completion-kit 0.28.35 → 0.28.37

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f506b4c95279cd20d1085f9f3a2eb3eed2d30b03694d3232977a410044a36f8d
-  data.tar.gz: 4a5ea855e4fbb830f0df97dd1f957b86894710464bf011e3c17878cfaad6c416
+  metadata.gz: 28e080ca8be590088a2a99f16dc5e0f5e6c3f292d29bbc708e6fd8ef14b9b7d3
+  data.tar.gz: f5ff52512c81735d06941634973277915ab89d95a94a0640a47b3436f961d185
 SHA512:
-  metadata.gz: 561278c7bf724b085e2cc05e3204d89b839e16803de24f9accfa59630edba15e4bdf85a9edbd745e03c9385807295749b95fda1b757b62c781265ebc705ca795
-  data.tar.gz: d2744c24f3eff96c4290bfa333145fa45941c6bf3d26cf7768968e452c6bba4d85c1668be97a8054ef1548aec9e55f782af081b82974c3c6d7a12dae6122ee41
+  metadata.gz: 2a8140334350dc31588e0b6f29c8f44fc2b6b4b17a8251d422a9d1d7cc7a605e8bf05e913afcb7fbbd1d682273ce9dd339786050b492e1111ae752285b271518
+  data.tar.gz: 93bc707699f8081213f5344ba2077be80a004adc24445b173a9d815bf4cffd1a68301b53bd1a34a17f9190dbfb5e0acc3790cfe316cdb26c7dcb291c984ed287

data/app/assets/stylesheets/completion_kit/application.css

@@ -4470,49 +4470,6 @@ table.ck-runs-table {
   -webkit-line-clamp: unset;
 }
 
-.ck-slider-row {
-  display: flex;
-  align-items: center;
-  gap: 0.75rem;
-}
-
-.ck-slider {
-  flex: 1;
-  -webkit-appearance: none;
-  appearance: none;
-  height: 4px;
-  background: var(--ck-line-strong);
-  border-radius: 2px;
-  outline: none;
-}
-
-.ck-slider::-webkit-slider-thumb {
-  -webkit-appearance: none;
-  appearance: none;
-  width: 16px;
-  height: 16px;
-  border-radius: 50%;
-  background: var(--ck-accent);
-  cursor: pointer;
-  border: 2px solid var(--ck-bg);
-}
-
-.ck-slider::-moz-range-thumb {
-  width: 16px;
-  height: 16px;
-  border-radius: 50%;
-  background: var(--ck-accent);
-  cursor: pointer;
-  border: 2px solid var(--ck-bg);
-}
-
-.ck-slider-value {
-  font-family: var(--ck-mono);
-  font-size: 0.85rem;
-  color: var(--ck-text);
-  min-width: 2rem;
-  text-align: right;
-}
 
 .ck-info-toggle {
   display: inline-flex;
@@ -6609,10 +6566,6 @@ a.tag-mark {
   margin-right: 0.2rem;
 }
 
-.ck-slider:focus-visible {
-  outline: 2px solid var(--ck-accent);
-  outline-offset: 4px;
-}
 .ck-select__option:focus-visible {
   box-shadow: inset 0 0 0 2px var(--ck-accent);
 }

data/app/controllers/completion_kit/runs_controller.rb

@@ -130,6 +130,7 @@ module CompletionKit
         format.turbo_stream do
           render turbo_stream: [
             turbo_stream.replace("run_status_header", partial: "completion_kit/runs/status_header", locals: { run: @run }),
+          turbo_stream.replace("run_config", partial: "completion_kit/runs/run_config", locals: { run: @run }),
             turbo_stream.replace("run_status_panel", partial: "completion_kit/runs/status_panel", locals: { run: @run }),
             turbo_stream.replace("run_responses_region", partial: "completion_kit/runs/responses_region",
                                  locals: { run: @run, responses: @responses, responses_offset: @responses_offset })
@@ -222,10 +223,7 @@ module CompletionKit
     end
 
     def run_params
-      permitted = params.require(:run).permit(:name, :prompt_id, :dataset_id, :judge_model, :temperature, :max_tokens, :judge_temperature, :output_column, :expected_column, :omit_temperature, metric_ids: [], tag_names: [])
-      omit = permitted.delete(:omit_temperature)
-      permitted[:temperature] = nil if omit == "1"
-      permitted
+      params.require(:run).permit(:name, :prompt_id, :dataset_id, :judge_model, :temperature, :max_tokens, :judge_temperature, :output_column, :expected_column, metric_ids: [], tag_names: [])
     end
 
