Explore OpenAI vector embedding with Neo4j, LangChain, and Wikipedia

22 min readSep 18, 2023

I’m a huge Neo4j fan. Since being introduced to graph databases, I’ve always had an ear out for exploring intriguing use cases. Over my nascent journey with AI and LLMs, I’ve noticed a lot of examples using Pinecone as a vector database for Retrieval-Augmented Generation (RAG) applications — but I was aware a Neo4j database could serve as both a knowledge graph AND a vector database after an official announcement on Tuesday, August 22nd, 2023.

On Saturday, September 16th, 2023, I spent my day participating in a hackathon hosted by Fixie AI exploring possibilities with AI and/or Large Language Models (LLMs) — eager to do a deep dive into exploring how Neo4j might work with vector embedding and a demo Retrieval-Augmented Generation (RAG) project.

All code for this blog post can be found on GitHub — along with the project walkthrough guide.

Plan of attack

My goals for today’s hackathon were:

Develop an understanding of how data can be stored as a Neo4j Vector and used in a lightweight Retrieval-Augmented Generation (RAG) example application
Review the Neo4j blog post LangChain Library Adds Full Support for Neo4j Vector Index
Load source data from Wikipedia based on an example query
Process and store the results as a Neo4j Vector
Explore sample queries and approaches for working with vector embeddings in Neo4j

Initial setup

Using my trusty 2021 14" MacBook Pro, I was ready to embark on today’s journey — needing to create the following:

An OpenAI API key
A free Neo4j graph database on Neo4j Aura
A local Python project for development using VS Code

OpenAI

Chances are good if you attended this hackathon, you’re probably comfortable creating an API key for OpenAI (ChatGPT, etc.). However, in the spirit of creating a walkthrough, let’s look at the steps involved.

The real magic is to sign into your OpenAI account. From there, you can navigate to the View API Keys section of your profile:

Let’s create a new secret key for this project — “DEMO: LangChain and Neo4j Vector embedding”:

Be sure to copy this value somewhere safe. We will add it to an environment variables file in our project momentarily.

Voila. Our OpenAI key has been created.

Neo4j

I’ll use Neo4j Aura for this demo to host my graph database on their free tier.

First, we’ll need to click New Instance:

Generally speaking, you can only create one free tier instance with a Neo4j Aura account.

Let’s click Create Free Instance to get underway:

We will want to declare several configuration settings in our environment variables file momentarily, so I recommend you click Download and continue to save these settings to a text file on your machine.

Here is an example configuration file downloaded from Neo4j Aura:

👮‍♂️ Don’t worry — this is an ephemeral instance that is no longer available.

After a few moments, we’ll see our instance is ready to rock and roll:

We’ll look at our Neo4j instance later in this tutorial — more eye candy to follow. I promise.

Python project for development using VS Code

If you don’t have Python 3.11.1 or newer in your development environment, please make sure you download and install Python before continuing with this walkthrough.

For this project, I created a new GitHub repo at https://github.com/TheRobBrennan/fixie-ai-llm-hackathon-20230916

If you’re using VS Code, you can use the built-in debugging capabilities within the IDE while developing and exploring the code base. If we load the main.py file, we can add a breakpoint in the left gutter (line 23 in this example) and click Debug Python File:

Once we hit our breakpoint, we see rich debugging information in the left sidebar.

All that’s left is to update our environment variables. Be sure to open .env.sample and save it as .env with the appropriate OpenAI API key and Neo4j credentials you saved.

Let’s get to it

I would recommend you read the original inspiration for this project — the Neo4j blog post LangChain Library Adds Full Support for Neo4j Vector Index — while continuing on the tour of my hackathon adventure.

I consider myself more of a React/JavaScript developer, so revisiting and refactoring Python doesn’t tend to light me up. However, the example code in the blog post kept calling out to me as something that would be fun to modularize and refactor once I truly knew what the fuck I was doing.

Besides, once I knew what I was doing, I could incorporate that into a Next.js/React project in the future.

🐍 Explore the Python code

main.py is our main script — which imports several modules that I wound up creating during a final refactor with an assist from ChatGPT 4:

For those unfamiliar with Python, this import pattern can be used to load files within subdirectories.

For example, I have an environment_utilities.py file in the ./modules/environment directory of my project:

Our example application is relatively straightforward. We will:

Load the environment variables defined in .env
Load raw data from Wikipedia based on an initial query
Process (chunk and clean) Wikipedia data
Store chunks of Wikipedia data in Neo4j using OpenAI embeddings and a Neo4j Vector
We’ll then ask a question against our Neo4j backend to see if our data was imported as expected
Lastly, we’ll follow a simple question/answer workflow using LangChain and our Neo4j backend

Set the stage — load your environment variables

Let’s look at modules/environment/environment_utilities.py

Please take note of the following:

~4 — We have defined a dictionary of environment variables that we require to be defined for this application.
~14 — We have a load_environment_variables function that will read our .env file
~31 — We have a verify_environment_variables function that will let us know if all required environment variables have been loaded

If we haven’t loaded all of our expected environment variables, this script will complain loudly at you by throwing an exception and terminating. Sad panda 🐼

NEO4J_URI is not set!
Traceback (most recent call last):
  File "/Users/rob/repos/fixie-ai-llm-hackathon-20230916/main.py", line 7, in <module>
    from modules.neo4j.credentials import neo4j_credentials
  File "/Users/rob/repos/fixie-ai-llm-hackathon-20230916/modules/neo4j/credentials.py", line 11, in <module>
    raise ValueError("Some environment variables are missing!")
ValueError: Some environment variables are missing!

Step 1 — Load data for a user query from Wikipedia

Once we are confident all of our environment variables have been defined, we can dive into the meat of this demo. Our goal — shamelessly borrowed from the initially referenced blog post on Neo4j — starts with seeing what data we can load from Wikipedia about Leonhard Euler — credited with being the first to develop graph theory.

If we look at our main.py file, we see a function defined toward the top to load_data_from_wikipedia_and_store_openai_embeddings_in_neo4j_vector:

Load raw data from Wikipedia

The first stop in our adventure is to use the WikipediaLoader from LangChain to load raw data from Wikipedia based on the supplied query.

But wait. WTF is a LangChain Document Loader? I thought you’d never ask!

If we look at the left sidebar, we can see a metric shit ton (the technical term) of the LangChain Document Loaders we can use. For this example, we can see a Wikipedia Document Loader ready for us to use.

If we peek under the covers using some helpful debugging in VS Code, we can see that raw_documents contains:

Sexy AF. Damn.💃

Process (chunk and clean) Wikipedia data

With our raw_documents in hand, let’s see what the blog post wants us to do next:

Next, we use the tiktoken text chunking module, which uses a tokenizer made by OpenAI, to split the article into chunks with 1000 tokens.

