MLCon24_Solution_LangGraph_AI_Workflows.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"provenance":[],"collapsed_sections":["lJy5wsZuHvXz","XUC6k6SzJE5a"]},"kernelspec":{"name":"python3","display_name":"Python 3"},"language_info":{"name":"python"}},"cells":[{"cell_type":"markdown","source":["<img src=\"https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png\" srcset=\"https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_130 130w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_260 260w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_390 390w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_520 520w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_650 650w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_780 780w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_910 910w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_1040 1040w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_1170 1170w, https://71022.cdn.cke-cs.com/RructTCFEHceQFc13ldy/images/6dbe93b28dbb43fbc9d50623b68a675a1fedd7608af93b46.png/w_1290 1290w\" sizes=\"100vw\" width=\"1290\">\n","<p style='margin-top: 1rem; margin-bottom: 1rem;'>Developed by Marco Frodl, Principal Consultant for Generative AI @ <a href='https://go.mfr.one/tt-en' _target='blank'>Thinktecture AG</a> -- More about me on my <a href='https://go.mfr.one/marcofrodl-en' _target='blank'>profile page</a></p>"],"metadata":{"id":"Vigmlask6ys0"}},{"cell_type":"markdown","source":["## Prettify Colab Notebook outputs"],"metadata":{"id":"lJy5wsZuHvXz"}},{"cell_type":"code","source":["from IPython.display import HTML, display\n","\n","def set_css():\n","  display(HTML('''\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  '''))\n","get_ipython().events.register('pre_run_cell', set_css)"],"metadata":{"id":"xOA06-GRFTeK","executionInfo":{"status":"ok","timestamp":1732635916570,"user_tz":-60,"elapsed":303,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":1,"outputs":[]},{"cell_type":"markdown","source":["## Set API keys via Colab Secrets"],"metadata":{"id":"XUC6k6SzJE5a"}},{"cell_type":"code","source":["# import Colab Secrets userdata module\n","from google.colab import userdata\n","import os\n","\n","# set Langfuse API keys\n","os.environ[\"LANGFUSE_PUBLIC_KEY\"] = userdata.get('LANGFUSE_PUBLIC_KEY')\n","os.environ[\"LANGFUSE_SECRET_KEY\"] = userdata.get('LANGFUSE_SECRET_KEY')"],"metadata":{"id":"7Aj0mSvhyJI5","colab":{"base_uri":"https://localhost:8080/","height":0},"outputId":"eb6c4672-13d1-417a-a4bc-85283c8518e4","executionInfo":{"status":"ok","timestamp":1732635921150,"user_tz":-60,"elapsed":4241,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":2,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"markdown","source":["## Load libraries"],"metadata":{"id":"6jh11ajrHjFC"}},{"cell_type":"code","source":["!pip -q install protobuf==3.20.3 langchain==0.3.8 langchain-core==0.3.21 langchain-community==0.3.8 langchain-openai==0.2.9\n","!pip -q install langgraph==0.2.53\n","!pip -q install langfuse==2.54.1\n","!pip -q install tavily-python"],"metadata":{"id":"pfrfCqwVVG_Y","colab":{"base_uri":"https://localhost:8080/","height":312},"outputId":"91c9fed0-8b11-47cb-a0c3-3cc340b94a76","executionInfo":{"status":"ok","timestamp":1732635955799,"user_tz":-60,"elapsed":34653,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":3,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}},{"output_type":"stream","name":"stdout","text":["\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.1/1.1 MB\u001b[0m \u001b[31m15.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.0/1.0 MB\u001b[0m \u001b[31m40.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m409.5/409.5 kB\u001b[0m \u001b[31m24.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m2.4/2.4 MB\u001b[0m \u001b[31m57.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m50.4/50.4 kB\u001b[0m \u001b[31m2.