     # Editing a run that already has results forks a new run — but only when a

data/app/services/completion_kit/llm_client.rb

@@ -14,11 +14,16 @@ module CompletionKit
     # recoverable refusal into a failed row.
     #
     # The wording can sit on either side of the parameter name, so both orders
-    # carry the same phrase set. The gap cannot cross a brace, which keeps a
-    # refusal aimed at some other parameter from matching a temperature echoed
-    # back in a neighbouring object of the same error body. That would
-    # otherwise strip a temperature the model was perfectly happy with and
-    # flag the run as having had it ignored.
+    # carry the same phrase set, and a provider that names the offending field
+    # outright gets a branch of its own.
+    #
+    # Neither gap may cross a brace, which stops a refusal aimed at some other
+    # parameter from matching a temperature echoed back in a neighbouring
+    # object of the same body. The phrase-first gap additionally bars commas,
+    # because "unsupported parameter: logprobs. Supported: model, prompt,
+    # temperature" is a list of what IS allowed and must not read as a refusal.
+    # The temperature-first gap has to allow commas, since "temperature, top_p
+    # and top_k are not supported" is a genuine refusal.
     REFUSAL_PHRASE = /
       deprecated | not\s+supported | does\s+not\s+support |
       only\s+the\s+default | unsupported\s+(?:parameter|value)
@@ -26,7 +31,9 @@ module CompletionKit
     TEMPERATURE_REFUSAL = /
       temperature [^{}]{0,80}? #{REFUSAL_PHRASE}
       |
-      #{REFUSAL_PHRASE} [^{}]{0,80}? temperature
+      #{REFUSAL_PHRASE} [^{},]{0,80}? temperature
+      |
+      "param" \s*:\s* "temperature"
     /xi
 
     def initialize(config = {})

data/app/services/completion_kit/mcp_tools/runs.rb

@@ -3,8 +3,12 @@ module CompletionKit
     module Runs
       extend Base
 
-      TEMPERATURE_DESCRIPTION = "Sampling temperature for generation, 0 to 1. Defaults to the column default. " \
-                                "Reasoning models ignore it and the run is flagged temperature_ignored.".freeze
+      TEMPERATURE_DESCRIPTION = "Sampling temperature for generation, 0 to 1. Leave it unset, which is the " \
+                                "default, and no temperature is sent at all, so the model applies its own. " \
+                                "Most current frontier models refuse the parameter outright; set it only when " \
+                                "you are targeting a model that honours it, such as anything served locally " \
+                                "through Ollama. A refused value is re-sent without one and the run is " \
+                                "flagged temperature_ignored.".freeze
 
       MAX_TOKENS_DESCRIPTION = "Cap on generated tokens per row. Leave unset to use the provider client's default, " \
                                "which is what silently truncates long outputs and makes the judge score malformed " \

data/app/views/completion_kit/runs/_form.html.erb

@@ -83,25 +83,6 @@
       <p class="ck-field-hint" id="expected-column-warn"></p>
     </div>
 
-    <div class="ck-field">
-      <label class="ck-label" for="run_temperature" style="position: relative;">
-        Temperature<span class="ck-info-toggle" tabindex="0">?</span><span class="ck-info-popup">Controls how random the model's output is. Lower values are more focused and deterministic, so the model picks the most likely words. Higher values are more varied and creative, with more risk of odd phrasing. Most LLMs default to 1.0; for evaluation, try a few values and see how your prompt holds up. Newer reasoning models (Claude Opus 4.7, GPT-5 family, etc.) refuse temperature outright. CompletionKit detects that and re-sends without the parameter, and flags the run so you know the value was never applied. Tick the box below to skip sending it in the first place.</span>
-      </label>
-      <div class="ck-slider-row">
-        <%= form.range_field :temperature, min: 0, max: 1, step: 0.1, class: "ck-slider", id: "run_temperature", disabled: run.temperature.nil?, oninput: "document.getElementById('temp-value').textContent = this.value" %>
-        <span class="ck-slider-value" id="temp-value"><%= run.temperature.nil? ? "off" : run.temperature %></span>
-      </div>
-      <label class="ck-checkbox-label">
-        <%= check_box_tag "run[omit_temperature]", "1", run.temperature.nil?, id: "run_omit_temperature", class: "ck-checkbox",
-              onchange: "document.getElementById('run_temperature').disabled = this.checked; document.getElementById('temp-value').textContent = this.checked ? 'off' : document.getElementById('run_temperature').value;" %>
-        <span class="ck-checkbox-label__box" aria-hidden="true"></span>
-        <span class="ck-checkbox-label__body">
-          <span class="ck-checkbox-label__text">Send no temperature</span>
-          <span class="ck-checkbox-label__hint">The model applies its own default. Required by models that refuse the parameter, which most current frontier models do.</span>
-        </span>
-      </label>
-    </div>
-
     <div class="ck-field">
       <label class="ck-label" for="run_max_tokens" style="position: relative;">
         Max tokens<span class="ck-info-toggle" tabindex="0">?</span><span class="ck-info-popup">The ceiling on how many tokens the model may generate per row. Leave it blank to use the provider's default. Set it when your prompt produces long output: if generation hits the cap the response is cut off mid-sentence, and the judge scores the truncation rather than the prompt. Match whatever your production caller uses so the evaluation reflects reality.</span>
@@ -135,14 +116,6 @@
       <% end %>
     </div>
 
