leva 0.2.1 → 0.3.1

data/app/views/leva/runner_results/show.html.erb

@@ -95,13 +95,6 @@
           <%
             score = eval_result.score
             score_pct = (score * 100).round
-            score_class = case score
-              when 0...0.2 then 'score-bad'
-              when 0.2...0.4 then 'score-poor'
-              when 0.4...0.6 then 'score-fair'
-              when 0.6...0.8 then 'score-good'
-              else 'score-excellent'
-            end
             score_label = case score
               when 0...0.2 then 'Bad'
               when 0.2...0.4 then 'Poor'
@@ -116,12 +109,12 @@
             short_name = short_name.presence || eval_result.evaluator_class.demodulize.gsub(/Eval(uator)?$/, '')
           %>
           <div class="eval-summary-card" title="<%= eval_result.evaluator_class %>">
-            <span class="eval-summary-score <%= score_class %>"><%= score_pct %><span class="eval-summary-pct">%</span></span>
+            <span class="eval-summary-score <%= score_class(score) %>"><%= score_pct %><span class="eval-summary-pct">%</span></span>
             <span class="eval-summary-name"><%= short_name %></span>
             <div class="eval-summary-bar">
-              <div class="eval-summary-bar-fill <%= score_class %>" style="width: <%= score_pct %>%"></div>
+              <div class="eval-summary-bar-fill <%= score_class(score) %>" style="width: <%= score_pct %>%"></div>
             </div>
-            <span class="eval-summary-label <%= score_class %>"><%= score_label %></span>
+            <span class="eval-summary-label <%= score_class(score) %>"><%= score_label %></span>
           </div>
         <% end %>
       </div>
@@ -217,49 +210,26 @@
     </div>
   </div>
 
-  <%# Raw Output - Enhanced Collapsible Section %>
-  <% if @runner_result.respond_to?(:raw_output) && @runner_result.raw_output.present? %>
+  <%# Raw LLM Response - Enhanced Collapsible Section %>
+  <% if @runner_result.prediction.present? && @runner_result.prediction != prediction_text %>
     <%
-      raw_output = @runner_result.raw_output
+      raw_output = @runner_result.prediction
       line_count = raw_output.lines.count
       char_count = raw_output.length
     %>
     <div class="result-section">
-      <details class="raw-output-collapsible">
-        <summary class="raw-output-header">
-          <div class="raw-output-header-left">
-            <svg class="icon-sm raw-output-icon" viewBox="0 0 20 20" fill="currentColor">
-              <path fill-rule="evenodd" d="M12.316 3.051a1 1 0 01.633 1.265l-4 12a1 1 0 11-1.898-.632l4-12a1 1 0 011.265-.633zM5.707 6.293a1 1 0 010 1.414L3.414 10l2.293 2.293a1 1 0 11-1.414 1.414l-3-3a1 1 0 010-1.414l3-3a1 1 0 011.414 0zm8.586 0a1 1 0 011.414 0l3 3a1 1 0 010 1.414l-3 3a1 1 0 11-1.414-1.414L16.586 10l-2.293-2.293a1 1 0 010-1.414z" clip-rule="evenodd" />
-            </svg>
-            <span class="raw-output-title">Raw Output</span>
-            <span class="raw-output-stats">
-              <span class="raw-output-stat"><%= line_count %> line<%= line_count == 1 ? '' : 's' %></span>
-              <span class="raw-output-stat-sep"></span>
-              <span class="raw-output-stat"><%= number_to_human_size(char_count) %></span>
-            </span>
-          </div>
-          <div class="raw-output-header-right">
-            <span class="raw-output-expand-hint">
-              <span class="expand-text">Show</span>
-              <span class="collapse-text">Hide</span>
-            </span>
-            <svg class="raw-output-chevron" viewBox="0 0 20 20" fill="currentColor">
-              <path fill-rule="evenodd" d="M5.293 7.293a1 1 0 011.414 0L10 10.586l3.293-3.293a1 1 0 111.414 1.414l-4 4a1 1 0 01-1.414 0l-4-4a1 1 0 010-1.414z" clip-rule="evenodd" />
-            </svg>
-          </div>
-        </summary>
-        <div class="raw-output-body">
-          <div class="raw-output-toolbar">
-            <span class="raw-output-lang-hint">
-              <svg class="icon-sm" viewBox="0 0 20 20" fill="currentColor">
-                <path fill-rule="evenodd" d="M18 10a8 8 0 11-16 0 8 8 0 0116 0zm-7-4a1 1 0 11-2 0 1 1 0 012 0zM9 9a1 1 0 000 2v3a1 1 0 001 1h1a1 1 0 100-2v-3a1 1 0 00-1-1H9z" clip-rule="evenodd" />
-              </svg>
-              Response from LLM
-            </span>
-          </div>
-          <pre class="raw-output-code"><code><%= raw_output %></code></pre>
-        </div>
-      </details>
+      <div class="result-section-header">
+        <h2 class="result-section-title">
+          <svg class="icon-sm" viewBox="0 0 20 20" fill="currentColor">
+            <path fill-rule="evenodd" d="M12.316 3.051a1 1 0 01.633 1.265l-4 12a1 1 0 11-1.898-.632l4-12a1 1 0 011.265-.633zM5.707 6.293a1 1 0 010 1.414L3.414 10l2.293 2.293a1 1 0 11-1.414 1.414l-3-3a1 1 0 010-1.414l3-3a1 1 0 011.414 0zm8.586 0a1 1 0 011.414 0l3 3a1 1 0 010 1.414l-3 3a1 1 0 11-1.414-1.414L16.586 10l-2.293-2.293a1 1 0 010-1.414z" clip-rule="evenodd" />
+          </svg>
+          Raw LLM Response<% if @experiment.metadata&.dig("model").present? %> <span class="text-muted font-mono text-sm">(<%= @experiment.metadata["model"] %>)</span><% end %>
+        </h2>
+        <span class="result-section-count"><%= line_count %> line<%= line_count == 1 ? '' : 's' %>, <%= number_to_human_size(char_count) %></span>
+      </div>
+      <div class="result-panel">
+        <pre class="result-code"><code><%= raw_output %></code></pre>
+      </div>
     </div>
   <% end %>
 

