ollama-client 0.2.5 → 0.2.7

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
-