completion-kit 0.27.7 → 0.27.9

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 69f57e2ec1d9363a088d85a4ecf6c3632c1f971e08a8df351c0b2bbbd9b1ba16
-  data.tar.gz: 8becf1e826e0a1dd537af4d4e762154715d693a719bd97a7464c2d54beb4fe78
+  metadata.gz: 3c963af7241cc97480aa462a422e3c34fb2f25ea813a380083443e68678f724b
+  data.tar.gz: 8a20da4fd8a0fac0e6395816cd65a4bc79ebf48756439414a343467f41ab2c2b
 SHA512:
-  metadata.gz: ebb8b90edfe00dbffc9b7339594fefb8884adbf04fa1e26bc6a42dcae66e17aaf3d77c03b873d8d5e91ab4edb6dbcb3e4c3b142c17170a4860f3b307510907cd
-  data.tar.gz: 5ec296ba854f8ad3162f67303d8898d1db83831ad9a443b5b20acf473b53c7ee97bb74d8ca4a213e0a27eb5fe6b55e68bb39c2efc7379e44029224337f45e269
+  metadata.gz: a9575cfcb4ef088a9e34c92d9852341f69eff3130f2d0e7845fc6517ce939dd9f8252c7c7722946b106de72dc2ad281388a0fc4a63134d8d8e69b9d037506825
+  data.tar.gz: 4050b952aea1c85a1871de868218fb9637cd662d30e4b1508f47ef2920f5d7e57c12b21b8a09c6cc7d0bc6e718f05032938ae6b47733767fcb458d3ccf374949

data/app/models/completion_kit/review.rb

@@ -21,10 +21,10 @@ module CompletionKit
     end
 
     def stale_against_current_judge?
-      return false unless metric_id && metric_version_id
-      current_id = MetricVersion.current.where(metric_id: metric_id).limit(1).pick(:id)
-      return false if current_id.nil?
-      metric_version_id != current_id
+      return false unless metric_id && metric_version
+      current_number = MetricVersion.current.where(metric_id: metric_id).limit(1).pick(:version_number)
+      return false if current_number.nil?
+      metric_version.version_number != current_number
     end
 
     def as_json(options = {})

data/app/models/completion_kit/run.rb

@@ -166,9 +166,9 @@ module CompletionKit
       review_pairs.each do |metric_id, version_id, metric_name|
         current = current_by_metric[metric_id]
         next if current.nil?
-        next if version_id == current[:id]
         label = label_by_version[version_id]
         next if label.nil?
+        next if label == current[:label]
         summary[metric_id] ||= { metric_name: metric_name, current_label: current[:label], stale_count: 0, scored_labels: [] }
         summary[metric_id][:stale_count] += 1
         summary[metric_id][:scored_labels] |= [label]

data/app/views/completion_kit/api_reference/_body.html.erb

@@ -204,7 +204,7 @@
       <p class="ck-copy">Data used as input for runs.</p>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">GET</span> /api/v1/datasets</p>
-        <p class="ck-meta-copy">List all datasets.</p>
+        <p class="ck-meta-copy">List all datasets. Supports pagination (<code>limit</code>, <code>offset</code>) and <code>tag[]</code> filtering.</p>
       </div>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">POST</span> /api/v1/datasets</p>
@@ -231,7 +231,7 @@
       <p class="ck-copy">Scoring dimensions used by the judge model.</p>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">GET</span> /api/v1/metrics</p>
-        <p class="ck-meta-copy">List all metrics.</p>
+        <p class="ck-meta-copy">List all metrics. Supports pagination (<code>limit</code>, <code>offset</code>) and <code>tag[]</code> filtering.</p>
       </div>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">POST</span> /api/v1/metrics</p>
@@ -289,7 +289,7 @@
       <p class="ck-copy">Named groups of metrics you can apply to a run as a set.</p>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">GET</span> /api/v1/metric_groups</p>
-        <p class="ck-meta-copy">List all metric groups with their metric IDs.</p>
+        <p class="ck-meta-copy">List all metric groups with their metric IDs. Supports pagination (<code>limit</code>, <code>offset</code>) and <code>tag[]</code> filtering.</p>
       </div>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">POST</span> /api/v1/metric_groups</p>
@@ -337,7 +337,7 @@
 
     <div class="ck-api-tabs__panel">
       <h2 class="ck-section-title">Tags</h2>
-      <p class="ck-copy">Domain labels you can attach to metrics, prompts, runs, datasets, and metric groups. Tags are auto-assigned a color from a 10-color palette. Each index page can be filtered by one or more tags using <code>?tag[]=name</code> query params (OR semantics).</p>
+      <p class="ck-copy">Domain labels you can attach to metrics, prompts, runs, datasets, and metric groups. Tags are auto-assigned a color from a 10-color palette. Each of those five resources' list endpoints can be filtered by one or more tags using <code>?tag[]=name</code> query params (OR semantics: a record matching any of the given tags is returned).</p>
       <div class="ck-api-endpoint">
         <p class="ck-api-method"><span class="ck-chip ck-chip--soft">GET</span> /api/v1/tags</p>
         <p class="ck-meta-copy">List all tags with name and color.</p>

data/lib/completion_kit/version.rb

@@ -1,3 +1,3 @@
 module CompletionKit
-  VERSION = "0.27.7"
+  VERSION = "0.27.9"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: completion-kit
 version: !ruby/object:Gem::Version
-  version: 0.27.7
+  version: 0.27.9
 platform: ruby
 authors:
 - Damien Bastin