What is the LangChain CharacterTextSplitter?

Let’s use the LangChain CharacterTextSplitter in our module:

This will generate processed_docs that we can then use for preparing to store the data in Neo4j:

Store chunks of data in Neo4j using OpenAI embeddings and a Neo4j Vector

The last stop on this leg of the adventure is to store our processed_docs in Neo4j:

We will be importing our credentials to connect to Neo4j (using the environment variables we loaded) from modules/neo4j/credentials.py:

Wait. Why is there an open_api_secret_key in the mix? We will use OpenAI to generate the vector embedding details we need, so I’ve included it in the configuration for simplicity.

Two important points here:

LangChain makes it easy to import the documents into Neo4j and index them using the newly added vector index.
Neo4j vector index is wrapped as a LangChain vector store and, therefore, follows the syntax used to interact with other vector databases.

Clear as mud?

Let’s look at LangChain Vector Stores:

Let’s take a look at modules/neo4j/vector.py and see how we will import the processed_documents into Neo4j and index them using the newly added vector index via the store_data_in_neo4j function:

Before we run our script, let’s take a peek at our clean Neo4j database

Before we execute the main.py script in its entirety, take a look at your Neo4j graph database in Neo4j Aura. Click Open in the upper right of the card displaying your instance.

This is your Neo4j Workspace. We will use this to explore, query, and interact with our graph database.

It’s showtime. Run the script 🍿

Here we go. It’s time.

# Create a new virtual environment for the project
% python3 -m venv .venv

# Activate your virtual environment
% source .venv/bin/activate
(.venv) %

# Install the packages from requirements.txt
(.venv) % pip install -r requirements.txt

# Copy the sample environment variables file to .env
(.venv) % cp .env.sample .env

# Update .env with your OpenAI API key and Neo4j credentials

# Load your environment variables (defined in ".env")
(.venv) % source .env

# Run the main script (~30 seconds or more to complete)
(.venv) % python3 main.py

## OPTIONAL: Use time to track the execution of your script
(.venv) % time python3 main.py
: : : : : : : : : : : : : : : :
python3 main.py 2.98s user 2.20s system 13% cpu 39.288 total

Once you run your script, click the refresh glyph near the last updated timestamp:

Our graph database has some data! We can see that there are twenty-one (21) Chunk nodes — with a variety of property keys available:

If this is your first adventure with Neo4j, don’t you worry. We will barely scratch the surface to see what data we’ve imported.

If we look at the Query a Neo4j database using Cypher guide, we can see Neo4j uses Cypher as the language to query the graph.

We will create a simple Cypher query that will match all nodes in the graph (referenced as the variable n) and display the results:

MATCH (n) RETURN n

This will show us something like this:

Huh? This graph view is intriguing. We have colored nodes with the Chunk label displayed at a high level. Let’s zoom in.

What happens if we click the Chunk node titled Euler’s formula?

Whoa! This node represents the chunk of text — contained in the text property — along with the OpenAI embeddings stored in embedding.

What is the LangChain OpenAIEmbedding model — besides being one of the many LangChain Text embedding models? And what do those embedding values represent?

Let’s take a look at the chunk of text that we processed — stored in the text property of the node we selected:

text: "Euler's formula, named after Leonhard Euler, is a mathematical formula in complex analysis that establishes the fundamental relationship between the trigonometric functions and the complex exponential function. Euler's formula states that for any real number x:

where e is the base of the natural logarithm, i is the imaginary unit, and cos and sin are the trigonometric functions cosine and sine respectively. This complex exponential function is sometimes denoted cis x ("cosine plus i sine"). The formula is still valid if x is a complex number, and so some authors refer to the more general complex version as Euler's formula. Euler's formula is ubiquitous in mathematics, physics, chemistry, and engineering. The physicist Richard Feynman called the equation "our jewel" and "the most remarkable formula in mathematics".When x = π, Euler's formula may be rewritten as eiπ + 1 = 0 or eiπ = -1, which is known as Euler's identity.

== History ==
In 1714, the English mathematician Roger Cotes presented a geometrical argument that can be interpreted (after correcting a misplaced factor of





            −
            1




    {\displaystyle {\sqrt {-1}}}
  ) as:
Exponentiating this equation yields Euler's formula. Note that the logarithmic statement is not universally correct for complex numbers, since a complex logarithm can have infinitely many values, differing by multiples of 2πi.
Around 1740 Leonhard Euler turned his attention to the exponential function and derived the equation named after him by comparing the series expansions of the exponential and trigonometric expressions. The formula was first published in 1748 in his foundational work Introductio in analysin infinitorum.Johann Bernoulli had found that
And since

the above equation tells us something about complex logarithms by relating natural logarithms to imaginary (complex) numbers. Bernoulli, however, did not evaluate the integral.
Bernoulli's correspondence with Euler (who also knew the above equation) shows that Bernoulli did not fully understand complex logarithms. Euler also suggested that complex logarithms can have infinitely many values.
The view of complex numbers as points in the complex plane was described about 50 years later by Caspar Wessel.

== Definitions of complex exponentiation ==

The exponential function ex for real values of x may be defined in a few different equivalent ways (see Characterizations of the exponential function). Several of these methods may be directly extended to give definitions of ez for complex values of z simply by substituting z in place of x and using the complex algebraic operations. In particular we may use any of the three following definitions, which are equivalent. From a more advanced perspective, each of these definitions may be interpreted as giving the unique analytic continuation of ex to the complex plane.

=== Differential equation definition ===
The exponential function



        f
        (
        z
        )
        =

          e

            z




    {\displaystyle f(z)=e^{z}}
   is the unique differentiable function of a complex variable for which the derivative equals the function  and

=== Power series definition ===
For complex z

Using the ratio test, it is possible to show that this power series has an infinite radius of convergence and so defines ez for all complex z.

=== Limit definition ===
For complex z

Here, n is restricted to positive integers, so there is no question about what the power with exponent n means.

== Proofs ==
Various proofs of the formula are possible."