-    <div class="ck-field" id="judge-temperature-field">
-      <label class="ck-label" for="run_judge_temperature" style="position: relative;">
-        Judge temperature<span class="ck-info-toggle" tabindex="0">?</span><span class="ck-info-popup">How random the judge is when it scores. This is separate from the generation temperature above. Keep it at 0 so re-judging the same output gives the same score, which is what makes one run comparable to another. Raise it only when you are deliberately measuring how much the judge disagrees with itself.</span>
-      </label>
-      <%= form.number_field :judge_temperature, in: 0..1, step: 0.1, class: "ck-input", id: "run_judge_temperature", **ck_field_aria(form, :judge_temperature) %>
-      <%= ck_field_error(form, :judge_temperature) %>
-      <p class="ck-field-hint" id="judge-temperature-hint">0 is deterministic and recommended. Above 0, the same output can score differently on a re-judge.</p>
-    </div>
 
     <div class="ck-field" id="metrics-field">
       <p class="ck-label" id="run_metrics_label">Metrics</p>

data/app/views/completion_kit/runs/_run_config.html.erb

@@ -0,0 +1,57 @@
+<div class="ck-run-config" id="run_config">
+  <div class="ck-run-config__row">
+    <span class="ck-run-config__key">Created</span>
+    <time datetime="<%= run.created_at.iso8601 %>" data-local-time><%= run.created_at.utc.strftime("%Y-%m-%d %H:%M UTC") %></time>
+  </div>
+  <div class="ck-run-config__row">
+    <span class="ck-run-config__key">Dataset</span>
+    <% if run.dataset %>
+      <span class="ck-run-config__dataset">
+        <%= link_to run.dataset.name, dataset_path(run.dataset), class: "ck-link" %>
+        <span class="ck-run-config__dataset-meta"><%= run.dataset.row_count %> rows</span>
+        <button type="button" class="ck-run-config__dataset-preview" onclick="document.getElementById('dataset-preview-<%= run.id %>').showModal()">Preview</button>
+      </span>
+    <% else %>
+      <span class="ck-run-config__none">None</span>
+    <% end %>
+  </div>
+  <% if run.judge_model.present? %>
+    <div class="ck-run-config__row">
+      <span class="ck-run-config__key">Judge</span>
+      <span style="text-transform: none;"><%= run.judge_model %></span>
+      <% unless run.judge_configured? %>
+        <span class="ck-run-config__warn">provider not configured</span>
+      <% end %>
+    </div>
+  <% end %>
+  <% if run.metrics.any? %>
+    <div class="ck-run-config__row">
+      <span class="ck-run-config__key">Metrics</span>
+      <span><%= run.metrics.map { |m| link_to(m.name, metric_path(m), class: "ck-link") }.join(", ").html_safe %></span>
+    </div>
+  <% end %>
+  <% if run.temperature.present? %>
+    <div class="ck-run-config__row">
+      <span class="ck-run-config__key">Temperature</span>
+      <span><%= run.temperature %></span>
+      <% if run.temperature_ignored? %>
+        <span class="ck-run-config__warn" style="color: var(--ck-dim);" title="The model rejected the temperature parameter, so CompletionKit re-sent the request without it.">ignored by model</span>
+      <% end %>
+    </div>
+  <% end %>
+  <div class="ck-run-config__row">
+    <span class="ck-run-config__key">Max tokens</span>
+    <span><%= run.max_tokens || "Provider default" %></span>
+  </div>
+  <% if run.nondeterministic_judge? %>
+    <div class="ck-run-config__row">
+      <span class="ck-run-config__key">Judge temperature</span>
+      <span><%= run.judge_temperature %></span>
+      <% if run.judge_temperature_ignored? %>
+        <span class="ck-run-config__warn" style="color: var(--ck-warning);" title="The judge model refused the temperature parameter, so CompletionKit re-sent the request without it and the provider applied its own default. The scores are not reproducible despite the setting above.">refused by judge, not reproducible</span>
+      <% else %>
+        <span class="ck-run-config__warn" style="color: var(--ck-warning);" title="Judging above temperature 0 makes scores irreproducible. The same output can get a different score on a re-judge.">scores not reproducible</span>
+      <% end %>
+    </div>
+  <% end %>
+</div>

