legion-llm 0.3.3 → 0.3.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7faf26458139d4c0e585e5c30e42602e85402e20abcbe2cd73ef8449ae17f947
-  data.tar.gz: 73fecb93dbfd407891e64a4278c21c9e9f7fcc2af8313660de4e075a3195bbc7
+  metadata.gz: e3e10dfcd60fe722290bec30017671fb261f3baf151b416c82defb082d4445f4
+  data.tar.gz: 0c0b062649f8ede281fada374681d551d437126a2efa0e604a069610e14b7069
 SHA512:
-  metadata.gz: 067c7e99927b675df13517a6ca5aa12b494fdedd8e277e7a42ae060e7597033597fbd0f88f595d5368937ba4a00aa7f6e2e02c14e56d01763f4253eb1cd3f421
-  data.tar.gz: 9dc754975461db838b49d1f9826d54a80c7bdf106fa44fa6fe3d3693eefe70a1d503ed5025e551a458b367ab53580c67b7c9cf1fb9ee01e69cd5c2394181f150
+  metadata.gz: eeb2cd074c2eb1c3b63ccb7644adbcf7cac6bab62f8d5cc966e318b2185267ab73fae920726b83e1f72cbf8753ba1245ab84ae914baa092aebdbef08e0548cd3
+  data.tar.gz: ccc52360f869421100f0bbda503570168f2f7eb86c5fda9069e6ee29bdaa4c60553d202a1ad2e2109e98ba451b9abf9c00dc8210b50cf5d0a925192cada2ab9d

data/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Legion LLM Changelog
 
+## [0.3.5] - 2026-03-18
+
+### Added
+- Gateway integration: `chat`, `embed`, `structured` delegate to `lex-llm-gateway` when loaded for automatic metering and fleet dispatch
+- `chat_direct`, `embed_direct`, `structured_direct` methods bypass gateway (used by gateway runners to avoid recursion)
+- Gateway integration spec (8 examples)
+
+## [0.3.4] - 2026-03-18
+
+### Added
+- Auto-configure LLM providers from environment variables (`AWS_BEARER_TOKEN_BEDROCK`, `ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, `CODEX_API_KEY`, `GEMINI_API_KEY`)
+- `ANTHROPIC_MODEL` env var sets default model for Anthropic and Bedrock providers
+- Import Claude CLI config from `~/.claude/settings.json` and `~/.claude.json`
+- Auto-detect Ollama via local port probe (no env var needed)
+- Auto-enable providers when credentials are found in environment
+
 ## [0.3.3] - 2026-03-17
 
 ### Added

data/CLAUDE.md

@@ -284,7 +284,7 @@ In-memory signal consumer with pluggable handlers. Adjusts effective priorities
 | `lib/legion/llm/embeddings.rb` | Embeddings module: generate, generate_batch, default_model |
 | `lib/legion/llm/shadow_eval.rb` | Shadow evaluation: enabled?, should_sample?, evaluate, compare |
 | `lib/legion/llm/structured_output.rb` | JSON schema enforcement with native response_format and prompt fallback |
-| `lib/legion/llm/version.rb` | Version constant (0.3.2) |
+| `lib/legion/llm/version.rb` | Version constant (0.3.3) |
 | `lib/legion/llm/quality_checker.rb` | QualityChecker module with QualityResult struct |
 | `lib/legion/llm/escalation_history.rb` | EscalationHistory mixin: `escalation_history`, `escalated?`, `final_resolution`, `escalation_chain` |
 | `lib/legion/llm/router/escalation_chain.rb` | EscalationChain value object |

data/Gemfile

@@ -4,6 +4,8 @@ source 'https://rubygems.org'
 
 gemspec
 
+gem 'lex-llm-gateway', path: '../extensions-core/lex-llm-gateway' if File.directory?('../extensions-core/lex-llm-gateway')
+
 group :test do
   gem 'rake'
   gem 'rspec'