Its corresponding embedding value is:

embedding: [0.031864188611507416, 0.020641086623072624, -0.013721482828259468, -0.017606403678655624, -0.01527806743979454, 0.0022841887548565865, -0.016573039814829826, -0.016978537663817406, 0.011393147520720959, -0.0271813552826643, 0.022197671234607697, 0.02434287965297699, -0.014872570522129536, 0.021766014397144318, -0.03238740935921669, -0.0003873065288644284, 0.017972657456994057, -0.0038129764143377542, 0.011674378998577595, 0.022433120757341385, -0.005987615790218115, 0.002248217351734638, 0.0018639764748513699, -0.009025570936501026, -0.028463248163461685, 0.00985618494451046, 0.025834061205387115, -0.03204731643199921, 0.007776379119604826, -0.032701343297958374, 0.01769796572625637, 0.0055036358535289764, -0.03960786759853363, 0.00996736902743578, -0.013786885887384415, 0.025507047772407532, -0.0005432510515674949, 0.001888502505607903, 0.016612282022833824, 0.034715745598077774, 0.042511746287345886, 0.0047776661813259125, -0.016599200665950775, -0.007331640925258398, -0.008469647727906704, 0.006330979987978935, 0.0026602542493492365, -0.04954907298088074, 0.008672395721077919, 0.013786885887384415, 0.027809221297502518, 0.010189738124608994, -0.017998818308115005, -0.017004698514938354, -0.008521969430148602, -0.028829503804445267, 0.007383963093161583, -0.011510872282087803, 0.0073708826676011086, 0.0033224564976990223, 0.0010129241272807121, -0.00021419378754217178, -0.01768488623201847, -0.007109271828085184, -0.004846339114010334, 0.004300226457417011, -0.002230231650173664, 0.005065437871962786, -0.01954232156276703, 0.013329067267477512, 0.04486624151468277, 0.016429154202342033, 0.003930701408535242, 0.004473543725907803, 0.03202115371823311, -0.013708402402698994, -0.02414667047560215, 0.005248565226793289, 0.004512785468250513, 0.013329067267477512, 0.005945104174315929, -0.013198262080550194, -0.03453261777758598, -0.0023070797324180603, 0.006628562230616808, -0.029038792476058006, -0.017384033650159836, 0.03314607962965965, -0.03340769186615944, -0.013577597215771675, 0.024408282712101936, -0.006474865600466728, -0.007612872403115034, 0.004967333748936653, 0.002352861687541008, -0.01912374421954155, -0.0029643767047673464, 0.0013595583150163293, -0.002410088898614049, -0.02393738180398941, 0.012786224484443665, 0.000947521417401731, 0.0026537140365689993, -0.014663281850516796, -0.022995583713054657, -0.0110726747661829, 0.016455315053462982, 0.0035415554884821177, 0.011595896445214748, -0.0021321275271475315, -0.0034761526621878147, 0.03882303461432457, 0.013021674938499928, -0.02856789343059063, 0.02799234911799431, -0.024395201355218887, 0.03233508765697479, -0.008731258101761341, 0.006402922794222832, -0.016651524230837822, 0.00763903371989727, -0.004751504864543676, 0.026043349876999855, -0.0015132545959204435, 0.007684815675020218, -0.0006070186500437558, -0.013093617744743824, -0.011124996468424797, -0.010706419125199318, -0.02028791233897209, 0.022825537249445915, -0.01470252312719822, 0.007397043518722057, 0.01419238280504942, -0.0016653158236294985, 0.004146530292928219, -0.01459787879139185, -0.006308089010417461, -0.01993473805487156, -0.027102872729301453, 0.03319840133190155, 0.004911741707473993, -0.021399758756160736, 0.01408773846924305, -0.0018705168040469289, 0.022838616743683815, 0.010385946370661259, 0.02455216832458973, 0.004117099102586508, 0.001006383798085153, 0.027076711878180504, 0.018260428681969643, 0.0135645167902112, -0.006958845537155867, -0.0007059402414597571, 0.016167543828487396, 0.0010987650603055954, 0.04122985154390335, -0.006736476439982653, -0.012936650775372982, 0.0008420595549978316, 0.006092260126024485, -0.02535008080303669, -0.001144547015428543, 0.00688690273091197, 0.03241357207298279, 0.015042617917060852, 0.022655490785837173, -0.015814369544386864, 0.00046599411871284246, -0.019764691591262817, 0.028253959491848946, -0.025258516892790794, 0.017174744978547096, 0.013028214685618877, -0.008816282264888287, 0.0017642374150454998, 0.004247904289513826, -0.007985668256878853, 0.010353244841098785, 0.027521450072526932, 0.016586121171712875, 0.009326422587037086, 0.009830023162066936, -0.0004148982698097825, 0.005294347181916237, 0.030843906104564667, -0.00026549401809461415, -0.0028401114977896214, -0.024408282712101936, -0.014741765335202217, 0.01930687204003334, 0.009411446750164032, -0.0013407550286501646, -0.6140527129173279, -0.026592731475830078, -0.029431208968162537, -0.002045468892902136, -0.0019228389719501138, -0.018757490441203117, 0.0014862760435789824, 0.006514107342809439, -0.028829503804445267, -0.003162219887599349, -0.01874440908432007, -0.01648147590458393, 0.003242338076233864, -0.010987650603055954, 0.00782870128750801, -0.018796730786561966, -0.012419969774782658, -0.020274832844734192, 0.0011919639073312283, -0.0025016528088599443, -0.012596556916832924, -0.004705723375082016, -0.01751483976840973, 0.025611691176891327, 0.0004713080998044461, 0.032884471118450165, 0.01671692542731762, -0.015461194328963757, 0.011412768624722958, 0.003260323777794838, -0.003535015042871237, 0.02149132266640663, 0.00845656730234623, 0.012727362103760242, 0.026043349876999855, -0.035526737570762634, -0.001187876216135919, 0.014218543656170368, 0.027024390175938606, 0.03863990679383278, -0.012550774961709976, -0.025271598249673843, 0.031497932970523834, -0.011700540781021118, -0.021229712292551994, 0.003688711440190673, 0.03293679282069206, -0.00759979197755456, -0.006517377682030201, -0.028463248163461685, -0.00011762264330172911, -0.008705097250640392, 0.0007030788692645729, -0.0047286138869822025, 0.01952924206852913, -0.019215308129787445, -0.008371544070541859, -0.017802610993385315, -0.013302906416356564, -0.011857506819069386, -0.0028842585161328316, -0.0032243523746728897, -0.05012461915612221, 0.02571633644402027, -0.01649455726146698, 0.009712298400700092, -0.04865959659218788, 0.00040794923552311957, 0.011635137721896172, 0.007776379119604826, 0.012635799124836922, 0.0251669529825449, 0.004244634415954351, -0.01731863059103489, 0.014506315812468529, 0.016834650188684464, 0.032125797122716904, -0.0044179512187838554, 0.012635799124836922, -0.007207375951111317, 0.02231539599597454, -0.011321204714477062, -0.029038792476058006, -0.006245956290513277, 0.01829967088997364, -0.01831275224685669, -0.009705758653581142, 0.0075343893840909, 0.009398365393280983, 0.016612282022833824, 0.015369631350040436, 0.011569734662771225, 0.018626684322953224, -0.033067598938941956, 0.011484711430966854, 0.007325100712478161, -0.00004792790423380211, 0.006471595726907253, 0.0059254830703139305, -0.011229640804231167, -0.013656080700457096, -0.024225154891610146, 0.019385356456041336, 0.008947087451815605, -0.021177388727664948, 0.02211918868124485, -0.00803798995912075, 0.0029006090480834246, 0.0624203234910965, -0.0602751150727272, -0.02413359098136425, -0.027416804805397987, -0.025467805564403534, 