ollama-client 0.2.4 → 0.2.6

data/examples/advanced_performance_testing.rb

@@ -1,186 +0,0 @@
-#!/usr/bin/env ruby
-# frozen_string_literal: true
-
-# Advanced Example: Performance Testing and Observability
-# Demonstrates: Latency measurement, throughput testing, error rate tracking, concurrent requests
-
-require "json"
-require "benchmark"
-require "time"
-require_relative "../lib/ollama_client"
-
-class PerformanceMonitor
-  def initialize(client:)
-    @client = client
-    @metrics = {
-      calls: [],
-      errors: [],
-      latencies: []
-    }
-  end
-
-  def measure_call(prompt:, schema:)
-    start_time = Time.now
-
-    begin
-      result = @client.generate(prompt: prompt, schema: schema)
-      latency = (Time.now - start_time) * 1000 # Convert to milliseconds
-
-      @metrics[:calls] << {
-        success: true,
-        latency_ms: latency,
-        timestamp: Time.now.iso8601
-      }
-      @metrics[:latencies] << latency
-
-      { success: true, result: result, latency_ms: latency }
-    rescue Ollama::Error => e
-      latency = (Time.now - start_time) * 1000
-      @metrics[:calls] << {
-        success: false,
-        latency_ms: latency,
-        error: e.class.name,
-        timestamp: Time.now.iso8601
-      }
-      @metrics[:errors] << { error: e.class.name, message: e.message, latency_ms: latency }
-
-      { success: false, error: e, latency_ms: latency }
-    end
-  end
-
-  def run_throughput_test(prompt:, schema:, iterations: 10)
-    puts "🚀 Running throughput test (#{iterations} iterations)..."
-    results = []
-
-    total_time = Benchmark.realtime do
-      iterations.times do |i|
-        print "  #{i + 1}/#{iterations}... "
-        result = measure_call(prompt: prompt, schema: schema)
-        results << result
-        puts result[:success] ? "✓" : "✗"
-      end
-    end
-
-    {
-      total_time: total_time,
-      iterations: iterations,
-      throughput: iterations / total_time,
-      results: results
-    }
-  end
-
-  def run_latency_test(prompt:, schema:, iterations: 10)
-    puts "⏱️  Running latency test (#{iterations} iterations)..."
-    latencies = []
-
-    iterations.times do |i|
-      print "  #{i + 1}/#{iterations}... "
-      result = measure_call(prompt: prompt, schema: schema)
-      if result[:success]
-        latencies << result[:latency_ms]
-        puts "#{result[:latency_ms].round(2)}ms"
-      else
-        puts "ERROR"
-      end
-    end
-
-    {
-      latencies: latencies,
-      min: latencies.min,
-      max: latencies.max,
-      avg: latencies.sum / latencies.length,
-      median: latencies.sort[latencies.length / 2],
-      p95: latencies.sort[(latencies.length * 0.95).to_i],
-      p99: latencies.sort[(latencies.length * 0.99).to_i]
-    }
-  end
-
-  def display_metrics
-    puts "\n" + "=" * 60
-    puts "Performance Metrics"
-    puts "=" * 60
-
-    total_calls = @metrics[:calls].length
-    successful = @metrics[:calls].count { |c| c[:success] }
-    failed = total_calls - successful
-
-    puts "Total calls: #{total_calls}"
-    puts "Successful: #{successful} (#{(successful.to_f / total_calls * 100).round(2)}%)"
-    puts "Failed: #{failed} (#{(failed.to_f / total_calls * 100).round(2)}%)"
-
-    if @metrics[:latencies].any?
-      latencies = @metrics[:latencies]
-      puts "\nLatency Statistics (ms):"
-      puts "  Min: #{latencies.min.round(2)}"
-      puts "  Max: #{latencies.max.round(2)}"
-      puts "  Avg: #{(latencies.sum / latencies.length).round(2)}"
-      puts "  Median: #{latencies.sort[latencies.length / 2].round(2)}"
-      puts "  P95: #{latencies.sort[(latencies.length * 0.95).to_i].round(2)}"
-      puts "  P99: #{latencies.sort[(latencies.length * 0.99).to_i].round(2)}"
-    end
-
-    return unless @metrics[:errors].any?
