Skip to content

Commit cb8838c

Browse files
gustavocidornelaswhoseoyster
gustavocidornelas
authored and
whoseoyster
committed
docs: add LangGraph notebook example
1 parent b89396c commit cb8838c

File tree

1 file changed

+390
-0
lines changed

1 file changed

+390
-0
lines changed

‎examples/tracing/langgraph/langgraph_tracing.ipynb

Lines changed: 390 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,390 @@
1+
{
2+
"cells": [
3+
{
4+
"cell_type": "markdown",
5+
"id": "2722b419",
6+
"metadata": {},
7+
"source": [
8+
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/openlayer-ai/openlayer-python/blob/main/examples/tracing/langgraph/langgraph_tracing.ipynb)\n",
9+
"\n",
10+
"\n",
11+
"# <a id=\"top\">LangGraph tracing</a>\n",
12+
"\n",
13+
"This notebook illustrates how use Openlayer's callback handler to monitor LangGraph workflows."
14+
]
15+
},
16+
{
17+
"cell_type": "markdown",
18+
"id": "75c2a473",
19+
"metadata": {},
20+
"source": [
21+
"## 1. Set the environment variables"
22+
]
23+
},
24+
{
25+
"cell_type": "code",
26+
"execution_count": null,
27+
"id": "f3f4fa13",
28+
"metadata": {},
29+
"outputs": [],
30+
"source": [
31+
"import os\n",
32+
"\n",
33+
"# OpenAI env variables\n",
34+
"os.environ[\"OPENAI_API_KEY\"] = \"YOUR_OPENAI_API_KEY_HERE\"\n",
35+
"\n",
36+
"# Openlayer env variables\n",
37+
"os.environ[\"OPENLAYER_API_KEY\"] = \"YOUR_OPENLAYER_API_KEY_HERE\"\n",
38+
"os.environ[\"OPENLAYER_INFERENCE_PIPELINE_ID\"] = \"YOUR_OPENLAYER_INFERENCE_PIPELINE_ID_HERE\""
39+
]
40+
},
41+
{
42+
"cell_type": "markdown",
43+
"id": "9758533f",
44+
"metadata": {},
45+
"source": [
46+
"## 2. Instantiate the `OpenlayerHandler`"
47+
]
48+
},
49+
{
50+
"cell_type": "code",
51+
"execution_count": null,
52+
"id": "e60584fa",
53+
"metadata": {},
54+
"outputs": [],
55+
"source": [
56+
"from openlayer.lib.integrations import langchain_callback\n",
57+
"\n",
58+
"openlayer_handler = langchain_callback.OpenlayerHandler()"
59+
]
60+
},
61+
{
62+
"cell_type": "markdown",
63+
"id": "72a6b954",
64+
"metadata": {},
65+
"source": [
66+
"## 3. Use LangGraph \n",
67+
"\n",
68+
"### 3.1 Simple chatbot example"
69+
]
70+
},
71+
{
72+
"cell_type": "markdown",
73+
"id": "76a350b4",
74+
"metadata": {},
75+
"source": [
76+
"We can start with a simple chatbot example similar to the one in the [LangGraph quickstart](https://langchain-ai.github.io/langgraph/tutorials/get-started/1-build-basic-chatbot/).\n",
77+
"\n",
78+
"The idea is passing the `openlayer_handler` as a callback to the LangGraph graph. After running the graph,\n",
79+
"you'll be able to see the traces in the Openlayer platform."
80+
]
81+
},
82+
{
83+
"cell_type": "code",
84+
"execution_count": null,
85+
"id": "cc351618",
86+
"metadata": {},
87+
"outputs": [],
88+
"source": [
89+
"from typing import Annotated\n",
90+
"from typing_extensions import TypedDict\n",
91+
"\n",
92+
"from langgraph.graph import StateGraph\n",
93+
"from langchain_openai import ChatOpenAI\n",
94+
"from langchain_core.messages import HumanMessage\n",
95+
"from langgraph.graph.message import add_messages"
96+
]
97+
},
98+
{
99+
"cell_type": "code",
100+
"execution_count": null,
101+
"id": "4595c63b",
102+
"metadata": {},
103+
"outputs": [],
104+
"source": [
105+
"class State(TypedDict):\n",
106+
" # Messages have the type \"list\". The `add_messages` function in the annotation defines how this state key should be updated\n",
107+
" # (in this case, it appends messages to the list, rather than overwriting them)\n",
108+
" messages: Annotated[list, add_messages]\n",