0.005506906192749739, -0.008548131212592125, -0.031681060791015625, 0.020876536145806313, 0.004326387774199247, -0.0023348757531493902, -0.04104672372341156, 0.021399758756160736, -0.005425152834504843, -0.009280640631914139, -0.03584067150950432, 0.01650763675570488, 0.023008665069937706, 0.0030036182142794132, -0.012151818722486496, -0.03945089876651764, 0.008384624496102333, 0.024225154891610146, -0.008620074018836021, 0.005696573760360479, -0.011020352132618427, 0.002773073734715581, 0.0062590367160737514, 0.011844426393508911, -0.017344791442155838, 0.008168795146048069, -0.020392557606101036, -0.028829503804445267, -0.0069784666411578655, 0.004326387774199247, -0.013813046738505363, -0.029457369819283485, -0.03950322046875954, -0.010477510280907154, 0.012485372833907604, 0.004780936054885387, 0.007717516738921404, 0.003208001609891653, 0.013656080700457096, -0.01975161023437977, 0.03675631061196327, 0.021975301206111908, 0.011327745392918587, 0.006762637756764889, -0.008391164243221283, -0.020366396754980087, -0.0059745353646576405, -0.004754775203764439, 0.03353849798440933, -0.0069196042604744434, 0.008221117779612541, -0.024617571383714676, -0.00654353853315115, -0.025075389072299004, 0.004898661281913519, -0.0018901375588029623, -0.051825087517499924, 0.0068411207757890224, 0.008665855973958969, 0.004283875692635775, 0.016416074708104134, 0.01569664478302002, 0.024996906518936157, -0.02111198753118515, -0.008914385922253132, -0.005706384312361479, -0.010562533512711525, 0.026893584057688713, -0.006703775376081467, -0.042145490646362305, 0.008116473443806171, 0.01792033575475216, 0.010536372661590576, 0.004993495065718889, -0.000909097318071872, -0.018038060516119003, 0.007717516738921404, 0.007325100712478161, 0.027312161400914192, 0.007253157906234264, -0.007436285261064768, -0.0005771786672994494, 0.011301583610475063, 0.03199499472975731, -0.0020307532977312803, 0.012805845588445663, 0.05075248330831528, 0.018378155305981636, 0.006985006853938103, 0.02070648968219757, -0.024434443563222885, 0.019359195604920387, -0.0036396593786776066, -0.003242338076233864, -0.0041628810577094555, 0.06597822904586792, 0.0019162986427545547, 0.010889546945691109, -0.015840530395507812, -0.0381690077483654, -0.01853512041270733, 0.010431728325784206, 0.03544825688004494, 0.029064953327178955, 0.015238825231790543, 0.01691313460469246, -0.002534354105591774, 0.005107949487864971, -0.006402922794222832, 0.0005469299503602087, -0.00026283704210072756, -0.006971926428377628, 0.02030099369585514, 0.01627218723297119, 0.007501687854528427, -0.0008412420284003019, -0.04057582467794418, -0.001448669470846653, 0.007442825473845005, -0.009012489579617977, 0.031471770256757736, 0.006501026917248964, 0.052217502146959305, 0.0044179512187838554, -0.0019490000559017062, 0.0382474884390831, -0.014113899320363998, -0.02290401980280876, 0.0220930278301239, -0.00763903371989727, -0.004028805531561375, -0.005882971454411745, -0.013100157491862774, 0.012727362103760242, 0.00042429991299286485, 0.021648287773132324, 0.012753523886203766, -0.013067456893622875, 0.005477475002408028, -0.011909828521311283, -0.00288752862252295, 0.01669076457619667, 0.01387844979763031, 0.0035742567852139473, 0.018456637859344482, 0.02011786587536335, 0.03516048192977905, 0.022851698100566864, -0.023898141458630562, 0.006510837469249964, -0.0017642374150454998, -0.007004627492278814, 0.0005914855282753706, -0.0004966516280546784, -0.01848279871046543, -0.024473683908581734, -0.03986947610974312, -0.009686137549579144, -0.021844496950507164, 0.0018852323992177844, -0.002957836491987109, 0.020379476249217987, -0.009104053489863873, 0.023649610579013824, 0.008848982863128185, 0.0060301274061203, 0.001947364886291325, -0.03924161195755005, -0.020680328831076622, -0.0014225083868950605, 0.024042027071118355, -0.007272778544574976, -0.019856255501508713, 0.002895703772082925, 0.010314003564417362, 0.001518977340310812, 0.01359067764133215, 0.0001279440039070323, 0.0057815974578261375, -0.018077302724123, -0.012917030602693558, -0.02129511348903179, -0.01176594290882349, 0.009077892638742924, 0.0020013221073895693, -0.00795296672731638, -0.0044441125355660915, 0.018849054351449013, -0.01752791926264763, -0.017161663621664047, 0.002944755833595991, 0.037305690348148346, -0.012197600677609444, -0.0013333972310647368, -0.00843694619834423, 0.004875770304352045, -0.013865369372069836, -0.0042707952670753, -0.029509691521525383, -0.01467636227607727, -0.016219865530729294, 0.0342710055410862, -0.006736476439982653, 0.027861542999744415, -0.013629919849336147, 0.02236771769821644, -0.02011786587536335, -0.015421953052282333, -0.0030657509341835976, 0.0022449472453445196, 0.010516751557588577, 0.04952291399240494, 0.017227066680788994, -0.016049819067120552, 0.026919744908809662, 0.014820248819887638, -0.010791443288326263, -0.016350671648979187, -0.026697376742959023, 0.021164309233427048, 0.03060845658183098, 0.0015786573057994246, -0.009659976698458195, 0.006693964824080467, -0.0025016528088599443, -0.01874440908432007, -0.004931362345814705, 0.01086992584168911, -0.011275422759354115, 0.03361697867512703, 0.0100654736161232, -0.008476188406348228, 0.014689442701637745, 0.037122562527656555, 0.017057020217180252, -0.0031278834212571383, 0.0006863194284960628, 0.0231133084744215, 0.028881825506687164, -0.010555993765592575, -0.01828659139573574, -0.01287124864757061, 0.011491252109408379, -0.00045168728684075177, -0.01054945308715105, -0.022851698100566864, 0.02351880632340908, 0.007802540436387062, 0.007213916163891554, 0.006608941126614809, -0.03434948995709419, -0.014205463230609894, 0.008123013190925121, 0.0018214647425338626, -0.00462397001683712, -0.009954288601875305, -0.030922388657927513, -0.011844426393508911, -0.006311358883976936, -0.014336268417537212, 0.00893400702625513, -0.00007557075878139585, -0.02133435569703579, -0.008319221436977386, -0.01871824823319912, -0.009679597802460194, 0.018966779112815857, 0.014951054006814957, -0.02270781248807907, -0.007436285261064768, -0.009849644266068935, -0.0201440267264843, -0.014833329245448112, 0.02173985168337822, -0.005510176066309214, 0.024081269279122353, -0.02958817407488823, -0.011530493386089802, 0.000004841331701754825, -0.0190975833684206, -0.01608906127512455, 0.019071422517299652, -0.026291878893971443, -0.03633773326873779, 0.006631832104176283, 0.03879687190055847, 0.015997497364878654, 0.005320508498698473, -0.0061151511035859585, -0.0031000871676951647, -0.0030036182142794132, -0.010778362862765789, -0.03157641738653183, -0.019385356456041336, -0.03895384073257446, -0.003930701408535242, 0.021268952637910843, 0.016573039814829826, -0.007861402817070484, 0.011419308371841908, 0.015251905657351017, 0.0014756480231881142, 0.006046478170901537, 0.007769838906824589, -0.0044146813452243805, -0.019764691591262817, 