-
-    puts "\nErrors by type:"
-    error_counts = @metrics[:errors].group_by { |e| e[:error] }
-    error_counts.each do |error_type, errors|
-      puts "  #{error_type}: #{errors.length}"
-    end
-  end
-
-  def export_metrics(filename: "metrics.json")
-    File.write(filename, JSON.pretty_generate(@metrics))
-    puts "\n📊 Metrics exported to #{filename}"
-  end
-end
-
-# Run performance tests
-if __FILE__ == $PROGRAM_NAME
-  # Use longer timeout for performance testing
-  config = Ollama::Config.new
-  config.timeout = 60 # 60 seconds for complex operations
-  client = Ollama::Client.new(config: config)
-  monitor = PerformanceMonitor.new(client: client)
-
-  schema = {
-    "type" => "object",
-    "required" => ["response"],
-    "properties" => {
-      "response" => { "type" => "string" }
-    }
-  }
-
-  puts "=" * 60
-  puts "Performance Testing Suite"
-  puts "=" * 60
-
-  # Test 1: Latency
-  latency_results = monitor.run_latency_test(
-    prompt: "Respond with a simple acknowledgment",
-    schema: schema,
-    iterations: 5
-  )
-
-  puts "\nLatency Results:"
-  puts "  Average: #{latency_results[:avg].round(2)}ms"
-  puts "  P95: #{latency_results[:p95].round(2)}ms"
-  puts "  P99: #{latency_results[:p99].round(2)}ms"
-
-  # Test 2: Throughput
-  throughput_results = monitor.run_throughput_test(
-    prompt: "Count to 5",
-    schema: schema,
-    iterations: 5
-  )
-
-  puts "\nThroughput Results:"
-  puts "  Total time: #{throughput_results[:total_time].round(2)}s"
-  puts "  Throughput: #{throughput_results[:throughput].round(2)} calls/sec"
-
-  # Display all metrics
-  monitor.display_metrics
-
-  # Export metrics
-  monitor.export_metrics
-end
-

data/examples/chat_console.rb

@@ -1,143 +0,0 @@
-#!/usr/bin/env ruby
-# frozen_string_literal: true
-
-require_relative "../lib/ollama_client"
-require "tty-reader"
-require "tty-screen"
-require "tty-cursor"
-
-def build_config
-  config = Ollama::Config.new
-  config.base_url = ENV["OLLAMA_BASE_URL"] if ENV["OLLAMA_BASE_URL"]
-  config.model = ENV["OLLAMA_MODEL"] if ENV["OLLAMA_MODEL"]
-  config.temperature = ENV["OLLAMA_TEMPERATURE"].to_f if ENV["OLLAMA_TEMPERATURE"]
-  config
-end
-
-def exit_command?(text)
-  %w[/exit /quit exit quit].include?(text.downcase)
-end
-
-def add_system_message(messages)
-  system_prompt = ENV.fetch("OLLAMA_SYSTEM", nil)
-  return unless system_prompt && !system_prompt.strip.empty?
-
-  messages << { role: "system", content: system_prompt }
-end
-
-def print_banner(config)
-  puts "Ollama chat console"
-  puts "Model: #{config.model}"
-  puts "Base URL: #{config.base_url}"
-  puts "Type /exit to quit."
-  puts "Screen: #{TTY::Screen.width}x#{TTY::Screen.height}"
-  puts
-end
-
-HISTORY_PATH = ".ollama_chat_history"
-MAX_HISTORY = 200
-COLOR_RESET = "\e[0m"
-COLOR_USER = "\e[32m"
-COLOR_LLM = "\e[36m"
-USER_PROMPT = "#{COLOR_USER}you>#{COLOR_RESET} ".freeze
-LLM_PROMPT = "#{COLOR_LLM}llm>#{COLOR_RESET} ".freeze
-
-def build_reader
-  TTY::Reader.new
-end
-
-def read_input(reader)
-  reader.read_line(USER_PROMPT)
-end
-
-def load_history(reader, path)
-  history = load_history_list(path)
-  history.reverse_each { |line| reader.add_to_history(line) }
-end
-
-def load_history_list(path)
-  return [] unless File.exist?(path)
-
-  unique_history(normalize_history(File.readlines(path, chomp: true)))
-end
-
-def normalize_history(lines)
-  lines.map(&:strip).reject(&:empty?)
