Added OpenAI example [skip ci]

Author: ankane

.gitignore

@@ -1,2 +1,2 @@
-/build/
+build
 /test/pqxx

README.md

@@ -22,6 +22,10 @@ And follow the instructions for your database library:
 
 - [libpqxx](#libpqxx)
 
+Or check out an example:
+
+- [Embeddings](examples/openai/example.cpp) with OpenAI
+
 ## libpqxx
 
 Include the header
@@ -94,3 +98,13 @@ createdb pgvector_cpp_test
 g++ -std=c++17 -Wall -Wextra -Wno-unknown-attributes -Werror -o test/pqxx test/pqxx_test.cpp -lpqxx -lpq
 test/pqxx
 ```
+
+To run an example:
+
+```sh
+cd examples/openai
+createdb pgvector_example
+cmake -S . -B build
+cmake --build build
+build/example
+```

examples/openai/CMakeLists.txt

@@ -0,0 +1,17 @@
+cmake_minimum_required(VERSION 3.18)
+
+project(example)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_FLAGS "-Wno-unknown-attributes")
+
+include(FetchContent)
+
+FetchContent_Declare(cpr GIT_REPOSITORY https://github.com/libcpr/cpr.git GIT_TAG 1.11.1)
+FetchContent_Declare(json GIT_REPOSITORY https://github.com/nlohmann/json.git GIT_TAG v3.11.3)
+FetchContent_Declare(libpqxx GIT_REPOSITORY https://github.com/jtv/libpqxx.git GIT_TAG 7.10.0)
+FetchContent_MakeAvailable(cpr json libpqxx)
+
+add_executable(example example.cpp)
+target_include_directories(example PRIVATE ${CMAKE_SOURCE_DIR}/../../include)
+target_link_libraries(example PRIVATE cpr::cpr nlohmann_json::nlohmann_json pqxx)

examples/openai/example.cpp

@@ -0,0 +1,70 @@
+#include <cstdint>
+#include <iostream>
+
+#include <cpr/cpr.h>
+#include <nlohmann/json.hpp>
+// TODO make <pgvector/pqxx.hpp>
+#include <pqxx.hpp>
+#include <pqxx/pqxx>
+
+using json = nlohmann::json;
+
+// https://platform.openai.com/docs/guides/embeddings/how-to-get-embeddings
+// input can be an array with 2048 elements
+std::vector<std::vector<float>> fetch_embeddings(const std::vector<std::string>& input, char *api_key) {
+  std::string url = "https://api.openai.com/v1/embeddings";
+  json data = {
+    {"input", input},
+    {"model", "text-embedding-3-small"}
+  };
+
+  cpr::Response r = cpr::Post(
+    cpr::Url{url},
+    cpr::Body{data.dump()},
+    cpr::Bearer{api_key},
+    cpr::Header{{"Content-Type", "application/json"}}
+  );
+  json response = json::parse(r.text);
+
+  std::vector<std::vector<float>> embeddings;
+  for (auto& v: response["data"]) {
+    embeddings.emplace_back(v["embedding"]);
+  }
+  return embeddings;
+}
+
+int main() {
+  char *api_key = std::getenv("OPENAI_API_KEY");
+  if (!api_key) {
+    std::cout << "Set OPENAI_API_KEY" << std::endl;
+    return 1;
+  }
+
+  pqxx::connection conn("dbname=pgvector_example");
+
+  pqxx::work tx(conn);
+  tx.exec("CREATE EXTENSION IF NOT EXISTS vector");
+  tx.exec("DROP TABLE IF EXISTS documents");
+  tx.exec("CREATE TABLE documents (id bigserial PRIMARY KEY, content text, embedding vector(1536))");
+  tx.commit();
+
+  std::vector<std::string> input = {
+    "The dog is barking",
+    "The cat is purring",
+    "The bear is growling"
+  };
+  auto embeddings = fetch_embeddings(input, api_key);
+
+  for (size_t i = 0; i < input.size(); i++) {
+    tx.exec("INSERT INTO documents (content, embedding) VALUES ($1, $2)", {input[i], pgvector::Vector(embeddings[i])});
+  }
+  tx.commit();
+
+  int64_t document_id = 1;
+  pqxx::result result = tx.exec("SELECT content FROM documents WHERE id != $1 ORDER BY embedding <=> (SELECT embedding FROM documents WHERE id = $1) LIMIT 5", {document_id});
+  for (auto const& row : result) {
+    std::cout << row[0].c_str() << std::endl;
+  }
+
+  return 0;
+}