0.017763368785381317, -0.0021909899078309536, 0.01517342310398817, 0.023623449727892876, -0.019398435950279236, -0.006932684686034918, -0.0321781225502491, 0.008848982863128185, 0.008502349257469177, 0.00224004196934402, 0.0051406510174274445, 0.023087147623300552, 0.0016497827600687742, -0.01076528150588274, -0.030477650463581085, -0.009516091085970402, -0.021373597905039787, 0.04081127420067787, 0.017841853201389313, -0.0011658028233796358, 0.005873160902410746, 0.01751483976840973, 0.008208037354052067, -0.0043034967966377735, -0.0241074301302433, -0.012694661505520344, 0.0022939990740269423, 0.002241677138954401, -0.020588764920830727, -0.015212664380669594, 0.008371544070541859, -0.014440912753343582, -0.020654167979955673, -0.013852288946509361, 0.011013812385499477, 0.04544178396463394, 0.018064221367239952, 0.008404244668781757, -0.01416622195392847, -0.03283214941620827, -0.006151122506707907, -0.030451489612460136, -0.003933971282094717, -0.0015165247023105621, 0.003960132598876953, 0.02193606086075306, 0.01947692036628723, -0.010163577273488045, -0.005932023283094168, 0.004313306882977486, -0.029640497639775276, -0.016612282022833824, 0.019660046324133873, 0.012851627543568611, -0.011798644438385963, 0.0044768135994672775, -0.009372204542160034, 0.011667839251458645, -0.014741765335202217, 0.0034172902815043926, -0.016978537663817406, -0.004548756871372461, -0.0029431208968162537, 0.014048496261239052, -0.0005399809451773763, 0.01973853074014187, 0.022289235144853592, 0.0062263356521725655, 0.04340122267603874, -0.019267631694674492, 0.02372809313237667, -0.029640497639775276, 0.02074573189020157, -0.02638344280421734, 0.014257784932851791, 0.04060198739171028, 0.021582886576652527, 0.02294326201081276, -0.008443486876785755, 0.01519958395510912, 0.018064221367239952, -0.01690005324780941, 0.005307428073137999, -0.004408141132444143, -0.0502554215490818, -0.0001834340946516022, 0.023649610579013824, 0.002058549551293254, 0.008397704921662807, 0.0003603279183153063, -0.001079144305549562, -0.017344791442155838, -0.005742355715483427, 0.008822822012007236, 0.01688697375357151, 0.014741765335202217, -0.013551436364650726, 0.03594531491398811, 0.010529831983149052, 0.023885060101747513, -0.007076570298522711, 0.0014429467264562845, -0.010399026796221733, -0.019450757652521133, 0.021412838250398636, -0.020798053592443466, 0.03484655171632767, -0.002493477426469326, 0.011589355766773224, -0.013943852856755257, -0.028070831671357155, 0.00759979197755456, -0.017854932695627213, 0.00342383049428463, -0.027286000549793243, 0.0023038096260279417, -0.02274705469608307, 0.001616263878531754, -0.004869230091571808, -0.02414667047560215, 0.00021930337243247777, 0.01377380546182394, 0.008907845243811607, -0.025245435535907745, -0.013747644610702991, 0.02575557678937912, -0.014467073604464531, 0.042511746287345886, -0.00906481221318245, 0.029038792476058006, 0.033512335270643234, 0.0008960167760960758, -0.008031449280679226, -0.008417325094342232, 0.006697234697639942, 0.01226300373673439, 0.012014472857117653, -0.0018688817508518696, -0.025847140699625015, -0.01136698666960001, 0.017043938860297203, 0.005274726543575525, -0.02697206661105156, -0.002078170422464609, 0.03725336864590645, 0.0015590365510433912, 0.014650201424956322, -0.00015277658530976623, 0.016455315053462982, -0.01197523158043623, 0.007756758481264114, -0.02255084551870823, -0.004218473099172115, -0.032675180584192276, -0.002271108329296112, -0.009404906071722507, 0.03699176013469696, -0.008875144645571709, 0.033904753625392914, -0.019398435950279236, 0.013152480125427246, -0.010320543311536312, -0.02313946932554245, -0.024696053937077522, -0.0058633508160710335, -0.020196348428726196, 0.02030099369585514, -0.02574249729514122, 0.012426510453224182, 0.021857576444745064, 0.01046442985534668, -0.024813778698444366, -0.012341486290097237, 0.008482728153467178, 0.03060845658183098, -0.024996906518936157, 0.005618090741336346, 0.0062263356521725655, -0.010634476318955421, -0.022838616743683815, -0.00863969512283802, -0.022053785622119904, -0.025062309578061104, -0.0010202819248661399, 0.00003852626832667738, 0.026069510728120804, 0.01950308121740818, -0.008881684392690659, -0.0008559576235711575, -0.005516716279089451, 0.002298904350027442, -0.008319221436977386, -0.012943191453814507, -0.01692621409893036, -0.02937888540327549, -0.008848982863128185, -0.013865369372069836, 0.020183268934488297, 0.020837295800447464, -0.007253157906234264, -0.010555993765592575, -0.009934667497873306, 0.029248081147670746, -0.00713543314486742, 0.024853020906448364, 0.02452600747346878, 0.0057554361410439014, 0.0025163684040308, 0.0018361803377047181, -0.008070691488683224, -0.007050409447401762, 0.040732793509960175, -0.015539677813649178, -0.023898141458630562, -0.016625363379716873, -0.0069196042604744434, 0.00853505078703165, 0.0014625675976276398, -0.007907184772193432, -0.016769248992204666, 0.0022433120757341385, -0.0017642374150454998, -0.020392557606101036, -0.00668742461130023, -0.006762637756764889, 0.0042740656062960625, -0.005284537095576525, 0.013943852856755257, -0.012256463058292866, 0.027469128370285034, 0.015552758239209652, -0.01247883215546608, -0.01086992584168911, -0.003770464798435569, 0.004208662547171116, -0.018221188336610794, 0.009404906071722507, -0.0014306837692856789, -0.015147262252867222, -0.01973853074014187, -0.015186503529548645, -0.017541000619530678, 0.02859405428171158, 0.02235463820397854, 0.01848279871046543, -0.01297589298337698, 0.015801288187503815, 0.009110594168305397, 0.028096992522478104, 0.007272778544574976, -0.03311992064118385, 0.007436285261064768, 0.008600452914834023, -0.022236913442611694, -0.006419273559004068, 0.010588694363832474, 0.007266238331794739, 0.026304960250854492, -0.00943760760128498, -0.0019113934831693769, -0.0029725520871579647, -0.03890151530504227, 0.0020977910608053207, -0.04120369255542755, 0.02194914035499096, 0.01971236988902092, 0.009607654064893723, -0.008286519907414913, 0.008724718354642391, -0.02052336186170578, 0.019633885473012924, -0.02476145699620247, 0.002146843122318387, 0.013839208520948887, 0.0036527400370687246, -0.01993473805487156, -0.009823483414947987, -0.020000141113996506, -0.011779023334383965, 0.032492052763700485, -0.00005773830707767047, 0.0070896511897444725, 0.021399758756160736, -0.026998227462172508, -0.005624630954116583, -0.015081859193742275, 0.014310107566416264, 0.0019898766186088324, 0.027259839698672295, 0.024434443563222885, -0.04567723348736763, 0.004839798901230097, -0.02031407319009304, 0.00007424226350849494, -0.0301375575363636, 0.008763959631323814, 0.004048426169902086, -0.024617571383714676, -0.009090973064303398, 0.0034990436397492886, -0.005036006681621075, 0.013322526589035988, -0.02736448310315609, 0.03675631061196327, -0.02393738180398941, 0.03366930037736893, 0.030032912269234657, 0.010850305669009686, 