ruby_llm-tribunal 0.1.0 → 0.1.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0cfe5bd072c4cc3499736cf095cdc4faea9778bb0feb368ac13c6735cd6239ce
-  data.tar.gz: 6730343af6bd441998357fdc5a56c13ba5a3b1e226877e0d77d704947fe84883
+  metadata.gz: 95f293e61287cbd2f94763540bf159667d3e02223e0c8950d0f5e46622e1be40
+  data.tar.gz: b0e0c398b24b7847d3e4defd7f2200733493621f44f65129a818e0bef6c08894
 SHA512:
-  metadata.gz: aacf8935874a75b51fcc3e6cd63b3d65d01b9f437bfc6b86a3fb496d50163b9fa122e615cbe3bd5177b8c05dd3305aadbac779385637203b1fc0e142099026a4
-  data.tar.gz: 44415c718a94108c0f7416054dd4e6f2ecc4ee05647272b176402ff0419ad23efb547f5dd7de088a31027003d0c2a32ce508bcb8bca3be77dd23d74f51e25b69
+  metadata.gz: 05ee643802ca6de43a4cf89bbb5e9a528030c663e69097fadd26a15543be3b44bfa520956a6f1033c8add68cc505c92043786febcb7fe2b7c77972580102c541
+  data.tar.gz: 231b545756fca7e53210c9bc4598cc10ae5da16e8a576a6d2801c59c11bb65a4378c252b10bd269369530de26ece1f51e923342bc512cf73ea4b2df667787310

data/CHANGELOG.md

@@ -6,6 +6,17 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 
 ## [Unreleased]
 
+## [0.1.1] - 2026-01-16
+
+### Fixed
+
+- **Critical**: Fixed incorrect threshold logic for negative metrics (toxicity, hallucination, bias, etc.) with `partial` verdicts. High scores on negative metrics now correctly result in failures.
+- **Concurrency**: The `--concurrency` option now gracefully falls back to sequential execution when the `parallel` gem is not installed, with a helpful warning message.
+
+### Added
+
+- Tests for negative metric edge cases (partial verdicts with inverted threshold logic)
+
 ## [0.1.0] - 2026-01-15
 
 ### Added
@@ -28,5 +39,6 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Requires `ruby_llm` >= 1.0
 - Optional: `neighbor` gem for embedding-based similarity
 
-[Unreleased]: https://github.com/Alqemist-labs/ruby_llm-tribunal/compare/v0.1.0...HEAD
+[Unreleased]: https://github.com/Alqemist-labs/ruby_llm-tribunal/compare/v0.1.1...HEAD
+[0.1.1]: https://github.com/Alqemist-labs/ruby_llm-tribunal/compare/v0.1.0...v0.1.1
 [0.1.0]: https://github.com/Alqemist-labs/ruby_llm-tribunal/releases/tag/v0.1.0

data/lib/ruby_llm/tribunal/assertions/judge.rb

@@ -140,7 +140,12 @@ module RubyLLM
             passed = case verdict
                      when 'yes' then !negative_metric
                      when 'no' then negative_metric
-                     when 'partial' then score.is_a?(Numeric) && score >= threshold
+                     when 'partial'
+                       return [:fail, details] unless score.is_a?(Numeric)
+
+                       # For negative metrics (toxicity, hallucination, etc.), high score = bad
+                       # So we invert the comparison: pass if score is LOW enough
+                       negative_metric ? score <= (1.0 - threshold) : score >= threshold
                      end
 
             passed ? [:pass, details] : [:fail, details]

data/lib/ruby_llm/tribunal/reporters/github.rb

@@ -10,8 +10,8 @@ module RubyLLM
             annotations = results[:cases]
                           .select { |c| c[:status] == :failed }
                           .map do |c|
-                            reasons = c[:failures].map { |type, reason| "#{type}: #{reason}" }.join('; ')
-                            "::error::#{c[:input]}: #{reasons}"
+              reasons = c[:failures].map { |type, reason| "#{type}: #{reason}" }.join('; ')
+              "::error::#{c[:input]}: #{reasons}"
             end
 
             summary = "::notice::Tribunal: #{results[:summary][:passed]}/#{results[:summary][:total]} passed " \

data/lib/ruby_llm/tribunal/version.rb

@@ -2,6 +2,6 @@
 
 module RubyLLM
   module Tribunal
-    VERSION = '0.1.0'
+    VERSION = '0.1.1'
   end
 end

data/lib/tasks/tribunal.rake

@@ -133,9 +133,15 @@ namespace :tribunal do
     cases = RubyLLM::Tribunal::Dataset.load_with_assertions(path)
 
     if concurrency > 1
-      require 'parallel'
-      Parallel.map(cases, in_threads: concurrency) do |test_case, assertions|
-        run_case(test_case, assertions, provider)
+      begin
+        require 'parallel'
+        Parallel.map(cases, in_threads: concurrency) do |test_case, assertions|
+          run_case(test_case, assertions, provider)
+        end
+      rescue LoadError
+        warn "Warning: 'parallel' gem not installed, falling back to sequential execution."
+        warn '  Install with: gem install parallel'
+        cases.map { |test_case, assertions| run_case(test_case, assertions, provider) }
       end
     else
       cases.map { |test_case, assertions| run_case(test_case, assertions, provider) }

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm-tribunal
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.1.1
 platform: ruby
 authors:
 - Florian
-autorequire: 
+autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-01-15 00:00:00.000000000 Z
+date: 2026-01-16 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ruby_llm
@@ -77,7 +77,7 @@ metadata:
   source_code_uri: https://github.com/Alqemist-labs/ruby_llm-tribunal
   changelog_uri: https://github.com/Alqemist-labs/ruby_llm-tribunal/blob/main/CHANGELOG.md
   rubygems_mfa_required: 'true'
-post_install_message: 
+post_install_message:
 rdoc_options: []
 require_paths:
 - lib
@@ -92,8 +92,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.0.3.1
-signing_key: 
+rubygems_version: 3.4.19
+signing_key:
 specification_version: 4
 summary: LLM evaluation framework for Ruby
 test_files: []