109+
"\n",
110+
"graph_builder = StateGraph(State)\n",
111+
"\n",
112+
"llm = ChatOpenAI(model = \"gpt-4o\", temperature = 0.2)\n"
113+
]
114+
},
115+
{
116+
"cell_type": "code",
117+
"execution_count": null,
118+
"id": "00a6fa80",
119+
"metadata": {},
120+
"outputs": [],
121+
"source": [
122+
"# The chatbot node function takes the current State as input and returns an updated messages list. This is the basic pattern for all LangGraph node functions.\n",
123+
"def chatbot(state: State):\n",
124+
" return {\"messages\": [llm.invoke(state[\"messages\"])]}\n"
125+
]
126+
},
127+
{
128+
"cell_type": "code",
129+
"execution_count": null,
130+
"id": "a36e5160",
131+
"metadata": {},
132+
"outputs": [],
133+
"source": [
134+
"# Add a \"chatbot\" node. Nodes represent units of work. They are typically regular python functions.\n",
135+
"graph_builder.add_node(\"chatbot\", chatbot)\n",
136+
"\n",
137+
"# Add an entry point. This tells our graph where to start its work each time we run it.\n",
138+
"graph_builder.set_entry_point(\"chatbot\")\n",
139+
"\n",
140+
"# Set a finish point. This instructs the graph \"any time this node is run, you can exit.\"\n",
141+
"graph_builder.set_finish_point(\"chatbot\")\n",
142+
"\n",
143+
"# To be able to run our graph, call \"compile()\" on the graph builder. This creates a \"CompiledGraph\" we can use invoke on our state.\n",
144+
"graph = graph_builder.compile()"
145+
]
146+
},
147+
{
148+
"cell_type": "code",
149+
"execution_count": null,
150+
"id": "deef517e",
151+
"metadata": {},
152+
"outputs": [],
153+
"source": [
154+
"# Pass the openlayer_handler as a callback to the LangGraph graph. After running the graph,\n",
155+
"# you'll be able to see the traces in the Openlayer platform.\n",
156+
"for s in graph.stream({\"messages\": [HumanMessage(content = \"What is the meaning of life?\")]},\n",
157+
" config={\"callbacks\": [openlayer_handler]}):\n",
158+
" print(s)"
159+
]
160+
},
161+
{
162+
"cell_type": "markdown",
163+
"id": "c049c8fa",
164+
"metadata": {},
165+
"source": [
166+
"### 3.2 Multi-agent example\n",
167+
"\n",
168+
"Now, we're going to use a more complex example. The principle, however, remains the same: passing the `openlayer_handler` as a callback to the LangGraph graph."
169+
]
170+
},
171+
{
172+
"cell_type": "code",
173+
"execution_count": null,
174+
"id": "213fc402",
175+
"metadata": {},
176+
"outputs": [],
177+
"source": [
178+
"from typing import Annotated\n",
179+
"from datetime import datetime\n",
180+
"\n",
181+
"from langchain.tools import Tool\n",
182+
"from langchain_community.tools import WikipediaQueryRun\n",
183+
"from langchain_community.utilities import WikipediaAPIWrapper\n",
184+
"\n",
185+
"# Define a tools that searches Wikipedia\n",
186+
"wikipedia_tool = WikipediaQueryRun(api_wrapper=WikipediaAPIWrapper())\n",
187+
"\n",
188+
"# Define a new tool that returns the current datetime\n",
189+
"datetime_tool = Tool(\n",
190+
" name=\"Datetime\",\n",
191+
" func = lambda x: datetime.now().isoformat(),\n",
192+
" description=\"Returns the current datetime\",\n",
193+
")"
194+
]
195+
},
196+
{
197+
"cell_type": "code",
198+
"execution_count": null,
199+
"id": "c76c8935",
200+
"metadata": {},
201+
"outputs": [],
202+
"source": [
203+
"from langchain.agents import AgentExecutor, create_openai_tools_agent\n",
204+
"from langchain_openai import ChatOpenAI\n",
205+
"from langchain_core.messages import BaseMessage, HumanMessage\n",
206+
"\n",
207+
"\n",
208+