0.004885580390691757, -0.014126979745924473, 0.012099497020244598, -0.029509691521525383, -0.01176594290882349, 0.025271598249673843, 0.0007754305843263865, 0.0037050622049719095, -0.031497932970523834, 0.007756758481264114, -0.010490590706467628, -0.010778362862765789, -0.03199499472975731, -0.008593913167715073, -0.01570972427725792, 0.03482038900256157, 0.0090713519603014, -0.037514980882406235, -0.007527849171310663, -0.005124300252646208, -0.005428422708064318, -0.014493235386908054, -0.011406227946281433, 0.020366396754980087, 0.008757419884204865, 0.0032635938841849566, -0.010196278803050518, -0.015421953052282333, -0.031053194776177406, 0.013760725036263466, -0.008397704921662807, -0.008214577101171017, 0.013433711603283882, 0.20813749730587006, 0.011281963437795639, 0.01377380546182394, 0.006873822305351496, -0.004872499965131283, 0.009836563840508461, -0.019254550337791443, -0.0030248742550611496, -0.02088961750268936, 0.0071223522536456585, -0.01752791926264763, -0.001199321704916656, -0.03578834980726242, 0.012217221781611443, -0.013230962678790092, -0.037358012050390244, -0.01827351003885269, -0.024839939549565315, -0.00349577353335917, -0.01035978551954031, 0.021242791786789894, -0.0036560101434588432, -0.004362359177321196, -0.007501687854528427, 0.018417395651340485, -0.003848948050290346, -0.013512195087969303, -0.016991617158055305, 0.0201309472322464, -0.004395060241222382, -0.009450688026845455, -0.016965456306934357, -0.02733832225203514, 0.034689582884311676, -0.03665166348218918, -0.0021010611671954393, 0.028698697686195374, -0.00943760760128498, 0.03311992064118385, 0.009267560206353664, 0.006170743145048618, -0.03202115371823311, 0.0016734912060201168, -0.020222509279847145, 0.03157641738653183, 0.018679006025195122, 0.006321169435977936, -0.018391234800219536, -0.011648218147456646, 0.022053785622119904, -0.03924161195755005, -0.008221117779612541, -0.0160759799182415, 0.03317224234342575, 0.021661369130015373, -0.0011911464389413595, 0.02694590575993061, 0.024669893085956573, 0.014310107566416264, 0.03654702007770538, -0.023885060101747513, 0.009705758653581142, 0.0021370328031480312, -0.005637711379677057, -0.0021844496950507164, 0.013904610648751259, -0.027495289221405983, 0.008378083817660809, -0.011648218147456646, -0.003848948050290346, -0.005742355715483427, -0.03704408183693886, -0.023466482758522034, 0.01773720793426037, -0.019149906933307648, -0.014205463230609894, 0.02576865814626217, -0.00471880380064249, 0.012289164587855339, 0.015971336513757706, -0.0008097669924609363, -0.005605010315775871, -0.016167543828487396, 0.0016841191099956632, -0.018822893500328064, -0.0201440267264843, 0.00709619140252471, 0.017357872799038887, -0.005732545163482428, 0.0004974691546522081, 0.008489268831908703, -0.004408141132444143, -0.008757419884204865, -0.0052812667563557625, 0.011667839251458645, 0.005961454939097166, 0.012348026968538761, -0.007737137842923403, -0.024264395236968994, 0.033904753625392914, -0.004712263587862253, 0.042302455753088, 0.030660778284072876, -0.008319221436977386, 0.006131501868367195, -0.016625363379716873, 0.01015703659504652, 0.01567048393189907, 0.012590017169713974, -0.027207516133785248, -0.026618892326951027, 0.010928788222372532, -0.005068708211183548, -0.024081269279122353, -0.0261741541326046, -0.010516751557588577, 0.006579509936273098, 0.018862133845686913, 0.009986990131437778, 0.006736476439982653, -0.007508228067308664, -0.023060986772179604, -0.030686939135193825, 0.010307462885975838, -0.0010137415956705809, -0.016586121171712875, -0.06472249329090118, 0.011432389728724957, -0.008829362690448761, -0.010745661333203316, 0.023649610579013824, -0.008077231235802174, 0.015356549993157387, -0.0068411207757890224, 0.005376100540161133, -0.014231624081730843, -0.024447523057460785, 0.0006209167186170816, -0.007926804944872856, 0.023047905415296555, -0.02434287965297699, -0.005869891028851271, 0.007874483242630959, -0.0034696124494075775, 0.00291532464325428, -0.004457192961126566, -0.001396347302943468, 0.016546878963708878, 0.0024656811729073524, -0.03346001356840134, -0.01348603330552578, -0.015801288187503815, -0.0010031136916950345, -0.04609581083059311, 0.00018813491624314338, -0.031105516478419304, -0.030817745253443718, -0.047351542860269547, 0.04559875279664993, -0.00643889419734478, -0.030216040089726448, 0.006196904461830854, 0.012681580148637295, -0.023492643609642982, -0.01948999986052513, -0.01005239225924015, -0.16251258552074432, 0.00044759962474927306, 0.012943191453814507, -0.009202158078551292, -0.007678274996578693, 0.013459872454404831, 0.035866834223270416, 0.0033943993039429188, 0.007573631126433611, -0.004924822133034468, 0.021622126922011375, 0.00029104194254614413, -0.023087147623300552, -0.02614799328148365, -0.006834580563008785, 0.0035415554884821177, 0.0029431208968162537, 0.01975161023437977, 0.04463079199194908, 0.02251160331070423, 0.038142845034599304, -0.03356465697288513, 0.0231656301766634, 0.002552339807152748, -0.003384588984772563, -0.000005460868578666123, -0.012805845588445663, 0.0007014438160695136, 0.02255084551870823, -0.012112577445805073, -0.0003035093250218779, -0.02449984662234783, 0.026278799399733543, -0.014388591051101685, -0.0080641508102417, -0.004198852460831404, -0.002763263415545225, -0.018077302724123, -0.01587977260351181, 0.011968690901994705, 0.03039916791021824, 0.031471770256757736, -0.008273439481854439, 0.0051995133981108665, 0.0026553489733487368, 0.006814959924668074, 0.015971336513757706, 0.001718455576337874, 0.002045468892902136, -0.017763368785381317, 0.03366930037736893, -0.03155025467276573, 0.016023658215999603, 0.012439590878784657, 0.02733832225203514, -0.009130214340984821, 0.020052462816238403, 0.018038060516119003, -0.015997497364878654, -0.012177979573607445, -0.00482344813644886, -0.03194267302751541, 0.005006575491279364, -0.0030297792982310057, 0.035500578582286835, 0.002609567018225789, -0.01297589298337698, 0.029248081147670746, -0.03181186690926552, 0.015932094305753708, 0.012897409498691559, -0.014467073604464531, 0.0161152221262455, -0.028070831671357155, 0.01408773846924305, 0.030817745253443718, -0.009417986497282982, -0.010143956169486046, 0.013956933282315731, -0.018338913097977638, -0.033695463091135025, 0.005124300252646208, 0.019267631694674492, -0.009509550407528877, -0.014113899320363998, -0.015343469567596912, -0.004993495065718889, 0.004489894490689039, -0.011458550579845905, -0.004964063875377178, 0.011720160953700542, -0.028463248163461685, -0.006641642656177282, 0.006520647555589676, -0.020955020561814308, 0.01852204091846943, 0.0031883809715509415, 0.002763263415545225, 0.004725344013422728, -0.00533685926347971, 0.014244704507291317, -0.008391164243221283, -0.014610960148274899, 0.010641016997396946, 0.03884919360280037, 0.005225674714893103, -0.021582886576652527, 0.02052336186170578, 