legion-llm 0.8.19 → 0.8.21

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e7ec5141c77d40965d835a36858aa44893158036c4a8052768edb19a181868d9
-  data.tar.gz: da2934e668432b45ef002979ce4b683b988fcf4ea72aa0d420ddc781c0c6781b
+  metadata.gz: 5fdd5beb0c0e3a464cd06848369ec4cf9c07beab7a9c1560cfc0d13e1a6dc721
+  data.tar.gz: 2539e7e1a1cebcd9ba363ac66f79c08f9314d7477d482f1d17a7bf6185e1fd0b
 SHA512:
-  metadata.gz: a6b6300de67bd0a3895b443c29f59b38f2e11fa80e84b4d327becddefcc1dcb4e62470d9ff304a240ee241b35450961fff793a24d8a3ce752d1befa1d9e8f7d7
-  data.tar.gz: 9a40fc3b19c9667b729b3959efccb270b799084b2eee6023e6c87b9103c92ea2f345a65a0abbd8a43805fece5754bc204e73571d47b62fad2100ac12cee647c2
+  metadata.gz: 377c8498427bbe07d0dd171ce8fba340683ac8fb0286b781092caa24eb19c038fc9e621115bfcbfe12aee25c1f84aa267b4c289b03cd9424a2beee0d31ccf128
+  data.tar.gz: 611ab93b2732f10aaac6015b2fc042af6de8963b23012c73a00a6c6a36477639a61763ace1be0e65eb2f38c815cd0418c9d118a644090324a6de53b98cc2f5a6

data/CHANGELOG.md

@@ -1,5 +1,16 @@
 # Legion LLM Changelog
 
+## [0.8.21] - 2026-04-22
+
+### Fixed
+- Tool audit events are now published to `llm.audit` exchange via `Audit.emit_tools` after each tool execution completes. Previously `emit_tools` was defined but never called — the `llm.audit.tools` queue was always empty.
+- Metering events now include `request_type: 'chat'` so the routing key is `metering.chat` instead of `metering.` (empty suffix).
+
+## [0.8.20] - 2026-04-22
+
+### Fixed
+- `tool_trigger_defaults[:tool_limit]` raised from 10 to 25. The v0.8.19 reduction from 50→10 was too aggressive — extension tools (Teams, etc.) are only injected via trigger matching, and a limit of 10 crowded them out when multiple extensions matched.
+
 ## [0.8.19] - 2026-04-22
 
 ### Fixed

data/lib/legion/llm/api/native/inference.rb

@@ -179,6 +179,7 @@ module Legion
                     "conversation_id=#{pipeline_response.conversation_id || conversation_id || 'none'} " \
                     "provider=#{routing[:provider] || routing['provider'] || 'unknown'} " \
                     "model=#{routing[:model] || routing['model'] || 'unknown'} " \
+                    "input_tokens=#{token_value(tokens, :input) || 0} output_tokens=#{token_value(tokens, :output) || 0} " \
                     "tool_calls=#{extract_tool_calls(pipeline_response).size} " \
                     "tool_executions=#{Array(pipeline_response.timeline).count { |event| event[:key].to_s.start_with?('tool:execute:') }} " \
                     "stop_reason=#{pipeline_response.stop&.dig(:reason) || 'unknown'} stream=true"
@@ -204,6 +205,7 @@ module Legion
                   "conversation_id=#{pipeline_response.conversation_id || conversation_id || 'none'} " \
                   "provider=#{routing[:provider] || routing['provider'] || 'unknown'} " \
                   "model=#{routing[:model] || routing['model'] || 'unknown'} " \
+                  "input_tokens=#{token_value(tokens, :input) || 0} output_tokens=#{token_value(tokens, :output) || 0} " \
                   "tool_calls=#{tool_calls.size} " \
                   "tool_executions=#{Array(pipeline_response.timeline).count { |event| event[:key].to_s.start_with?('tool:execute:') }} " \
                   "stop_reason=#{pipeline_response.stop&.dig(:reason) || 'unknown'} stream=false"

data/lib/legion/llm/inference/executor.rb

@@ -842,6 +842,33 @@ module Legion
             result: result_str[0, 4096], result_size: result_str.bytesize,
             started_at: started_at, finished_at: finished_at, duration_ms: duration_ms
           )
+
+          publish_tool_audit(tc_id, tc_name, result_str, is_error, duration_ms, started_at, finished_at)
+        end
+
+        def publish_tool_audit(tc_id, tc_name, result_str, is_error, duration_ms, started_at, finished_at)
+          Legion::LLM::Audit.emit_tools(
+            request_id:      @request.id,
+            conversation_id: @request.conversation_id,
+            exchange_id:     @exchange_id,
+            tool_name:       tc_name,
+            tool_call:       {
+              id:          tc_id,
+              name:        tc_name,
+              status:      is_error ? :error : :success,
+              duration_ms: duration_ms,
+              started_at:  started_at,
+              finished_at: finished_at
+            },
+            result:          result_str[0, 4096],
+            caller:          @request.caller,
+            classification:  @request.classification,
+            tracing:         @tracing,
+            timestamp:       finished_at,
+            request_type:    'tool'
+          )
+        rescue StandardError => e
+          handle_exception(e, level: :warn, operation: 'llm.pipeline.publish_tool_audit', tool_name: tc_name)
         end
 
         def tool_call_field(tool_call, field)
@@ -1004,6 +1031,7 @@ module Legion
             provider:      @resolved_provider,
             model_id:      @resolved_model,
             tier:          tier,
+            request_type:  'chat',
             input_tokens:  input_tokens,
             output_tokens: output_tokens,
             latency_ms:    latency_ms

data/lib/legion/llm/settings.rb

@@ -285,7 +285,7 @@ module Legion
       def self.tool_trigger_defaults
         {
           scan_depth: 10,
-          tool_limit: 10
+          tool_limit: 25
         }
       end
 

data/lib/legion/llm/version.rb

@@ -2,6 +2,6 @@
 
 module Legion
   module LLM
-    VERSION = '0.8.19'
+    VERSION = '0.8.21'
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: legion-llm
 version: !ruby/object:Gem::Version
-  version: 0.8.19
+  version: 0.8.21
 platform: ruby
 authors:
 - Esity