data/app/views/completion_kit/runs/show.html.erb

@@ -50,59 +50,7 @@
   <% end %>
 <% end %>
 
-<div class="ck-run-config">
-  <div class="ck-run-config__row">
-    <span class="ck-run-config__key">Created</span>
-    <time datetime="<%= @run.created_at.iso8601 %>" data-local-time><%= @run.created_at.utc.strftime("%Y-%m-%d %H:%M UTC") %></time>
-  </div>
-  <div class="ck-run-config__row">
-    <span class="ck-run-config__key">Dataset</span>
-    <% if @run.dataset %>
-      <span class="ck-run-config__dataset">
-        <%= link_to @run.dataset.name, dataset_path(@run.dataset), class: "ck-link" %>
-        <span class="ck-run-config__dataset-meta"><%= @run.dataset.row_count %> rows</span>
-        <button type="button" class="ck-run-config__dataset-preview" onclick="document.getElementById('dataset-preview-<%= @run.id %>').showModal()">Preview</button>
-      </span>
-    <% else %>
-      <span class="ck-run-config__none">None</span>
-    <% end %>
-  </div>
-  <% if @run.judge_model.present? %>
-    <div class="ck-run-config__row">
-      <span class="ck-run-config__key">Judge</span>
-      <span style="text-transform: none;"><%= @run.judge_model %></span>
-      <% unless @run.judge_configured? %>
-        <span class="ck-run-config__warn">provider not configured</span>
-      <% end %>
-    </div>
-  <% end %>
-  <% if @run.metrics.any? %>
-    <div class="ck-run-config__row">
-      <span class="ck-run-config__key">Metrics</span>
-      <span><%= @run.metrics.map { |m| link_to(m.name, metric_path(m), class: "ck-link") }.join(", ").html_safe %></span>
-    </div>
-  <% end %>
-  <div class="ck-run-config__row">
-    <span class="ck-run-config__key">Temperature</span>
-    <span><%= @run.temperature %></span>
-    <% if @run.temperature_ignored? %>
-      <span class="ck-run-config__warn" style="color: var(--ck-dim);" title="The model rejected the temperature parameter, so CompletionKit re-sent the request without it.">ignored by model</span>
-    <% end %>
-  </div>
-  <div class="ck-run-config__row">
-    <span class="ck-run-config__key">Max tokens</span>
-    <span><%= @run.max_tokens || "Provider default" %></span>
-  </div>
-  <div class="ck-run-config__row">
-    <span class="ck-run-config__key">Judge temperature</span>
-    <span><%= @run.judge_temperature %></span>
-    <% if @run.judge_temperature_ignored? %>
-      <span class="ck-run-config__warn" style="color: var(--ck-warning);" title="The judge model refused the temperature parameter, so CompletionKit re-sent the request without it and the provider applied its own default. The scores are not reproducible despite the setting above.">refused by judge, not reproducible</span>
-    <% elsif @run.nondeterministic_judge? %>
-      <span class="ck-run-config__warn" style="color: var(--ck-warning);" title="Judging above temperature 0 makes scores irreproducible. The same output can get a different score on a re-judge.">scores not reproducible</span>
-    <% end %>
-  </div>
-</div>
+<%= render "run_config", run: @run %>
 
 <% if @run.prompt %>
   <div class="ck-prompt-preview">

data/db/migrate/20260804000002_default_runs_to_no_temperature.rb

@@ -0,0 +1,9 @@
+class DefaultRunsToNoTemperature < ActiveRecord::Migration[8.0]
+  def up
+    change_column_default :completion_kit_runs, :temperature, nil
+  end
+
+  def down
+    change_column_default :completion_kit_runs, :temperature, 1.0
+  end
+end

data/lib/completion_kit/version.rb

@@ -1,3 +1,3 @@
 module CompletionKit
-  VERSION = "0.28.35"
+  VERSION = "0.28.37"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: completion-kit
 version: !ruby/object:Gem::Version
-  version: 0.28.35
+  version: 0.28.37
 platform: ruby
 authors:
 - Damien Bastin
@@ -373,6 +373,7 @@ files:
 - app/views/completion_kit/runs/_response_row.html.erb
 - app/views/completion_kit/runs/_responses_region.html.erb
 - app/views/completion_kit/runs/_row.html.erb
+- app/views/completion_kit/runs/_run_config.html.erb
 - app/views/completion_kit/runs/_sort_toolbar.html.erb
 - app/views/completion_kit/runs/_status_header.html.erb
 - app/views/completion_kit/runs/_status_panel.html.erb
@@ -448,6 +449,7 @@ files:
 - db/migrate/20260730000001_add_judge_temperature_to_completion_kit_runs.rb
 - db/migrate/20260730000002_create_completion_kit_prompt_serves.rb
 - db/migrate/20260804000001_add_judge_temperature_ignored_to_runs.rb
+- db/migrate/20260804000002_default_runs_to_no_temperature.rb
 - lib/completion-kit.rb
 - lib/completion_kit.rb
 - lib/completion_kit/concurrency_check.rb