0.04889504611492157, -0.005765246693044901, -0.009123674593865871, -0.004149800166487694, 0.02434287965297699, 0.025585530325770378, -0.01747559756040573, 0.010935328900814056, -0.014977214857935905, 0.004506244789808989, 0.024277476593852043, 0.0014241434400901198, 0.06268193572759628, 0.013361768797039986, -0.01416622195392847, 0.00785486213862896, 0.008966708555817604, -0.0281754769384861, -0.10412107408046722, -0.010274761356413364, 0.000498695473652333, 0.027913866564631462, 0.02391122095286846, 0.01646839641034603, -0.028489409014582634, -0.015932094305753708, -0.016036737710237503, 0.033276885747909546, -0.02431671880185604, -0.01337484922260046, -0.004689372610300779, 0.00021112804824952036, -0.0022547575645148754, 0.006157662719488144, -0.003371508326381445, -0.004663211293518543, -0.05156347528100014, 0.02235463820397854, 0.00893400702625513, 0.0017920335521921515, 0.015997497364878654, 0.014100818894803524, 0.005807758308947086, -0.0015197949251160026, -0.030660778284072876, 0.01671692542731762, 0.013577597215771675, 0.0007026700768619776, -0.014283946715295315, -0.007240077015012503, -0.009908506646752357, -0.04298264533281326, 0.013799966312944889, -0.016429154202342033, 0.01145200990140438, 0.016402993351221085, 0.011432389728724957, -0.022812455892562866, 0.01565740257501602, 0.0017527919262647629, 0.02173985168337822, 0.007168134208768606, -0.008685476146638393, -0.00311480276286602, -0.0007325100596062839, 0.04729922115802765, -0.015997497364878654, -0.009777701459825039, -0.031471770256757736, -0.013034755364060402, -0.037907395511865616, 0.015474275685846806, 0.017593322321772575, 0.0025850411038845778, -0.01772412844002247, 0.016965456306934357, -0.015408872626721859, -0.02136051654815674, -0.014244704507291317, -0.00648794649168849, -0.028489409014582634, -0.005212593823671341, 0.008947087451815605, 0.012701201252639294, -0.012295704334974289, 0.006245956290513277, 0.010281302034854889, -0.011085755191743374, -0.0068411207757890224, 0.0011281963670626283, -0.007998748682439327, 0.01827351003885269, -0.055199865251779556, 0.02074573189020157, -0.008005288429558277, -0.012106036767363548, 0.029091114178299904, -0.020837295800447464, -0.006510837469249964, -0.020562604069709778, 0.013734564185142517, 0.007083110976964235, 0.013499113731086254, 0.04444766417145729, 0.017541000619530678, -0.01015703659504652, 0.027521450072526932, -0.0271813552826643, -0.004594538826495409, -0.001396347302943468, 0.06718163937330246, -0.008986328728497028, -0.004571647848933935, -0.03155025467276573, 0.0115501144900918, -0.03364314138889313, 0.013145939446985722, 0.009195617400109768, -0.013708402402698994, -0.004905201494693756, -0.052008215337991714, 0.01971236988902092, -0.0018230997957289219, -0.01747559756040573, 0.00023749349929858, -0.008986328728497028, 0.03686095401644707, -0.026200314983725548, 0.006474865600466728, 0.010601774789392948, -0.003443451365455985, -0.0030232390854507685, -0.014362429268658161, -0.01247229240834713, -0.017370952293276787, -0.03722720965743065, 0.016808489337563515, 0.006111880764365196, 0.0037573841400444508, 0.005003305617719889, 0.0007754305843263865, 0.031236322596669197, 0.01669076457619667, 0.006814959924668074, -0.014964134432375431, -0.017200905829668045, 0.002591581316664815, 0.0016579580260440707, -0.02374117448925972, -0.005915672983974218, -0.008181875571608543, -0.015788208693265915, 0.014375509694218636, 0.051223382353782654, -0.021452080458402634, -0.02372809313237667, 0.022786295041441917, 0.034323327243328094, 0.01646839641034603, 0.032125797122716904, -0.0005796313052996993, -0.021007342264056206, 0.004950983449816704, 0.018757490441203117, 0.00043043142068199813, -0.018783651292324066, -0.003960132598876953, -0.0016342495800927281, 0.013708402402698994, 0.02773073874413967, 0.014807167463004589, -0.00584045983850956, -0.022668570280075073, 0.005124300252646208, 0.019372275099158287, 0.0032635938841849566, 0.016154462471604347, -0.010667177848517895, 0.026488088071346283, -0.012903950177133083, 0.036520857363939285, 0.003629849059507251, 0.011818265542387962, 0.0014151505893096328, -0.005071978084743023, -0.027626093477010727, -0.04954907298088074, 0.005860080476850271, 0.011033432558178902, -0.015565838664770126, -0.009496469981968403, 0.013970013707876205, -0.004754775203764439, 0.010000070556998253, 0.019803933799266815, -0.013616839423775673, -0.020392557606101036, -0.0021321275271475315, 0.0031278834212571383, 0.007410124409943819, 0.024212073534727097, -0.002486936980858445, -0.003255418734624982, 0.009516091085970402, 0.02234155684709549, 0.013250583782792091, -0.020170187577605247, -0.002581770997494459, -0.0011739782057702541, 0.021033503115177155, 0.007737137842923403, 0.02171369083225727, -0.018862133845686913, 0.007357802242040634, 0.019149906933307648, -0.005242025014013052, -0.02794002741575241, -0.009110594168305397, -0.004149800166487694, 0.0017724127974361181, 0.0014511221088469028, 0.011458550579845905, 0.0191368255764246, -0.029457369819283485, -0.009411446750164032, -0.00010254936933051795, -0.02031407319009304, -0.019633885473012924, 0.0007721604197286069, 0.01751483976840973, 0.009378745220601559, 0.013244044035673141, -0.0009916682029142976, -0.0019081233767792583, -0.007782919332385063, 0.029509691521525383, 0.0231133084744215, -0.01790725439786911, -0.004882310517132282, 0.029614334926009178, 0.008606993593275547, 0.009136755019426346, 0.0010284571908414364, -0.011883667670190334, 0.029875947162508965, 0.008947087451815605, 0.02572941593825817, -0.018587442114949226, 0.03275366500020027, -0.028437087312340736, -0.009823483414947987, 0.009110594168305397, 0.0019898766186088324, -0.008404244668781757, -0.013348688371479511, -0.005546147469431162, -0.006465055514127016, -0.01005239225924015, -0.0035742567852139473, 0.08512813597917557, 0.025507047772407532, -0.017410194501280785, 0.04484008252620697, 0.010130875743925571, 0.020628007128834724, 0.009431066922843456, -0.006893442943692207, 0.006579509936273098, -0.05394413322210312, -0.001287615392357111, 0.0020961561240255833, -0.004852879326790571, 0.0026880502700805664, -0.01075874175876379, 0.015539677813649178, -0.005588659550994635, 0.026448845863342285, -0.021268952637910843, -0.006906523369252682, 0.021556725725531578, 0.017240148037672043, 0.0062557668425142765, 0.00423155352473259, -0.02232847735285759, -0.015605080872774124, 0.02637036330997944, 0.012956271879374981, -0.001310506253503263, -0.01811654306948185, -0.015317308716475964, -0.009640355594456196, -0.021282033994793892, -0.03196883201599121, 0.005428422708064318, -0.02171369083225727, 0.018600523471832275, -0.01887521520256996, 0.007861402817070484, 0.010719500482082367, -0.0033518876880407333, 0.0017446166602894664, 0.0037344933953136206, -0.043950602412223816, -0.026605812832713127, 0.006298278458416462, -0.022890940308570862, -0.0005546965403482318, -0.05284537002444267]