-end
-
-def unique_history(lines)
-  seen = {}
-  lines.each_with_object([]) do |line, unique|
-    next if seen[line]
-
-    unique << line
-    seen[line] = true
-  end
-end
-
-def update_history(path, text)
-  history = load_history_list(path)
-  history.delete(text)
-  history.unshift(text)
-  history = history.first(MAX_HISTORY)
-
-  File.write(path, history.join("\n") + (history.empty? ? "" : "\n"))
-end
-
-def chat_response(client, messages, config)
-  content = +""
-  prompt_printed = false
-
-  print "#{COLOR_LLM}...#{COLOR_RESET}"
-  $stdout.flush
-
-  client.chat_raw(
-    messages: messages,
-    allow_chat: true,
-    options: { temperature: config.temperature },
-    stream: true
-  ) do |chunk|
-    token = chunk.dig("message", "content").to_s
-    next if token.empty?
-
-    unless prompt_printed
-      print "\r#{LLM_PROMPT}"
-      prompt_printed = true
-    end
-
-    content << token
-    print token
-    $stdout.flush
-  end
-
-  puts
-  content
-end
-
-def run_console(client, config)
-  messages = []
-  add_system_message(messages)
-  print_banner(config)
-  reader = build_reader
-  load_history(reader, HISTORY_PATH)
-
-  loop do
-    input = read_input(reader)
-    break unless input
-
-    text = input.strip
-    next if text.empty?
-    break if exit_command?(text)
-
-    update_history(HISTORY_PATH, text)
-    messages << { role: "user", content: text }
-    content = chat_response(client, messages, config)
-    messages << { role: "assistant", content: content }
-  end
-rescue Interrupt
-  puts "\nExiting..."
-end
-
-config = build_config
-client = Ollama::Client.new(config: config)
-run_console(client, config)

data/examples/complete_workflow.rb

@@ -1,245 +0,0 @@
-#!/usr/bin/env ruby
-# frozen_string_literal: true
-
-# Complete example showing how to use structured outputs in a real workflow
-# This demonstrates the full cycle: schema definition -> LLM call -> using the result
-
-require "json"
-require_relative "../lib/ollama_client"
-
-# Example: Task Planning Agent
-# The LLM decides what action to take, and we execute it
-
-class TaskPlanner
-  def initialize(client:)
-    @client = client
-    @task_schema = {
-      "type" => "object",
-      "required" => ["action", "reasoning", "confidence", "next_step"],
-      "properties" => {
-        "action" => {
-          "type" => "string",
-          "description" => "The action to take",
-          "enum" => ["search", "calculate", "store", "retrieve", "finish"]
-        },
-        "reasoning" => {
-          "type" => "string",
-          "description" => "Why this action was chosen"
-        },
-        "confidence" => {
-          "type" => "number",
-          "minimum" => 0,
-          "maximum" => 1,
-          "description" => "Confidence in this decision (0.0 to 1.0, where 1.0 is 100% confident)"
-        },
-        "next_step" => {
-          "type" => "string",
-          "description" => "What to do next"
-        },
-        "parameters" => {
-          "type" => "object",
-          "description" => "Parameters needed for the action"
-        }
-      }
-    }
-  end
-
-  def plan(context:)
-    puts "🤔 Planning next action..."
-    puts "Context: #{context}\n\n"
-
-    begin
-      prompt = "Given this context: #{context}\n\n" \
-               "Decide the next action to take.\n\n" \
-               "IMPORTANT: Use decimal values for confidence " \
-               "(e.g., 0.95 for 95% confident, 0.80 for 80% confident, 1.0 for 100% confident)."