data/app/views/leva/workbench/_results_section.html.erb

@@ -131,14 +131,6 @@
           <% evaluation_result = @dataset_record&.evaluation_results&.for_evaluator(evaluator_class)&.last %>
           <% score = evaluation_result&.score %>
           <%
-            score_class = case score
-              when 0...0.2 then 'score-bad'
-              when 0.2...0.4 then 'score-poor'
-              when 0.4...0.6 then 'score-fair'
-              when 0.6...0.8 then 'score-good'
-              when 0.8..1.0 then 'score-excellent'
-              else ''
-            end
             bg_style = case score
               when 0...0.2 then 'background: rgba(207, 111, 98, 0.08);'
               when 0.2...0.4 then 'background: rgba(232, 161, 88, 0.08);'
@@ -153,7 +145,7 @@
               <span class="eval-name"><%= evaluator_class.name.demodulize.gsub(/Evaluator$/, '').gsub(/([a-z])([A-Z])/, '\1 \2') %></span>
               <div class="flex items-center gap-2">
                 <% if score %>
-                  <span class="eval-score <%= score_class %>"><%= sprintf('%.0f', score * 100) %></span>
+                  <span class="eval-score <%= score_class(score) %>"><%= sprintf('%.0f', score * 100) %></span>
                   <span class="text-xs text-muted" style="font-size: 10px;">%</span>
                 <% else %>
                   <span class="eval-score eval-score--empty">

data/db/migrate/20241204000001_create_leva_optimization_runs.rb

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+
+class CreateLevaOptimizationRuns < ActiveRecord::Migration[7.2]
+  def change
+    create_table :leva_optimization_runs do |t|
+      t.references :dataset, null: false, foreign_key: { to_table: :leva_datasets }
+      t.references :prompt, foreign_key: { to_table: :leva_prompts }
+      t.string :status, default: "pending", null: false
+      t.string :current_step
+      t.integer :progress, default: 0, null: false
+      t.integer :examples_processed, default: 0
+      t.integer :total_examples
+      t.string :prompt_name, null: false
+      t.string :mode, default: "light", null: false
+      t.text :error_message
+      t.json :metadata
+      t.string :model
+      t.string :optimizer, default: "bootstrap", null: false
+
+      t.timestamps
+    end
+
+    add_index :leva_optimization_runs, :status
+  end
+end

data/lib/generators/leva/templates/eval.rb.erb

@@ -8,7 +8,9 @@ class <%= class_name %>Eval < Leva::BaseEval
     # Implement your evaluation logic here
     # You can access the ground truth using recordable.ground_truth
 