data/lib/legion/llm/claude_config_loader.rb

@@ -0,0 +1,59 @@
+# frozen_string_literal: true
+
+module Legion
+  module LLM
+    module ClaudeConfigLoader
+      CLAUDE_SETTINGS = File.expand_path('~/.claude/settings.json')
+      CLAUDE_CONFIG   = File.expand_path('~/.claude.json')
+
+      module_function
+
+      def load
+        config = read_json(CLAUDE_SETTINGS).merge(read_json(CLAUDE_CONFIG))
+        return if config.empty?
+
+        apply_claude_config(config)
+      end
+
+      def read_json(path)
+        return {} unless File.exist?(path)
+
+        require 'json'
+        ::JSON.parse(File.read(path), symbolize_names: true)
+      rescue StandardError
+        {}
+      end
+
+      def apply_claude_config(config)
+        apply_api_keys(config)
+        apply_model_preference(config)
+      end
+
+      def apply_api_keys(config)
+        llm = Legion::LLM.settings
+        providers = llm[:providers]
+
+        if config[:anthropicApiKey] && providers.dig(:anthropic, :api_key).nil?
+          providers[:anthropic][:api_key] = config[:anthropicApiKey]
+          Legion::Logging.debug 'Imported Anthropic API key from Claude CLI config'
+        end
+
+        return unless config[:openaiApiKey] && providers.dig(:openai, :api_key).nil?
+
+        providers[:openai][:api_key] = config[:openaiApiKey]
+        Legion::Logging.debug 'Imported OpenAI API key from Claude CLI config'
+      end
+
+      def apply_model_preference(config)
+        return unless config[:preferredModel] || config[:model]
+
+        model = config[:preferredModel] || config[:model]
+        llm = Legion::LLM.settings
+        return if llm[:default_model]
+
+        llm[:default_model] = model
+        Legion::Logging.debug "Imported model preference from Claude CLI config: #{model}"
+      end
+    end
+  end
+end

data/lib/legion/llm/providers.rb

@@ -4,6 +4,7 @@ module Legion
   module LLM
     module Providers
       def configure_providers
+        auto_enable_from_resolved_credentials
         settings[:providers].each do |provider, config|
           next unless config[:enabled]
 
@@ -11,6 +12,37 @@ module Legion
         end
       end
 
+      def auto_enable_from_resolved_credentials
+        settings[:providers].each do |provider, config|
+          next if config[:enabled]
+
+          has_creds = case provider
+                      when :bedrock
+                        config[:bearer_token] || (config[:api_key] && config[:secret_key])
+                      when :ollama
+                        ollama_running?(config)
+                      else
+                        config[:api_key]
+                      end
+          next unless has_creds
+
+          config[:enabled] = true
+          Legion::Logging.info "Auto-enabled #{provider} provider (credentials found)"
+        end
+      end
+
+      def ollama_running?(config)
+        require 'socket'
+        url = config[:base_url] || 'http://localhost:11434'
+        host_part = url.gsub(%r{^https?://}, '').split(':')
+        addr = host_part[0]
+        port = (host_part[1] || '11434').to_i
+        Socket.tcp(addr, port, connect_timeout: 1).close
+        true
+      rescue StandardError
+        false
+      end
+
       def apply_provider_config(provider, config)
         case provider
         when :bedrock

data/lib/legion/llm/settings.rb

@@ -4,10 +4,11 @@ module Legion
   module LLM
     module Settings
       def self.default
+        model_override = ENV.fetch('ANTHROPIC_MODEL', nil)
         {
           enabled:          true,
           connected:        false,
-          default_model:    nil,
+          default_model:    model_override,
           default_provider: nil,
           providers:        providers,
           routing:          routing_defaults,
@@ -68,23 +69,23 @@ module Legion
             api_key:       nil,
             secret_key:    nil,
             session_token: nil,
-            bearer_token:  nil,
+            bearer_token:  'env://AWS_BEARER_TOKEN_BEDROCK',
             region:        'us-east-2'
           },
           anthropic: {
             enabled:       false,
             default_model: 'claude-sonnet-4-6',
-            api_key:       nil
+            api_key:       'env://ANTHROPIC_API_KEY'
           },
           openai:    {
             enabled:       false,
             default_model: 'gpt-4o',
-            api_key:       nil
+            api_key:       ['env://OPENAI_API_KEY', 'env://CODEX_API_KEY']
           },
           gemini:    {
             enabled:       false,
             default_model: 'gemini-2.0-flash',
-            api_key:       nil
+            api_key:       'env://GEMINI_API_KEY'
           },
           ollama:    {
             enabled:       false,

data/lib/legion/llm/version.rb

@@ -2,6 +2,6 @@
 
 module Legion
   module LLM
-    VERSION = '0.3.3'
+    VERSION = '0.3.5'
   end
 end

data/lib/legion/llm.rb

@@ -9,6 +9,12 @@ require 'legion/llm/compressor'
 require 'legion/llm/quality_checker'
 require 'legion/llm/escalation_history'
 
+begin
+  require 'legion/extensions/llm/gateway'
+rescue LoadError
+  nil
+end
+
 module Legion
   module LLM
     class EscalationExhausted < StandardError; end
@@ -19,6 +25,9 @@ module Legion
       def start
         Legion::Logging.debug 'Legion::LLM is running start'
 