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.1/3.1 MB\u001b[0m \u001b[31m63.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.2/1.2 MB\u001b[0m \u001b[31m46.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m49.5/49.5 kB\u001b[0m \u001b[31m2.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n","grpcio-status 1.62.3 requires protobuf>=4.21.6, but you have protobuf 3.20.3 which is incompatible.\u001b[0m\u001b[31m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m125.1/125.1 kB\u001b[0m \u001b[31m3.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m247.9/247.9 kB\u001b[0m \u001b[31m4.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m90.4/90.4 kB\u001b[0m \u001b[31m5.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n","jupyter-server 1.24.0 requires anyio<4,>=3.1.0, but you have anyio 4.6.2.post1 which is incompatible.\u001b[0m\u001b[31m\n","\u001b[0m"]}]},{"cell_type":"code","source":["# let's check the imported versions of the most important libraries\n","!pip show langchain\n","!pip show langchain-core\n","!pip show langchain_openai\n","!pip show langchain_community\n","!pip show langgraph\n","!pip show langfuse"],"metadata":{"id":"wsfvcpLnQyw_"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Prepare LLM"],"metadata":{"id":"PflymEKFgBBl"}},{"cell_type":"code","source":["# for this notebook we use the free OVH AI endpoints and get access to a bunch of different open source LLMs\n","# you must register for an account to get your own (free) access token\n","from langchain_openai import ChatOpenAI\n","\n","TEMPERATURE = 0\n","MAX_TOKENS = 1500\n","\n","# OVH AI Endpoints Overview: https://endpoints.ai.cloud.ovh.net/\n","\n","# OVH Mistral Nemo\n","#MODEL_NAME = \"Mistral-Nemo-Instruct-2407\"\n","#BASE_URL = \"https://mistral-nemo-instruct-2407.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Mistral-7B-0.2\n","#MODEL_NAME = \"Mistral-7B-Instruct-v0.2\"\n","#BASE_URL = \"https://mistral-7b-instruct-v02.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Mixtral 8x22B\n","#MODEL_NAME = \"Mixtral-8x22B-Instruct-v0.1\"\n","#BASE_URL = \"https://mixtral-8x22b-instruct-v01.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Llama3-70B\n","#MODEL_NAME = \"Meta-Llama-3-70B-Instruct\"\n","#BASE_URL = \"https://llama-3-70b-instruct.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Llama3-8B\n","#MODEL_NAME = \"Meta-Llama-3-8B-Instruct\"\n","#BASE_URL = \"https://llama-3-8b-instruct.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Llama3.1-70B\n","MODEL_NAME = \"Meta-Llama-3_1-70B-Instruct\"\n","BASE_URL = \"https://llama-3-1-70b-instruct.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","# OVH Mamba Codestral 7B\n","#MODEL_NAME = \"mamba-codestral-7B-v0.1\"\n","#BASE_URL = \"https://mamba-codestral-7b-v0-1.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1/\"\n","\n","# OVH CodeLlama-13B\n","#MODEL_NAME = \"CodeLlama-13b-Instruct-hf\"\n","#BASE_URL = \"https://codellama-13b-instruct-hf.endpoints.kepler.ai.cloud.ovh.net/api/openai_compat/v1\"\n","\n","llm = ChatOpenAI(model=MODEL_NAME, temperature=TEMPERATURE, max_tokens=MAX_TOKENS, base_url=BASE_URL, api_key=userdata.get('OVH_AI_ENDPOINTS_ACCESS_TOKEN'))"],"metadata":{"id":"YJ3o8h7OgBBm","colab":{"base_uri":"https://localhost:8080/","height":17},"outputId":"fbdd16f6-d73e-456f-e6ec-d2d989ffec49","executionInfo":{"status":"ok","timestamp":1732635967853,"user_tz":-60,"elapsed":12057,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":4,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"markdown","source":["## Prepare LangFuse"],"metadata":{"id":"vOP866IZG87z"}},{"cell_type":"code","source":["# prepare Langfuse as debugging and tracing framework for our Generative AI application - never develop GenAI apps without that!