-    # Example implementation:
-    runner_result.parsed_predictions.first == recordable.ground_truth ? 1.0 : 0.0
+    # Example implementation (case-insensitive comparison):
+    prediction = runner_result.parsed_predictions.first.to_s.downcase
+    expected = recordable.ground_truth.to_s.downcase
+    prediction == expected ? 1.0 : 0.0
   end
 end

data/lib/leva/errors.rb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+module Leva
+  # Base error class for all Leva errors
+  class Error < StandardError; end
+
+  # Raised when a dataset has insufficient records for optimization
+  class InsufficientDataError < Error; end
+
+  # Raised when DSPy is not properly configured
+  class DspyConfigurationError < Error; end
+
+  # Raised when optimization fails
+  class OptimizationError < Error; end
+
+  # Raised when a runner encounters an error during execution
+  class RunnerError < Error; end
+end

data/lib/leva/version.rb

@@ -1,3 +1,3 @@
 module Leva
-  VERSION = "0.2.1"
+  VERSION = "0.3.1"
 end

data/lib/leva.rb

@@ -1,5 +1,6 @@
 require "leva/version"
 require "leva/engine"
+require "leva/errors"
 require "liquid"
 
 module Leva

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: leva
 version: !ruby/object:Gem::Version
-  version: 0.2.1
+  version: 0.3.1
 platform: ruby
 authors:
 - Kieran Klaassen
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2025-11-23 00:00:00.000000000 Z
+date: 2025-12-07 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rails
@@ -53,6 +53,7 @@ files:
 - app/assets/config/leva_manifest.js
 - app/assets/stylesheets/leva/application.css
 - app/controllers/leva/application_controller.rb
+- app/controllers/leva/dataset_optimizations_controller.rb
 - app/controllers/leva/dataset_records_controller.rb
 - app/controllers/leva/datasets_controller.rb
 - app/controllers/leva/design_system_controller.rb
@@ -72,8 +73,17 @@ files:
 - app/models/leva/dataset_record.rb
 - app/models/leva/evaluation_result.rb
 - app/models/leva/experiment.rb
+- app/models/leva/optimization_run.rb
 - app/models/leva/prompt.rb
 - app/models/leva/runner_result.rb
+- app/services/leva/class_loader.rb
+- app/services/leva/dataset_converter.rb
+- app/services/leva/optimizers/base.rb
+- app/services/leva/optimizers/bootstrap.rb
+- app/services/leva/optimizers/gepa_optimizer.rb
+- app/services/leva/optimizers/miprov2_optimizer.rb
+- app/services/leva/prompt_optimizer.rb
+- app/services/leva/signature_generator.rb
 - app/views/layouts/leva/application.html.erb
 - app/views/leva/dataset_records/index.html.erb
 - app/views/leva/dataset_records/show.html.erb
@@ -90,6 +100,7 @@ files:
 - app/views/leva/experiments/index.html.erb
 - app/views/leva/experiments/new.html.erb
 - app/views/leva/experiments/show.html.erb
+- app/views/leva/optimization_runs/show.html.erb
 - app/views/leva/runner_results/show.html.erb
 - app/views/leva/workbench/_evaluation_area.html.erb
 - app/views/leva/workbench/_prompt_content.html.erb
@@ -107,12 +118,14 @@ files:
 - db/migrate/20240813173035_create_leva_experiments.rb
 - db/migrate/20240813173040_create_leva_runner_results.rb
 - db/migrate/20240813173050_create_leva_evaluation_results.rb
+- db/migrate/20241204000001_create_leva_optimization_runs.rb
 - lib/generators/leva/eval_generator.rb
 - lib/generators/leva/runner_generator.rb
 - lib/generators/leva/templates/eval.rb.erb
 - lib/generators/leva/templates/runner.rb.erb
 - lib/leva.rb
 - lib/leva/engine.rb
+- lib/leva/errors.rb
 - lib/leva/version.rb
 - lib/tasks/leva_tasks.rake
 homepage: https://github.com/kieranklaassen/leva
@@ -137,7 +150,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.4.10
+rubygems_version: 3.5.3
 signing_key:
 specification_version: 4
 summary: Flexible Evaluation Framework for Language Models in Rails