-
-      result = @client.generate(
-        prompt: prompt,
-        schema: @task_schema
-      )
-
-      # The result is guaranteed to match our schema
-      display_decision(result)
-      execute_action(result)
-
-      result
-    rescue Ollama::SchemaViolationError => e
-      puts "❌ Invalid response structure: #{e.message}"
-      puts "   This shouldn't happen with format parameter, but we handle it gracefully"
-      nil
-    rescue Ollama::Error => e
-      puts "❌ Error: #{e.message}"
-      nil
-    end
-  end
-
-  private
-
-  def display_decision(result)
-    puts "📋 Decision:"
-    puts "   Action: #{result['action']}"
-    puts "   Reasoning: #{result['reasoning']}"
-    puts "   Confidence: #{(result['confidence'] * 100).round}%"
-    puts "   Next Step: #{result['next_step']}"
-    puts "   Parameters: #{JSON.pretty_generate(result['parameters'] || {})}\n"
-  end
-
-  def execute_action(result)
-    case result["action"]
-    when "search"
-      query = result.dig("parameters", "query") || "default"
-      puts "🔍 Executing search: #{query}"
-      # In real code, you'd call your search function here
-      puts "   → Search results would appear here\n"
-
-    when "calculate"
-      operation = result.dig("parameters", "operation") || "unknown"
-      puts "🧮 Executing calculation: #{operation}"
-      # In real code, you'd call your calculator here
-      puts "   → Calculation result would appear here\n"
-
-    when "store"
-      key = result.dig("parameters", "key") || "unknown"
-      puts "💾 Storing data with key: #{key}"
-      # In real code, you'd save to your storage
-      puts "   → Data stored successfully\n"
-
-    when "retrieve"
-      key = result.dig("parameters", "key") || "unknown"
-      puts "📂 Retrieving data with key: #{key}"
-      # In real code, you'd fetch from your storage
-      puts "   → Data retrieved successfully\n"
-
-    when "finish"
-      puts "✅ Task complete!\n"
-
-    else
-      puts "⚠️ Unknown action: #{result['action']}\n"
-    end
-  end
-end
-
-# Example: Data Analyzer
-# The LLM analyzes data and returns structured insights
-
-class DataAnalyzer
-  def initialize(client:)
-    @client = client
-    @analysis_schema = {
-      "type" => "object",
-      "required" => ["summary", "confidence", "key_points"],
-      "properties" => {
-        "summary" => {
-          "type" => "string",
-          "description" => "Brief summary of the analysis"
-        },
-        "confidence" => {
-          "type" => "number",
-          "minimum" => 0,
-          "maximum" => 1,
-          "description" => "Confidence level (0.0 to 1.0, where 1.0 is 100% confident)"
-        },
-        "key_points" => {
-          "type" => "array",
-          "items" => { "type" => "string" },
-          "minItems" => 1,
-          "maxItems" => 5
-        },
-        "sentiment" => {
-          "type" => "string",
-          "enum" => ["positive", "neutral", "negative"]
-        },
-        "recommendations" => {
-          "type" => "array",
-          "items" => { "type" => "string" }
-        }
-      }
-    }
-  end
-
-  def analyze(data:)
-    puts "📊 Analyzing data..."
-    puts "Data: #{data}\n\n"
-
-    begin
-      prompt = "Analyze this data and provide insights: #{data}\n\n" \
-               "IMPORTANT: Express confidence as a decimal between 0.0 and 1.0 " \
-               "(e.g., 0.85 for 85% confidence, not 85)."
-
-      result = @client.generate(
-        prompt: prompt,
-        schema: @analysis_schema
-      )
-
-      display_analysis(result)
-      make_recommendations(result)
-
-      result
-    rescue Ollama::Error => e
-      puts "❌ Error: #{e.message}"
-      nil
-    end
-  end
-
-  private
-
-  def display_analysis(result)
-    puts "📈 Analysis Results:"
-    puts "   Summary: #{result['summary']}"
-    puts "   Confidence: #{(result['confidence'] * 100).round}%"
-    puts "   Sentiment: #{result['sentiment']}"
-    puts "\n   Key Points:"
-    result["key_points"].each_with_index do |point, i|
-      puts "     #{i + 1}. #{point}"
-    end
-
-    if result["recommendations"] && !result["recommendations"].empty?
-      puts "\n   Recommendations:"
-      result["recommendations"].each_with_index do |rec, i|
-        puts "     #{i + 1}. #{rec}"
-      end
-    end
-    puts
-  end
-
-  def make_recommendations(result)
-    if result["confidence"] > 0.8 && result["sentiment"] == "positive"
-      puts "✅ High confidence positive analysis - safe to proceed"
-    elsif result["confidence"] < 0.5
-      puts "⚠️ Low confidence - manual review recommended"
-    elsif result["sentiment"] == "negative"
-      puts "⚠️ Negative sentiment detected - investigate further"
-    end
-    puts
-  end
-end
-
-# Main execution
-if __FILE__ == $PROGRAM_NAME
-  client = Ollama::Client.new
-
-  puts "=" * 60
-  puts "Example 1: Task Planning Agent"
-  puts "=" * 60
-  puts
-
-  planner = TaskPlanner.new(client: client)
-  planner.plan(context: "User wants to know the weather in Paris")
-
-  puts "\n" + "=" * 60
-  puts "Example 2: Data Analysis"
-  puts "=" * 60
-  puts
-
-  analyzer = DataAnalyzer.new(client: client)
-  analyzer.analyze(
-    data: "Sales increased 25% this quarter. Customer satisfaction is at 4.8/5. " \
-          "Revenue: $1.2M. New customers: 150."
-  )
-
-  puts "=" * 60
-  puts "Examples complete!"
-  puts "=" * 60
-end
-