\n","from langfuse.callback import CallbackHandler\n","handler = CallbackHandler()"],"metadata":{"id":"gdG06ivx5xV1","colab":{"base_uri":"https://localhost:8080/","height":17},"outputId":"d8fd9fe9-45e6-4521-dffa-53a1c564d7f5","executionInfo":{"status":"ok","timestamp":1732635968570,"user_tz":-60,"elapsed":721,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":5,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"markdown","source":["## Prepare prompts and instructions"],"metadata":{"id":"aVMRD664Vnyj"}},{"cell_type":"markdown","source":["### Prompt for final answer generation"],"metadata":{"id":"QK9KxjO7O6sC"}},{"cell_type":"code","source":["# Define prompt for answer generation\n","prompt_template = \"\"\"You are an assistant for question-answering tasks at ACME GmbH.\n","Think carefully about the context.\n","Just say 'Diese Frage kann ich nicht beantworten' if there is not enough or no context given.\n","Provide an answer to the user question using only the given context.\n","Use three sentences maximum and keep the answer concise.\n","Here is the context to use to answer the question:\n","\n","{context}\n","\n","Now, review the user question:\n","\n","{question}\n","\n","Write the answer in German. Don't output an English translation.\n","\n","Answer:\"\"\""],"metadata":{"id":"zGoTLGhqbgPa","colab":{"base_uri":"https://localhost:8080/","height":17},"outputId":"d247964f-14a1-4d69-c0eb-2cd63b999d10","executionInfo":{"status":"ok","timestamp":1732635968571,"user_tz":-60,"elapsed":7,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":6,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"markdown","source":["### Prompt for topic routing at workflow start"],"metadata":{"id":"1wuZ3mpzO-zo"}},{"cell_type":"code","source":["# define prompt for routing\n","import json\n","from langchain_core.messages import HumanMessage, SystemMessage\n","\n","router_instructions = \"\"\"You are an expert at routing a user question to a devsearch or websearch.\n","The devsearch contains documents related to coding, programming, development practices, single page applications, the Angular framework.\n","If the question is related to coding or development but not specifically covered by the devsearch, still return 'devsearch'.\n","Use 'websearch' for non-coding questions.\n","Return JSON with a single key, \"datasource,\" that is 'websearch' or 'devsearch' depending on the question.\n","Output JSON only, without any introduction or explanation.\n","\n","Example output 1:\n","{\n","  \"datasource\": \"websearch\"\n","}\n","\n","Example output 2:\n","{\n","  \"datasource\": \"devsearch\"\n","}\n","\"\"\""],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"5xLVYY8sVr6d","outputId":"56b77c2c-3012-4b95-884e-1f4ca5758617","executionInfo":{"status":"ok","timestamp":1732635968571,"user_tz":-60,"elapsed":6,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":7,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"markdown","source":["### Setup Tavily search client"],"metadata":{"id":"EZ1bz8lTRBye"}},{"cell_type":"code","source":["# Tavily playground: https://app.tavily.com/playground\n","# Instantiating your TavilyClient\n","from tavily import TavilyClient\n","search_client = TavilyClient(api_key=userdata.get('TAVILY_API_KEY'))"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"TS1-t8FbRFt1","outputId":"a0de14a9-92ab-4d21-8367-cb1f300c8f54","executionInfo":{"status":"ok","timestamp":1732635969534,"user_tz":-60,"elapsed":968,"user":{"displayName":"Marco Frodl","userId":"15184785398550850138"}}},"execution_count":8,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"code","source":["# Test open web search\n","from langchain.schema import Document\n","results = search_client.search(\"Wie hoch ist der Berliner Fernsehturm?