API Integrations

library(lares)

default:
  openai:
    secret_key: "sk-your-openai-key-here"
  
  gemini:
    api_key: "your-gemini-key-here"
  
  database:
    server: "localhost"
    database: "mydb"
    uid: "user"
    pwd: "password"

# lares will prompt you to set the directory
creds <- get_credentials("openai")

# Or specify manually
creds <- get_credentials("openai", dir = "~/my_credentials")

# Simple question
response <- gpt_ask("Explain linear regression in 2 sentences")

# Create a structured prompt
prompt <- gpt_prompter(
  instruction = "Classify these items",
  input = c("Dog", "Rose", "Car", "Oak"),
  context = c("Animal", "Plant", "Object"),
  output = "table",
  cols = c("Item", "Category")
)

# Send to ChatGPT
result <- gpt_ask(prompt$prompt)

# See all previous prompts and responses from this session
history <- gpt_history()
head(history)

# Customize model and parameters
response <- gpt_ask(
  "Write a haiku about data science",
  model = "gpt-4",
  temperature = 0.9, # More creative (0-2)
  num_retries = 3
)

# Text generation
response <- gemini_ask(
  "Explain the Central Limit Theorem",
  api_key = get_credentials("gemini")$api_key
)

# Analyze an image
response <- gemini_image(
  prompt = "Describe this plot in detail",
  image_path = "path/to/plot.png",
  api_key = get_credentials("gemini")$api_key
)

# Fetch Apple stock data
aapl <- stocks_hist(
  symbols = "AAPL",
  from = Sys.Date() - 90, # Last 90 days
  to = Sys.Date()
)

head(aapl, 3)

# Compare multiple stocks
tech_stocks <- stocks_hist(
  symbols = c("AAPL", "GOOGL", "MSFT"),
  from = "2024-01-01"
)

head(tech_stocks, 3)

# Real-time quotes
quote <- stocks_quote("AAPL")
print(quote)

# Track a portfolio
portfolio <- daily_portfolio(
  symbols = c("AAPL", "GOOGL", "TSLA"),
  shares = c(10, 5, 8),
  from = "2024-01-01"
)

# Visualize performance
splot_summary(portfolio)

# Growth over time
splot_growth(tech_stocks, symbols = c("AAPL", "GOOGL"))

# Returns on investment
splot_roi(tech_stocks, investment = 10000)

# Sector performance
splot_types(tech_stocks)

# Read a Google Sheet
data <- readGS(
  title = "My Spreadsheet",
  sheet = "Sheet1",
  creds = get_credentials("google")
)

# Write data to Google Sheets
writeGS(
  data = mtcars,
  title = "My Data",
  sheet = "Cars",
  creds = get_credentials("google")
)

# Query a database
query <- "SELECT * FROM users WHERE active = TRUE LIMIT 10"
result <- queryDB(
  query = query,
  from = "my_database_creds"
)

# Get trend data
trends <- gtrends_time(
  keyword = "machine learning",
  from = "2023-01-01",
  to = "2024-01-01"
)

head(trends)

# Find related search terms
related <- gtrends_related(
  keyword = "data science",
  categories = c("top", "rising")
)

# ❌ DON'T hardcode
# api_key <- "sk-1234567890"

# ✅ DO use credential management
api_key <- get_credentials("openai")$secret_key

# Wrap in try-catch
result <- tryCatch(
  {
    gpt_ask("Your question here")
  },
  error = function(e) {
    message("API error: ", e$message)
    NULL
  }
)

# Add delays for bulk operations
items <- c("item1", "item2", "item3")

results <- lapply(items, function(item) {
  result <- gpt_ask(item)
  Sys.sleep(1) # 1 second delay
  result
})

# Cache expensive API calls
result <- cache_pipe(
  {
    # This only runs once, then cached and loaded on subsequent calls
    gpt_ask("Expensive query here")
  },
  base = "my_gpt_query"
)

# Use cheaper models for simple tasks
simple_task <- gpt_ask(
  "Summarize: The meeting is at 3pm",
  model = "gpt-3.5-turbo" # Cheaper than gpt-4
)

# Use expensive models only when needed
complex_task <- gpt_ask(
  "Analyze this complex dataset: ...",
  model = "gpt-4"
)

# Verify credentials are loaded
creds <- get_credentials("openai")
if (is.null(creds$secret_key)) {
  stop("OpenAI API key not found!")
}

# Check internet connection
if (haveInternet()) {
  message("Connected to internet")
} else {
  stop("No internet connection")
}