"def create_agent(llm: ChatOpenAI, system_prompt: str, tools: list):\n",
209+
" # Each worker node will be given a name and some tools.\n",
210+
" prompt = ChatPromptTemplate.from_messages(\n",
211+
" [\n",
212+
" (\n",
213+
" \"system\",\n",
214+
" system_prompt,\n",
215+
" ),\n",
216+
" MessagesPlaceholder(variable_name=\"messages\"),\n",
217+
" MessagesPlaceholder(variable_name=\"agent_scratchpad\"),\n",
218+
" ]\n",
219+
" )\n",
220+
" agent = create_openai_tools_agent(llm, tools, prompt)\n",
221+
" executor = AgentExecutor(agent=agent, tools=tools)\n",
222+
" return executor\n",
223+
"\n",
224+
"def agent_node(state, agent, name):\n",
225+
" result = agent.invoke(state)\n",
226+
" return {\"messages\": [HumanMessage(content=result[\"output\"], name=name)]}"
227+
]
228+
},
229+
{
230+
"cell_type": "code",
231+
"execution_count": null,
232+
"id": "f626e7f4",
233+
"metadata": {},
234+
"outputs": [],
235+
"source": [
236+
"from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder\n",
237+
"from langchain_core.output_parsers.openai_functions import JsonOutputFunctionsParser\n",
238+
"\n",
239+
"members = [\"Researcher\", \"CurrentTime\"]\n",
240+
"system_prompt = (\n",
241+
" \"You are a supervisor tasked with managing a conversation between the\"\n",
242+
" \" following workers: {members}. Given the following user request,\"\n",
243+
" \" respond with the worker to act next. Each worker will perform a\"\n",
244+
" \" task and respond with their results and status. When finished,\"\n",
245+
" \" respond with FINISH.\"\n",
246+
")\n",
247+
"# Our team supervisor is an LLM node. It just picks the next agent to process and decides when the work is completed\n",
248+
"options = [\"FINISH\"] + members\n",
249+
"\n",
250+
"# Using openai function calling can make output parsing easier for us\n",
251+
"function_def = {\n",
252+
" \"name\": \"route\",\n",
253+
" \"description\": \"Select the next role.\",\n",
254+
" \"parameters\": {\n",
255+
" \"title\": \"routeSchema\",\n",
256+
" \"type\": \"object\",\n",
257+
" \"properties\": {\n",
258+
" \"next\": {\n",
259+
" \"title\": \"Next\",\n",
260+
" \"anyOf\": [\n",
261+
" {\"enum\": options},\n",
262+
" ],\n",
263+
" }\n",
264+
" },\n",
265+
" \"required\": [\"next\"],\n",
266+
" },\n",
267+
"}\n",
268+
"\n",
269+
"# Create the prompt using ChatPromptTemplate\n",
270+
"prompt = ChatPromptTemplate.from_messages(\n",
271+
" [\n",
272+
" (\"system\", system_prompt),\n",
273+
" MessagesPlaceholder(variable_name=\"messages\"),\n",
274+
" (\n",
275+
" \"system\",\n",
276+
" \"Given the conversation above, who should act next?\"\n",
277+
" \" Or should we FINISH? Select one of: {options}\",\n",
278+
" ),\n",
279+
" ]\n",
280+
").partial(options=str(options), members=\", \".join(members))\n",
281+
"\n",
282+
"llm = ChatOpenAI(model=\"gpt-4o\")\n",
283+
"\n",
284+
"# Construction of the chain for the supervisor agent\n",
285+
"supervisor_chain = (\n",
286+
" prompt\n",
287+
" | llm.bind_functions(functions=[function_def], function_call=\"route\")\n",
288+
" | JsonOutputFunctionsParser()\n",
289+
")"
290+
]
291+
},
292+
{
293+
"cell_type": "code",
294+
"execution_count": null,
295+
"id": "ec307b80",
296+
"metadata": {},
297+
"outputs": [],
298+
"source": [
299+
"import operator\n",
300+
"import functools\n",
301+
"from typing import Sequence, TypedDict\n",
302+
"\n",
303+
"from langgraph.graph import END, START, StateGraph\n",
304+
"from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder\n",
305+
"\n",
306+
"\n",
307+
"# The agent state is the input to each node in the graph\n",