\", search_depth=\"advanced\", max_results=2)\n","\n","# List to store the generated Document objects\n","documents = []\n","\n","# Iterate over each entry in the feed\n","for entry in results[\"results\"]:\n","    # Extract the page content\n","    page_content = entry.get('content', 'No content')\n","\n","    # Extract metadata\n","    metadata = {\n","        \"title\": entry.get('title', 'No Title'),\n","        \"link\": entry.get('url', 'No Link'),\n","        \"score\": entry.get('score', '0'),\n","    }\n","\n","    # Create a Document object for this entry\n","    document = Document(page_content=page_content, metadata=metadata)\n","\n","    # Append the document to the list\n","    documents.append(document)\n","\n","# Output results\n","for doc in documents:\n","    # Extract metadata and page content\n","    title = doc.metadata.get('title', 'N/A')\n","    link = doc.metadata.get('link', 'N/A')\n","    score = doc.metadata.get('score', 'N/A')\n","    content = doc.page_content[:350]  # Limit content for readability\n","\n","    # Print in a readable format\n","    print(f\"Title: {title}\")\n","    print(f\"Link: {link}\")\n","    print(f\"Score: {score}\")\n","    print(f\"Content: {content}\")\n","    print(\"-\" * 50 )\n"],"metadata":{"id":"7k0bGO36UPkI"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# Test open web search with domain filter\n","from langchain.schema import Document\n","results = search_client.search(\"Wie viele Komponenten brauche ich mindestens für das Setup von Routing?\", search_depth=\"advanced\", max_results=2, include_domains=[\"angular.dev\"], include_raw_content=True)\n","\n","# List to store the generated Document objects\n","documents = []\n","\n","# Iterate over each entry in the feed\n","for entry in results[\"results\"]:\n","    # Extract the page content\n","    page_content = entry.get('raw_content', 'No content')\n","\n","    # Extract metadata\n","    metadata = {\n","        \"title\": entry.get('title', 'No Title'),\n","        \"link\": entry.get('url', 'No Link'),\n","        \"score\": entry.get('score', '0'),\n","    }\n","\n","    # Create a Document object for this entry\n","    document = Document(page_content=page_content, metadata=metadata)\n","\n","    # Append the document to the list\n","    documents.append(document)\n","\n","# Output results\n","for doc in documents:\n","    # Extract metadata and page content\n","    title = doc.metadata.get('title', 'N/A')\n","    link = doc.metadata.get('link', 'N/A')\n","    score = doc.metadata.get('score', 'N/A')\n","    content = doc.page_content[:80]  # Limit content for readability\n","\n","    # Print in a readable format\n","    print(f\"Title: {title}\")\n","    print(f\"Link: {link}\")\n","    print(f\"Score: {score}\")\n","    print(f\"Content: {content}\")\n","    print(\"-\" * 50 )"],"metadata":{"id":"hsAKfag0Wq8T","colab":{"base_uri":"https://localhost:8080/","height":463},"outputId":"37c3cba8-36b4-4346-adda-db8411a29be8"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Title: Common routing tasks • Angular\n","Link: https://angular.dev/guide/routing/common-router-tasks/\n","Score: 0.48486787\n","Content: Common routing tasks • Angular\n","Skip to main content\n","menu\n","close\n","\n","\n","\n","v18\n","\n","\n","K\n","\n","Docs\n","\n","--------------------------------------------------\n","Title: Routing • Overview • Angular\n","Link: https://angular.dev/guide/routing/\n","Score: 0.272218\n","Content: Angular Routing\n","\n","Routing helps you change what the user sees in a single-page ap\n","--------------------------------------------------\n"]}]},{"cell_type":"markdown","source":["## Setup Graph"],"metadata":{"id":"xbuD7l0FvrgM"}},{"cell_type":"code","source":["from typing_extensions import TypedDict\n","from typing import List\n","\n","class GraphState(TypedDict):\n","    \"\"\"\n","    Graph state is a dictionary that contains information we want to propagate to, and modify in, each graph node.