No raw_content found for source https://calhospital.org/advocacy/bill-tracker/\n",
+ "Warning: No raw_content found for source https://health-access.org/2025-26-state-health-policy-priorities/\n",
+ "Warning: No raw_content found for source https://www.calhealthplans.org/legislation/\n",
+ "Warning: No raw_content found for source https://www.chcf.org/resource/defending-medi-cal-2025/analysis-of-federal-proposals/\n",
+ "Warning: No raw_content found for source https://calmatters.digitaldemocracy.org/bills/ca_202320240ab2200\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB2200/id/2919344\n",
+ "Warning: No raw_content found for source https://www.calpers.ca.gov/documents/202411-full-agenda-item9a-01-a/download?inline\n",
+ "Warning: No raw_content found for source https://calchamberalert.com/2024/07/19/calchamber-status-update-report-on-major-legislation-for-business-19/\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB1415/id/3137100\n",
+ "Warning: No raw_content found for source https://www.polsinelli.com/ron-s-grace/publications/ab-1415-california-assembly-passed\n",
+ "Warning: No raw_content found for source https://health-access.org/ab-1415-bonta-reining-in-health-care-costs/\n",
+ "Warning: No raw_content found for source https://www.polsinelli.com/paul-a-gomez/publications/ab-1415-california-assembly-passed\n",
+ "Warning: No raw_content found for source https://legiscan.com/CA/text/AB1415/id/3260454\n",
+ "Warning: No raw_content found for source https://www.polsinelli.com/publications/california-bill-seeks-to-expand-scope-of-ohca-review\n",
+ "Warning: No raw_content found for source https://www.polsinelli.com/ashley-n-osak/publications/ab-1415-california-assembly-passed\n",
+ "Warning: No raw_content found for source https://today.westlaw.com/Document/Ia004aad029e711f08595d8f275c7741d/View/FullText.html?transitionType=Default&contextData=(sc.Default)\n",
+ "Warning: No raw_content found for source https://legiscan.com/TX/text/SB763/id/2817871\n",
+ "Warning: No raw_content found for source https://legiscan.com/TX/bill/SB763/2023\n",
+ "Warning: No raw_content found for source https://www.courthousenews.com/christianity-to-make-a-comeback-in-texas-public-schools-under-gop-proposals/\n",
+ "Warning: No raw_content found for source http://www.leginfo.ca.gov/pub/15-16/bill/sen/sb_0751-0800/sb_763_cfa_20150427_162839_sen_comm.html\n",
+ "Warning: No raw_content found for source https://www.sciencedirect.com/science/article/abs/pii/S0925838821017527\n",
+ "Warning: No raw_content found for source https://link.springer.com/article/10.1186/1556-276X-9-526\n",
+ "Warning: No raw_content found for source https://www.texaspolicyresearch.com/bills/89th-legislature-sb-763/\n"
+ ]
+ }
+ ],
+ "source": [
+ "report = await graph.ainvoke({\"topic\": report_topic,\n",
+ " \"report_structure\": report_structure,\n",
+ " \"number_of_queries\": 5,\n",
+ " \"tavily_topic\": tavily_topic,\n",
+ " \"tavily_days\": tavily_days})"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "id": "tjTrR5kUfKlG"
+ },
+ "source": [
+ "# Final Report\n",
+ "Let's validate all sections. This is the final report the Agent has written.\n",
+ "Finally, let's look at the final output."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 46,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 1000
+ },
+ "id": "ztXSbfpSfKlG",
+ "outputId": "0eb1df17-adb5-4ce2-f1f4-9b0167c30c17"
+ },
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ ""
+ ],
+ "text/markdown": "# Brief Overview of California Health Care Legislation\nThe California health care landscape is undergoing significant changes with the introduction of several key pieces of legislation. Bills such as AB 2200, AB 1415, and SB 763 aim to reform various aspects of the state's health care system, including universal coverage, oversight of health care transactions, and penalties for antitrust violations. This report provides an in-depth examination of these bills, exploring their core features, architecture, and potential impact on California's health care system.\n\n## Core Features and Architecture of AB 2200\n**The California Guaranteed Health Care for All Act, AB 2200, proposes a comprehensive framework for a single-payer health care system in California**. This system, known as CalCare, aims to provide universal health care coverage to all California residents, regardless of citizenship status. The core features of AB 2200 include comprehensive benefits, freedom of choice, no premiums or copays, and addressing health care disparities.\n\nFor example, the bill would provide coverage for a wide range of medical benefits and services, including primary care, hospitalization, prescription drugs, and long-term care. Additionally, CalCare would incorporate the health care benefits and standards of other existing federal and state provisions, such as the federal Children's Health Insurance Program and Medi-Cal.\n\nThe architecture of AB 2200 includes the creation of a CalCare Board to govern the system, which would be responsible for determining eligibility, negotiating pricing for covered pharmaceuticals and medical supplies, and establishing a prescription drug formulary. The bill would also establish an Advisory Commission on Long-Term Services and Supports to advise the board on matters related to long-term care.\n\n### Sources\n- CA AB2200 | 2023-2024 | Regular Session | Introduced: https://legiscan.com/CA/text/AB2200/id/2919344\n- Most Recent Attempt at Establishing Universal Single-Payer Healthcare in California Fails Again: https://sourceonhealthcare.org/most-recent-attempt-at-establishing-universal-single-payer-healthcare-in-california-fails-again/\n- AB 2200 CA Guaranteed Health Care for All Act | Official Website: https://a25.asmdc.org/ab-2200-ca-guaranteed-health-care-all-act\n- AB 2200 CalCare Frequently Asked Questions: https://a25.asmdc.org/sites/a25.asmdc.org/files/2024-02/AB%202200%20CalCare%20FAQ_2-7-24.pdf\n- CalChamber-Led Coalition Opposing State-Run Health Care System: https://advocacy.calchamber.com/2024/04/18/calchamber-led-coalition-opposing-state-run-health-care-system/\n- Assembly Bill Policy Committee Analysis: https://billtexts.s3.amazonaws.com/ca/ca-analysishttps-leginfo-legislature-ca-gov-faces-billAnalysisClient-xhtml-bill-id-202320240AB2885-ca-analysis-371357.pdf\n- A.B. 2200 CALIFORNIA GUARANTEED HEALTH CARE FOR ALL ACT: https://www.nationalnursesunited.org/sites/default/files/nnu/documents/0524_CA-LegFactSheet_CalCare.pdf\n- 2023 - 2024 Legislation - Assemblymember Ash Kalra: https://a25.asmdc.org/2023-2024-legislation\n- CalChamber Status Update Report on Major Legislation for Business: https://calchamberalert.com/2024/07/19/calchamber-status-update-report-on-major-legislation-for-business-19/\n- AB 2200: Guaranteed Health Care for All. | Digital Democracy: https://calmatters.digitaldemocracy.org/bills/ca_202320240ab2200\n\n## Core Features and Implementation of AB 1415\n**AB 1415 proposes significant changes to California's healthcare regulations**, aiming to expand the jurisdiction of the Office of Health Care Affordability (OHCA). The bill seeks to bring healthcare investments by private equity groups and hedge funds under OHCA's review process. Key features of AB 1415 include:\n* Expanded definition of \"health care entity\" to include management services organizations (MSOs)\n* Notification requirements for private equity groups, hedge funds, and newly created business entities involved in transactions with healthcare entities\n* Expanded definition of \"provider\" to include private and public healthcare providers, health systems, and entities that own, operate, or control a provider\nFor example, AB 1415 would require private equity groups and hedge funds to report directly to OHCA on significant healthcare transactions, allowing for a more thorough review of these transactions and their potential impact on market competition, access, affordability, quality, and the healthcare workforce.\n\n### Sources\n- Title: California: AB 1415 and Expanded OHCA Oversight\n https://www.foley.com/insights/publications/2025/03/california-ab-1415-expanded-ohca-oversight/\n- Title: AB 1415 (Bonta) \n https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB1415\n- Title: California Amends AB 1415, Bill That Expands OHCA's Transaction Review Authority\n https://www.mwe.com/insights/california-amends-ab-1415-bill-that-expands-ohcas-healthcare-transaction-review-authority/\n\n## Examine the Core Features, Architecture, and Implementation of SB 763\n**The implementation of SB 763 allows school districts to employ or accept as volunteers chaplains to provide support, services, and programs for students.** This bill, passed during the 2023 Texas legislative session, requires all school boards to vote by March 1, 2024, on whether to adopt a policy to hire or accept as volunteers chaplains. \n- Key aspects of SB 763 include:\n - Chaplains are not required to be trained in counseling nor certified by the State Board for Educator Certification.\n - They could be compensated using funds earmarked for school safety.\n - The bill does not define or establish minimum standards for chaplain programs or any form of oversight of chaplain practices.\nAn example of the impact of SB 763 can be seen in the Louise school district, which encouraged its social and emotional counselor to obtain chaplain certification after the bill became law. \n### Sources\n- Title: TX SB763 | 2023-2024 | 88th Legislature | Enrolled: https://legiscan.com/TX/text/SB763/id/2817871\n- Title: SB 763 - Texas Education Agency: https://tea.texas.gov/texas-schools/school-boards/sb-763-faq.pdf\n- Title: FACT SHEET: SB 763 (CHAPLAINS BILL): https://www.aclutx.org/sites/default/files/sb_763_factsheet_twopager.pdf\n\n# Comparative Analysis of Key Bills\nThe following section presents a structured comparison of AB 2200, AB 1415, and SB 763, highlighting their relative strengths and weaknesses.\n\n## Comparison Table and Recommendations\n| Bill | Key Features | Strengths | Weaknesses |\n| --- | --- | --- | --- |\n| AB 2200 | Single-payer health care system, comprehensive benefits, no premiums or copays | Universal coverage, reduced administrative costs | High implementation costs, potential for increased taxes |\n| AB 1415 | Expanded oversight on health care transactions, notification requirements for private equity groups | Increased transparency, improved market competition | Potential for increased regulatory burden, complexity |\n| SB 763 | Increased penalties for antitrust violations, chaplain programs in schools | Enhanced enforcement, support for students | Limited scope, potential for inconsistent implementation |\n\nBased on this comparison, we recommend prioritizing AB 2200 for its potential to provide universal health care coverage, despite its high implementation costs. AB 1415 is also a valuable proposal, as it promotes transparency and competition in health care transactions. SB 763, while limited in scope, can still contribute to enhanced enforcement and support for students. Next steps include further analysis of the implementation costs and regulatory implications of these bills."
+ },
+ "metadata": {},
+ "execution_count": 46
+ }
+ ],
+ "source": [
+ "from IPython.display import Markdown\n",
+ "Markdown(report['final_report'])"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 47,
+ "metadata": {
+ "colab": {
+ "base_uri": "https://localhost:8080/",
+ "height": 125
+ },
+ "id": "W8W2xTOUfKlG",
+ "outputId": "d4609488-0165-471b-aed5-245127372f96"
+ },
+ "outputs": [
+ {
+ "output_type": "execute_result",
+ "data": {
+ "text/plain": [
+ "'# Brief Overview of California Health Care Legislation\\nThe California health care landscape is undergoing significant changes with the introduction of several key pieces of legislation. Bills such as AB 2200, AB 1415, and SB 763 aim to reform various aspects of the state\\'s health care system, including universal coverage, oversight of health care transactions, and penalties for antitrust violations. This report provides an in-depth examination of these bills, exploring their core features, architecture, and potential impact on California\\'s health care system.\\n\\n## Core Features and Architecture of AB 2200\\n**The California Guaranteed Health Care for All Act, AB 2200, proposes a comprehensive framework for a single-payer health care system in California**. This system, known as CalCare, aims to provide universal health care coverage to all California residents, regardless of citizenship status. The core features of AB 2200 include comprehensive benefits, freedom of choice, no premiums or copays, and addressing health care disparities.\\n\\nFor example, the bill would provide coverage for a wide range of medical benefits and services, including primary care, hospitalization, prescription drugs, and long-term care. Additionally, CalCare would incorporate the health care benefits and standards of other existing federal and state provisions, such as the federal Children\\'s Health Insurance Program and Medi-Cal.\\n\\nThe architecture of AB 2200 includes the creation of a CalCare Board to govern the system, which would be responsible for determining eligibility, negotiating pricing for covered pharmaceuticals and medical supplies, and establishing a prescription drug formulary. The bill would also establish an Advisory Commission on Long-Term Services and Supports to advise the board on matters related to long-term care.\\n\\n### Sources\\n- CA AB2200 | 2023-2024 | Regular Session | Introduced: https://legiscan.com/CA/text/AB2200/id/2919344\\n- Most Recent Attempt at Establishing Universal Single-Payer Healthcare in California Fails Again: https://sourceonhealthcare.org/most-recent-attempt-at-establishing-universal-single-payer-healthcare-in-california-fails-again/\\n- AB 2200 CA Guaranteed Health Care for All Act | Official Website: https://a25.asmdc.org/ab-2200-ca-guaranteed-health-care-all-act\\n- AB 2200 CalCare Frequently Asked Questions: https://a25.asmdc.org/sites/a25.asmdc.org/files/2024-02/AB%202200%20CalCare%20FAQ_2-7-24.pdf\\n- CalChamber-Led Coalition Opposing State-Run Health Care System: https://advocacy.calchamber.com/2024/04/18/calchamber-led-coalition-opposing-state-run-health-care-system/\\n- Assembly Bill Policy Committee Analysis: https://billtexts.s3.amazonaws.com/ca/ca-analysishttps-leginfo-legislature-ca-gov-faces-billAnalysisClient-xhtml-bill-id-202320240AB2885-ca-analysis-371357.pdf\\n- A.B. 2200 CALIFORNIA GUARANTEED HEALTH CARE FOR ALL ACT: https://www.nationalnursesunited.org/sites/default/files/nnu/documents/0524_CA-LegFactSheet_CalCare.pdf\\n- 2023 - 2024 Legislation - Assemblymember Ash Kalra: https://a25.asmdc.org/2023-2024-legislation\\n- CalChamber Status Update Report on Major Legislation for Business: https://calchamberalert.com/2024/07/19/calchamber-status-update-report-on-major-legislation-for-business-19/\\n- AB 2200: Guaranteed Health Care for All. | Digital Democracy: https://calmatters.digitaldemocracy.org/bills/ca_202320240ab2200\\n\\n## Core Features and Implementation of AB 1415\\n**AB 1415 proposes significant changes to California\\'s healthcare regulations**, aiming to expand the jurisdiction of the Office of Health Care Affordability (OHCA). The bill seeks to bring healthcare investments by private equity groups and hedge funds under OHCA\\'s review process. Key features of AB 1415 include:\\n* Expanded definition of \"health care entity\" to include management services organizations (MSOs)\\n* Notification requirements for private equity groups, hedge funds, and newly created business entities involved in transactions with healthcare entities\\n* Expanded definition of \"provider\" to include private and public healthcare providers, health systems, and entities that own, operate, or control a provider\\nFor example, AB 1415 would require private equity groups and hedge funds to report directly to OHCA on significant healthcare transactions, allowing for a more thorough review of these transactions and their potential impact on market competition, access, affordability, quality, and the healthcare workforce.\\n\\n### Sources\\n- Title: California: AB 1415 and Expanded OHCA Oversight\\n https://www.foley.com/insights/publications/2025/03/california-ab-1415-expanded-ohca-oversight/\\n- Title: AB 1415 (Bonta) \\n https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB1415\\n- Title: California Amends AB 1415, Bill That Expands OHCA\\'s Transaction Review Authority\\n https://www.mwe.com/insights/california-amends-ab-1415-bill-that-expands-ohcas-healthcare-transaction-review-authority/\\n\\n## Examine the Core Features, Architecture, and Implementation of SB 763\\n**The implementation of SB 763 allows school districts to employ or accept as volunteers chaplains to provide support, services, and programs for students.** This bill, passed during the 2023 Texas legislative session, requires all school boards to vote by March 1, 2024, on whether to adopt a policy to hire or accept as volunteers chaplains. \\n- Key aspects of SB 763 include:\\n - Chaplains are not required to be trained in counseling nor certified by the State Board for Educator Certification.\\n - They could be compensated using funds earmarked for school safety.\\n - The bill does not define or establish minimum standards for chaplain programs or any form of oversight of chaplain practices.\\nAn example of the impact of SB 763 can be seen in the Louise school district, which encouraged its social and emotional counselor to obtain chaplain certification after the bill became law. \\n### Sources\\n- Title: TX SB763 | 2023-2024 | 88th Legislature | Enrolled: https://legiscan.com/TX/text/SB763/id/2817871\\n- Title: SB 763 - Texas Education Agency: https://tea.texas.gov/texas-schools/school-boards/sb-763-faq.pdf\\n- Title: FACT SHEET: SB 763 (CHAPLAINS BILL): https://www.aclutx.org/sites/default/files/sb_763_factsheet_twopager.pdf\\n\\n# Comparative Analysis of Key Bills\\nThe following section presents a structured comparison of AB 2200, AB 1415, and SB 763, highlighting their relative strengths and weaknesses.\\n\\n## Comparison Table and Recommendations\\n| Bill | Key Features | Strengths | Weaknesses |\\n| --- | --- | --- | --- |\\n| AB 2200 | Single-payer health care system, comprehensive benefits, no premiums or copays | Universal coverage, reduced administrative costs | High implementation costs, potential for increased taxes |\\n| AB 1415 | Expanded oversight on health care transactions, notification requirements for private equity groups | Increased transparency, improved market competition | Potential for increased regulatory burden, complexity |\\n| SB 763 | Increased penalties for antitrust violations, chaplain programs in schools | Enhanced enforcement, support for students | Limited scope, potential for inconsistent implementation |\\n\\nBased on this comparison, we recommend prioritizing AB 2200 for its potential to provide universal health care coverage, despite its high implementation costs. AB 1415 is also a valuable proposal, as it promotes transparency and competition in health care transactions. SB 763, while limited in scope, can still contribute to enhanced enforcement and support for students. Next steps include further analysis of the implementation costs and regulatory implications of these bills.'"
+ ],
+ "application/vnd.google.colaboratory.intrinsic+json": {
+ "type": "string"
+ }
+ },
+ "metadata": {},
+ "execution_count": 47
+ }
+ ],
+ "source": [
+ "#print the final report in plain text\n",
+ "report['final_report']"
+ ]
+ }
+ ],
+ "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": "3.10.13"
+ },
+ "colab": {
+ "provenance": []
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 0
+}
\ No newline at end of file