308+
"class AgentState(TypedDict):\n",
309+
" # The annotation tells the graph that new messages will always be added to the current states\n",
310+
" messages: Annotated[Sequence[BaseMessage], operator.add]\n",
311+
" # The 'next' field indicates where to route to next\n",
312+
" next: str\n",
313+
"\n",
314+
"# Add the research agent using the create_agent helper function\n",
315+
"research_agent = create_agent(llm, \"You are a web researcher.\", [wikipedia_tool])\n",
316+
"research_node = functools.partial(agent_node, agent=research_agent, name=\"Researcher\")\n",
317+
"\n",
318+
"# Add the time agent using the create_agent helper function\n",
319+
"currenttime_agent = create_agent(llm, \"You can tell the current time at\", [datetime_tool])\n",
320+
"currenttime_node = functools.partial(agent_node, agent=currenttime_agent, name = \"CurrentTime\")\n",
321+
"\n",
322+
"workflow = StateGraph(AgentState)\n",
323+
"\n",
324+
"# Add a \"chatbot\" node. Nodes represent units of work. They are typically regular python functions.\n",
325+
"workflow.add_node(\"Researcher\", research_node)\n",
326+
"workflow.add_node(\"CurrentTime\", currenttime_node)\n",
327+
"workflow.add_node(\"supervisor\", supervisor_chain)\n",
328+
"\n",
329+
"# We want our workers to ALWAYS \"report back\" to the supervisor when done\n",
330+
"for member in members:\n",
331+
" workflow.add_edge(member, \"supervisor\")\n",
332+
"\n",
333+
"# Conditional edges usually contain \"if\" statements to route to different nodes depending on the current graph state.\n",
334+
"# These functions receive the current graph state and return a string or list of strings indicating which node(s) to call next.\n",
335+
"conditional_map = {k: k for k in members}\n",
336+
"conditional_map[\"FINISH\"] = END\n",
337+
"workflow.add_conditional_edges(\"supervisor\", lambda x: x[\"next\"], conditional_map)\n",
338+
"\n",
339+
"# Add an entry point. This tells our graph where to start its work each time we run it.\n",
340+
"workflow.add_edge(START, \"supervisor\")\n",
341+
"\n",
342+
"# To be able to run our graph, call \"compile()\" on the graph builder. This creates a \"CompiledGraph\" we can use invoke on our state.\n",
343+
"graph_2 = workflow.compile()"
344+
]
345+
},
346+
{
347+
"cell_type": "code",
348+
"execution_count": null,
349+
"id": "08e35ae9",
350+
"metadata": {},
351+
"outputs": [],
352+
"source": [
353+
"# Pass the openlayer_handler as a callback to the LangGraph graph. After running the graph,\n",
354+
"# you'll be able to see the traces in the Openlayer platform.\n",
355+
"for s in graph_2.stream({\"messages\": [HumanMessage(content = \"How does photosynthesis work?\")]},\n",
356+
" config={\"callbacks\": [openlayer_handler]}):\n",
357+
" print(s)"
358+
]
359+
},
360+
{
361+
"cell_type": "code",
362+
"execution_count": null,
363+
"id": "16acecc2",
364+
"metadata": {},
365+
"outputs": [],
366+
"source": []
367+
}
368+
],
369+
"metadata": {
370+
"kernelspec": {
371+
"display_name": "callback-improvements",
372+
"language": "python",
373+
"name": "python3"
374+
},
375+
"language_info": {
376+
"codemirror_mode": {
377+
"name": "ipython",
378+
"version": 3
379+
},
380+
"file_extension": ".py",
381+
"mimetype": "text/x-python",
382+
"name": "python",
383+
"nbconvert_exporter": "python",
384+
"pygments_lexer": "ipython3",
385+
"version": "3.9.19"
386+
}
387+
},
388+
"nbformat": 4,
389+
"nbformat_minor": 5
390+
}

0 commit comments

Comments
 (0)
pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy