Vektor-Embeddings

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22query = f"""Use the below article on the 2022 Winter Olympics to answer the subsequent question. If the answer cannot be found, write "I don't know."

Article:
\"\"\"
{wikipedia_article_on_curling}
\"\"\"

Question: Which athletes won the gold medal in curling at the 2022 Winter Olympics?"""

response = client.chat.completions.create(
    messages=[
        {
            "role": "system",
            "content": "You answer questions about the 2022 Winter Olympics.",
        },
        {"role": "user", "content": query},
    ],
    model=GPT_MODEL,
    temperature=0,
)

print(response.choices[0].message.content)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

String article =
    "At the 2022 Winter Olympics, Great Britain won women's curling and Sweden won men's curling.";
String question =
    "Use the below article on the 2022 Winter Olympics to answer the subsequent question. "
        + "If the answer cannot be found, write \"I don't know.\"\n\n"
        + "Article:\n"
        + article
        + "\n\nQuestion: Which athletes won the gold medal in curling at the 2022 Winter Olympics?";

ChatCompletionCreateParams params =
    ChatCompletionCreateParams.builder()
        .model("gpt-4.1-mini")
        .addSystemMessage("You answer questions about the 2022 Winter Olympics.")
        .addUserMessage(question)
        .temperature(0)
        .build();

client.chat().completions().create(params).choices().stream()
    .flatMap(choice -> choice.message().content().stream())
    .forEach(System.out::println);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29require "openai"

client = OpenAI::Client.new
article = "At the 2022 Winter Olympics, Great Britain won women's curling and Sweden won men's curling."
question = <<~QUESTION
  Use the article below to answer the question. If the answer cannot be found, say "I don't know."

  Article:
  #{article}

  Question: Which athletes won the gold medal in curling at the 2022 Winter Olympics?
QUESTION

response = client.chat.completions.create(
  model: "gpt-4.1-mini",
  messages: [
    {
      role: :system,
      content: "You answer questions about the 2022 Winter Olympics."
    },
    {
      role: :user,
      content: question
    }
  ],
  temperature: 0
)

puts(response.choices.fetch(0).message.content)

1
2
3
4
5
6
7
8
9
10def search_reviews(df, product_description, n=3, pprint=True):
    embedding = get_embedding(product_description, model="text-embedding-3-small")
    df["similarities"] = df.ada_embedding.apply(
        lambda x: cosine_similarity(x, embedding)
    )
    res = df.sort_values("similarities", ascending=False).head(n)
    return res


res = search_reviews(df, "delicious beans", n=3)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.embeddings.EmbeddingCreateParams;
import java.util.Comparator;
import java.util.List;
import java.util.stream.IntStream;

List<String> reviews =
    List.of(
        "A rich cup of coffee.",
        "Smooth beans in tomato sauce.",
        "Dark chocolate with orange.");
var reviewEmbeddings =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .inputOfArrayOfStrings(reviews)
                .build())
        .data();
List<Float> query =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .inputOfArrayOfStrings(List.of("delicious beans"))
                .build())
        .data()
        .get(0)
        .embedding();

IntStream.range(0, reviews.size())
    .boxed()
    .sorted(
        Comparator.comparingDouble(
                (Integer index) ->
                    cosineSimilarity(query, reviewEmbeddings.get(index).embedding()))
            .reversed())
    .limit(3)
    .map(reviews::get)
    .forEach(System.out::println);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30require "openai"

client = OpenAI::Client.new
reviews = [
  "A rich cup of coffee.",
  "Smooth beans in tomato sauce.",
  "Dark chocolate with orange."
]

response = client.embeddings.create(
  model: "text-embedding-3-small",
  input: reviews + ["delicious beans"]
)

query = response.data.fetch(-1).embedding
similarity = lambda do |embedding|
  dot_product = embedding.zip(query).sum { |value, query_value| value * query_value }
  magnitude = Math.sqrt(embedding.sum { |value| value**2 })
  query_magnitude = Math.sqrt(query.sum { |value| value**2 })
  dot_product / (magnitude * query_magnitude)
end

results = reviews.map.with_index do |review, index|
  {
    review: review,
    score: similarity.call(response.data.fetch(index).embedding)
  }
end.sort_by { |result| -result.fetch(:score) }.first(3)

puts(results)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16df["code_embedding"] = df["code"].apply(
    lambda x: get_embedding(x, model="text-embedding-3-small")
)


def search_functions(df, code_query, n=3, pprint=True, n_lines=7):
    embedding = get_embedding(code_query, model="text-embedding-3-small")
    df["similarities"] = df.code_embedding.apply(
        lambda x: cosine_similarity(x, embedding)
    )

    res = df.sort_values("similarities", ascending=False).head(n)
    return res


res = search_functions(df, "Completions API tests", n=3)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.embeddings.EmbeddingCreateParams;
import java.util.Comparator;
import java.util.List;
import java.util.stream.IntStream;

List<String> functions =
    List.of("def add(a, b): return a + b", "def complete(prompt): return prompt");
var functionEmbeddings =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .inputOfArrayOfStrings(functions)
                .build())
        .data();
List<Float> query =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .input("Completions API tests")
                .build())
        .data()
        .get(0)
        .embedding();
IntStream.range(0, functions.size())
    .boxed()
    .sorted(
        Comparator.comparingDouble(
                (Integer index) ->
                    cosineSimilarity(query, functionEmbeddings.get(index).embedding()))
            .reversed())
    .map(functions::get)
    .forEach(System.out::println);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29require "openai"

client = OpenAI::Client.new
functions = [
  "function add(a, b) { return a + b; }",
  "function complete(prompt) { return prompt; }"
]

response = client.embeddings.create(
  model: "text-embedding-3-small",
  input: functions + ["Completions API tests"]
)

query = response.data.fetch(-1).embedding
similarity = lambda do |embedding|
  dot_product = embedding.zip(query).sum { |value, query_value| value * query_value }
  magnitude = Math.sqrt(embedding.sum { |value| value**2 })
  query_magnitude = Math.sqrt(query.sum { |value| value**2 })
  dot_product / (magnitude * query_magnitude)
end

results = functions.map.with_index do |source, index|
  {
    source: source,
    score: similarity.call(response.data.fetch(index).embedding)
  }
end.sort_by { |result| -result.fetch(:score) }

puts(results)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23def recommendations_from_strings(
    strings: list[str],
    index_of_source_string: int,
    model="text-embedding-3-small",
) -> list[int]:
    """Return nearest neighbors of a given string."""

    # get embeddings for all strings
    embeddings = [embedding_from_string(string, model=model) for string in strings]

    # get the embedding of the source string
    query_embedding = embeddings[index_of_source_string]

    # get distances between the source embedding and other embeddings (function from embeddings_utils.py)
    distances = distances_from_embeddings(
        query_embedding, embeddings, distance_metric="cosine"
    )

    # get indices of nearest neighbors (function from embeddings_utils.py)
    indices_of_nearest_neighbors = indices_of_nearest_neighbors_from_distances(
        distances
    )
    return indices_of_nearest_neighbors

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.embeddings.EmbeddingCreateParams;
import java.util.Comparator;
import java.util.List;
import java.util.stream.IntStream;

List<String> strings =
    List.of(
        "A cheetah is a fast land animal.",
        "A peregrine falcon is a fast bird.",
        "A tortoise moves slowly.");

var embeddings =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .inputOfArrayOfStrings(strings)
                .build())
        .data();

List<Float> query = embeddings.get(0).embedding();
var nearestNeighbors =
    IntStream.range(0, embeddings.size())
        .boxed()
        .sorted(
            Comparator.comparingDouble(
                (Integer index) -> {
                  List<Float> candidate = embeddings.get(index).embedding();
                  double dotProduct = 0;
                  double queryMagnitude = 0;
                  double candidateMagnitude = 0;
                  for (int dimension = 0; dimension < query.size(); dimension++) {
                    dotProduct += query.get(dimension) * candidate.get(dimension);
                    queryMagnitude += query.get(dimension) * query.get(dimension);
                    candidateMagnitude += candidate.get(dimension) * candidate.get(dimension);
                  }
                  return 1 - dotProduct / Math.sqrt(queryMagnitude * candidateMagnitude);
                }))
        .toList();

System.out.println(nearestNeighbors);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31require "openai"

client = OpenAI::Client.new
strings = [
  "A cheetah is a fast land animal.",
  "A peregrine falcon is a fast bird.",
  "A tortoise moves slowly."
]

response = client.embeddings.create(
  model: "text-embedding-3-small",
  input: strings
)

query = response.data.fetch(0).embedding
similarity = lambda do |embedding|
  dot_product = embedding.zip(query).sum { |value, query_value| value * query_value }
  magnitude = Math.sqrt(embedding.sum { |value| value**2 })
  query_magnitude = Math.sqrt(query.sum { |value| value**2 })
  dot_product / (magnitude * query_magnitude)
end

recommendations = response.data.map.with_index do |embedding, index|
  {
    index: index,
    text: strings.fetch(index),
    similarity: similarity.call(embedding.embedding)
  }
end.sort_by { |recommendation| -recommendation.fetch(:similarity) }

puts(recommendations)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18df = df[df.Score != 3]
df["sentiment"] = df.Score.replace(
    {1: "negative", 2: "negative", 4: "positive", 5: "positive"}
)

labels = ["negative", "positive"]
label_embeddings = [get_embedding(label, model=model) for label in labels]


def label_score(review_embedding, label_embeddings):
    return cosine_similarity(review_embedding, label_embeddings[1]) - cosine_similarity(
        review_embedding, label_embeddings[0]
    )


prediction = (
    "positive" if label_score(get_embedding("Sample Review", model=model), label_embeddings) > 0 else "negative"
)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.embeddings.EmbeddingCreateParams;
import java.util.List;

var embeddings =
    client
        .embeddings()
        .create(
            EmbeddingCreateParams.builder()
                .model("text-embedding-3-small")
                .inputOfArrayOfStrings(List.of("negative", "positive", "Sample Review"))
                .build())
        .data();

List<Float> review = embeddings.get(2).embedding();
double negative = cosineSimilarity(review, embeddings.get(0).embedding());
double positive = cosineSimilarity(review, embeddings.get(1).embedding());
System.out.println(positive > negative ? "positive" : "negative");

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22require "openai"

client = OpenAI::Client.new
labels = ["negative", "positive"]

response = client.embeddings.create(
  model: "text-embedding-3-small",
  input: labels + ["The coffee arrived quickly and tastes great."]
)

review = response.data.fetch(-1).embedding
similarity = lambda do |embedding|
  dot_product = embedding.zip(review).sum { |value, review_value| value * review_value }
  magnitude = Math.sqrt(embedding.sum { |value| value**2 })
  review_magnitude = Math.sqrt(review.sum { |value| value**2 })
  dot_product / (magnitude * review_magnitude)
end

negative, positive = response.data.first(2).map do |embedding|
  similarity.call(embedding.embedding)
end
puts((positive > negative) ? "positive" : "negative")