\n","    \"\"\"\n","    question : str # User question\n","    generation : str # LLM generation\n","    documents : List[str] # List of retrieved documents"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"ywyPa2HLvueZ","outputId":"b3151a30-e6e9-41c4-d609-c9e252f65712"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"code","source":["from langchain_core.messages import HumanMessage\n","from langgraph.graph import START, END\n","\n","### Helper function\n","# Post-processing\n","def format_docs(docs):\n","    return \"\\n\\n\".join(doc.page_content for doc in docs)\n","\n","### Nodes\n","def web_search_angular(state):\n","    \"\"\"\n","    Run web search for Angular content on angular.dev\n","\n","    Args:\n","        state (dict): The current graph state\n","\n","    Returns:\n","        state (dict): New key added to state, documents, that contains retrieved documents\n","    \"\"\"\n","    print(\"--- WEB SEARCH ANGULAR ---\")\n","    question = state[\"question\"]\n","    # Instantiating your TavilyClient\n","    from tavily import TavilyClient\n","    search_client = TavilyClient(api_key=userdata.get('TAVILY_API_KEY'))\n","\n","    # Run open web search\n","    from langchain.schema import Document\n","    results = search_client.search(question, search_depth=\"advanced\", max_results=2, include_domains=[\"angular.dev\"], include_raw_content=True)\n","\n","    # List to store the generated Document objects\n","    documents = []\n","\n","    # Iterate over each entry in the feed\n","    for entry in results[\"results\"]:\n","        # Extract the page content: prefer raw_content, fall back to content, and use default if both are empty\n","        page_content = entry.get('raw_content') or entry.get('content') or \"No content found\"\n","\n","        # Extract metadata\n","        metadata = {\n","            \"title\": entry.get('title', 'No Title'),\n","            \"link\": entry.get('url', 'No Link'),\n","            \"score\": entry.get('score', '0'),\n","        }\n","\n","        # Create a Document object for this entry\n","        document = Document(page_content=page_content, metadata=metadata)\n","\n","        # Append the document to the list\n","        documents.append(document)\n","\n","\n","    # Write retrieved documents to documents key in graph state\n","    print(f\"> found {len(documents)} documents\")\n","    return{\"documents\": documents}\n","\n","def web_search_full(state):\n","    \"\"\"\n","    Run web search for any content for given question\n","\n","    Args:\n","        state (dict): The current graph state\n","\n","    Returns:\n","        state (dict): New key added to state, documents, that contains retrieved documents\n","    \"\"\"\n","    print(\"--- WEB SEARCH FULL ---\")\n","    question = state[\"question\"]\n","    # Instantiating your TavilyClient\n","    from tavily import TavilyClient\n","    search_client = TavilyClient(api_key=userdata.get('TAVILY_API_KEY'))\n","\n","    # Run open web search\n","    from langchain.schema import Document\n","    results = search_client.search(question, search_depth=\"advanced\", max_results=2)\n","\n","    # List to store the generated Document objects\n","    documents = []\n","\n","    # Iterate over each entry in the feed\n","    for entry in results[\"results\"]:\n","        # Extract the page content\n","        page_content = entry.get('content', 'No content')\n","\n","        # Extract metadata\n","        metadata = {\n","            \"title\": entry.get('title', 'No Title'),\n","            \"link\": entry.get('url', 'No Link'),\n","            \"score\": entry.get('score', '0'),\n","        }\n","\n","        # Create a Document object for this entry\n","        document = Document(page_content=page_content, metadata=metadata)\n","\n","        # Append the document to the list\n","        documents.append(document)\n","\n","\n","    # Write retrieved documents to documents key in state\n","    print(f\"> found {len(documents)} documents\")\n","    return{\"documents\": documents}\n","\n","def generate(state):\n","    \"\"\"\n","    Generate answer using RAG on retrieved documents\n","\n","    Args:\n","        state (dict): The current graph state\n","\n","    Returns:\n","        state (dict): New key added to state, generation, that contains LLM generation\n","    \"\"\"\n","    print(\"---GENERATE---\")\n","    question = state[\"question\"]\n","    documents = state[\"documents\"]\n","\n","    # RAG generation\n","    docs_txt = format_docs(documents)\n","    rag_prompt_formatted = prompt_template.format(context=docs_txt, question=question)\n","    generation = llm.invoke([HumanMessage(content=rag_prompt_formatted)])\n","    return {\"generation\": generation}"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"P2OUbm7lv5_o","outputId":"e7313b3c-d785-45d4-e579-44a0e19e348e"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"code","source":["### Conditional nodes\n","def route_question(state):\n","    \"\"\"\n","    Route question to web search full or websearch angular\n","\n","    Args:\n","        state (dict): The current graph state\n","\n","    Returns:\n","        str: Next node to call\n","    \"\"\"\n","\n","    print(\"--- TOPIC ROUTER ---\")\n","    route_question = llm.invoke(\n","        [SystemMessage(content=router_instructions)]\n","        + [HumanMessage(content=state[\"question\"])]\n","    )\n","    source = json.loads(route_question.content)[\"datasource\"]\n","    if source == \"websearch\":\n","        print(\"> ROUTE QUESTION TO WEB_SEARCH_FULL ---\")\n","        return \"websearch\"\n","    elif source == \"devsearch\":\n","        print(\"> ROUTE QUESTION TO WEB_SEARCH_ANGULAR ---\")\n","        return \"devsearch\""],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"iGNuYXT2ViVk","outputId":"10a9aa71-6a5b-43af-8971-a35a143213f3"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"code","source":["from langgraph.graph import StateGraph\n","from IPython.display import Image, display\n","\n","workflow = StateGraph(GraphState)\n","\n","# Define the nodes\n","workflow.add_node(\"web_search_full\", web_search_full) # full websearch\n","workflow.add_node(\"web_search_angular\", web_search_angular) # websearch angular.dev\n","workflow.add_node(\"generate\", generate) # generate\n","\n","# Define the edges\n","workflow.set_conditional_entry_point(\n","    route_question,\n","    {\n","        \"websearch\": \"web_search_full\",\n","        \"devsearch\": \"web_search_angular\",\n","    },\n",")\n","workflow.add_edge(\"web_search_full\", \"generate\")\n","workflow.add_edge(\"web_search_angular\", \"generate\")\n","workflow.add_edge(\"generate\", END)\n","\n","# Compile\n","graph = workflow.compile()\n","display(Image(graph.get_graph().draw_mermaid_png()))"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":374},"id":"K9I2Ugcz4sTq","outputId":"4c6072ec-be8a-42bf-a8eb-ff86fbeb82b4"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}},{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<IPython.core.display.Image object>"]},"metadata":{}}]},{"cell_type":"code","source":["# format result output\n","def print_graph_result(result):\n","    # Print question, generated answer, trust level, answer grade\n","    print(\"--- END OF AI-WORKFLOW ---\\n\")\n","    print(f\"Question: {result.get('question', 'N/A')}\")\n","    print(f\"Answer: {result.get('generation').content if result.get('generation') else 'N/A'}\")\n","\n","    # Print additional metadata from the generation\n","    generation = result.get('generation')\n","    if generation:\n","        response_metadata = generation.response_metadata\n","        print(\"\\nResponse Metadata:\")\n","        for key, value in response_metadata.items():\n","            print(f\"  {key}: {value}\")\n","        print()\n","\n","    # Print documents\n","    documents = result.get('documents', [])\n","    for index, doc in enumerate(documents):\n","        print(f\"Document {index + 1}:\")\n","        metadata = doc.metadata\n","        for key, value in metadata.items():\n","            print(f\"  {key}: {value}\")\n","        print(f\"  Content: {doc.page_content[:80]}...\")  # Limit content to 150 characters"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":17},"id":"AkZH23HEWBWE","outputId":"5ecfc0b3-bbef-4277-a83e-155fde063f39"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}}]},{"cell_type":"code","source":["result = graph.invoke({\"question\": \"Wie viele Komponenten brauche ich mindestens für das Setup von Routing?\"}, config={\"callbacks\": [handler]})\n","print_graph_result(result)"],"metadata":{"id":"ZSbDrhT2UWp1","colab":{"base_uri":"https://localhost:8080/","height":767},"outputId":"c791663d-ee43-49a8-fcf3-87930503a924"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}},{"output_type":"stream","name":"stdout","text":["--- TOPIC ROUTER ---\n","> ROUTE QUESTION TO WEB_SEARCH_ANGULAR ---\n","--- WEB SEARCH ANGULAR ---\n","> found 2 documents\n","---GENERATE---\n","--- END OF AI-WORKFLOW ---\n","\n","Question: Wie viele Komponenten brauche ich mindestens für das Setup von Routing?\n","Answer: Für das Setup von Routing benötigen Sie mindestens zwei Komponenten, damit die Anwendung von einer zur anderen navigieren kann.\n","\n","Response Metadata:\n","  token_usage: {'completion_tokens': 28, 'prompt_tokens': 6525, 'total_tokens': 6553, 'completion_tokens_details': None, 'prompt_tokens_details': None}\n","  model_name: Meta-Llama-3_1-70B-Instruct\n","  system_fingerprint: None\n","  finish_reason: stop\n","  logprobs: None\n","\n","Document 1:\n","  title: Common routing tasks • Angular\n","  link: https://angular.dev/guide/routing/common-router-tasks/\n","  score: 0.48486787\n","  Content: Common routing tasks • Angular\n","Skip to main content\n","menu\n","close\n","\n","\n","\n","v18\n","\n","\n","K\n","\n","Docs\n","...\n","Document 2:\n","  title: Routing • Overview • Angular\n","  link: https://angular.dev/guide/routing/\n","  score: 0.272218\n","  Content: Angular Routing\n","\n","Routing helps you change what the user sees in a single-page ap...\n"]}]},{"cell_type":"code","source":["result = graph.invoke({\"question\": \"Wie hoch ist der Berliner Fernsehturm?\"}, config={\"callbacks\": [handler]})\n","print_graph_result(result)"],"metadata":{"id":"H03_giSX-Rzj","colab":{"base_uri":"https://localhost:8080/","height":499},"outputId":"9bb88b73-8d00-49a0-b84d-f68f2ebbce80"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["\n","  <style>\n","    pre {\n","        white-space: pre-wrap;\n","    }\n","  </style>\n","  "]},"metadata":{}},{"output_type":"stream","name":"stdout","text":["--- TOPIC ROUTER ---\n","> ROUTE QUESTION TO WEB_SEARCH_FULL ---\n","--- WEB SEARCH FULL ---\n","> found 2 documents\n","---GENERATE---\n","--- END OF AI-WORKFLOW ---\n","\n","Question: Wie hoch ist der Berliner Fernsehturm?\n","Answer: Der Berliner Fernsehturm ist genau 368 Meter hoch.\n","\n","Response Metadata:\n","  token_usage: {'completion_tokens': 14, 'prompt_tokens': 616, 'total_tokens': 630, 'completion_tokens_details': None, 'prompt_tokens_details': None}\n","  model_name: Meta-Llama-3_1-70B-Instruct\n","  system_fingerprint: None\n","  finish_reason: stop\n","  logprobs: None\n","\n","Document 1:\n","  title: Wie hoch ist der Berliner Fernsehturm? Wieviele Stufen hat er?\n","  link: https://www.in-berlin-brandenburg.com/Sehenswuerdigkeiten/Fernsehturm/FAQ/wie-hoch-berliner-fernsehturm.html\n","  score: 0.99968004\n","  Content: Wie hoch ist der Fernsehturm in Berlin? Höhe des Berliner Fernsehturms Berlin Fe...\n","Document 2:\n","  title: - Der Berliner Fernsehturm: Fragen und Antworten - rbb24 Inforadio\n","  link: https://www.inforadio.de/hintergrund-material/region/der-berliner-fernsehturm--fragen-und-antworten.html\n","  score: 0.9995363\n","  Content: DER RBB Der Berliner Fernsehturm: Fragen und Antworten Wie lange wurde an dem Ge...\n"]}]}]}