Great! We have our vector embedding data.

WTF? How is this going to help us in our search? I’m glad you asked, my friend. We’ll look at that next.

Step 2 — Run an example query against our Neo4j Vector data

TIMEOUT: What is vector search?

According to the guide at https://www.elastic.co/what-is/vector-search:

Vector search leverages machine learning (ML) to capture the meaning and context of unstructured data, including text and images, transforming it into a numeric representation. Frequently used for semantic search, vector search finds similar data using approximate nearest neighbor (ANN) algorithms. Compared to traditional keyword search, vector search yields more relevant results and executes faster.

Vector Similarity Search

Let’s query our Neo4j graph to see if we can answer the question, “Where did Euler grow up?” with our imported data.

First, let’s initialize our Neo4j Vector. The default name of a Neo4j Vector index is vector — which is what is being passed in along with our Neo4j credentials.

Using the Neo4jVector vector store, we will want to use the from_existing_index method:

Once we have our Neo4j Vector Index, we can perform our similarity search using the LangChain module’s similarity_search method:

Let’s take a look at the results of our query!

Our application will display the first result’s page_content — a processed chunk of data.

Step 3 — A simple question/answer workflow using LangChain and our Neo4j backend

For the last step in our demo, we will explore a sample question/answer workflow using LangChain and our Neo4j backend.

What’s involved here? It is a straightforward plan to initialize and execute a LangChain Question Answering workflow.

So how is this going to work — at a high level? 🤔

The LangChain Question Answering Overview contains a beautiful visualization of what we’re going to be doing:

We will pick up at step 4 — Retrieval — since the neo4j_vector is our vectorstore. We will initialize our Neo4j Vector Index for searching and then generate a prompt for the LLM, including our original query and retrieved data.