+        require 'legion/llm/claude_config_loader'
+        ClaudeConfigLoader.load
+
         configure_providers
         run_discovery
         set_defaults
@@ -47,20 +56,24 @@ module Legion
         end
       end
 
-      # Create a new chat session
-      # @param model [String] model ID (e.g., "us.anthropic.claude-sonnet-4-6-v1")
-      # @param provider [Symbol] provider slug (e.g., :bedrock, :anthropic)
-      # @param intent [Hash, nil] routing intent (capability, privacy, etc.)
-      # @param tier [Symbol, nil] explicit tier override — skips rule matching
-      # @param escalate [Boolean, nil] enable escalation retry loop (nil = auto from settings)
-      # @param max_escalations [Integer, nil] max escalation attempts override
-      # @param quality_check [Proc, nil] custom quality check callable
-      # @param message [String, nil] message to send (required for escalation)
-      # @param kwargs [Hash] additional options passed to RubyLLM.chat
-      # @return [RubyLLM::Chat]
-      # TODO: fleet tier dispatch via Transport (Phase 3)
+      # Create a new chat session — delegates to lex-llm-gateway when available
+      # for automatic metering and fleet dispatch
       def chat(model: nil, provider: nil, intent: nil, tier: nil, escalate: nil,
                max_escalations: nil, quality_check: nil, message: nil, **)
+        if gateway_loaded? && message
+          return gateway_chat(model: model, provider: provider, intent: intent,
+                              tier: tier, message: message, escalate: escalate,
+                              max_escalations: max_escalations, quality_check: quality_check, **)
+        end
+
+        chat_direct(model: model, provider: provider, intent: intent, tier: tier,
+                    escalate: escalate, max_escalations: max_escalations,
+                    quality_check: quality_check, message: message, **)
+      end
+
+      # Direct chat bypassing gateway — used by gateway runners to avoid recursion
+      def chat_direct(model: nil, provider: nil, intent: nil, tier: nil, escalate: nil,
+                      max_escalations: nil, quality_check: nil, message: nil, **)
         escalate = escalation_enabled? if escalate.nil?
 
         if escalate && message
@@ -74,11 +87,15 @@ module Legion
         end
       end
 
-      # Generate embeddings via Embeddings module
-      # @param text [String, Array<String>] text to embed
-      # @param model [String] embedding model ID
-      # @return [Hash] { vector:, model:, dimensions:, tokens: }
+      # Generate embeddings — delegates to gateway when available
       def embed(text, **)
+        return Legion::Extensions::LLM::Gateway::Runners::Inference.embed(text: text, **) if gateway_loaded?
+
+        embed_direct(text, **)
+      end
+
+      # Direct embed bypassing gateway
+      def embed_direct(text, **)
         require 'legion/llm/embeddings'
         Embeddings.generate(text: text, **)
       end
@@ -91,11 +108,19 @@ module Legion
         Embeddings.generate_batch(texts: texts, **)
       end
 
-      # Generate structured JSON output from LLM
-      # @param messages [Array<Hash>] conversation messages
-      # @param schema [Hash] JSON schema to enforce
-      # @return [Hash] { data:, raw:, model:, valid: }
+      # Generate structured JSON output — delegates to gateway when available
       def structured(messages:, schema:, **)
+        if gateway_loaded?
+          return Legion::Extensions::LLM::Gateway::Runners::Inference.structured(
+            messages: messages, schema: schema, **
+          )
+        end
+
+        structured_direct(messages: messages, schema: schema, **)
+      end
+
+      # Direct structured bypassing gateway
+      def structured_direct(messages:, schema:, **)
         require 'legion/llm/structured_output'
         StructuredOutput.generate(messages: messages, schema: schema, **)
       end
@@ -110,6 +135,14 @@ module Legion
 
       private
 
+      def gateway_loaded?
+        defined?(Legion::Extensions::LLM::Gateway::Runners::Inference)
+      end
+
+      def gateway_chat(**)
+        Legion::Extensions::LLM::Gateway::Runners::Inference.chat(**)
+      end
+
       def chat_single(model:, provider:, intent:, tier:, **kwargs)
         if (intent || tier) && Router.routing_enabled?
           resolution = Router.resolve(intent: intent, tier: tier, model: model, provider: provider)

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: legion-llm
 version: !ruby/object:Gem::Version
-  version: 0.3.3
+  version: 0.3.5
 platform: ruby
 authors:
 - Esity
@@ -89,6 +89,7 @@ files:
 - legion-llm.gemspec
 - lib/legion/llm.rb
 - lib/legion/llm/bedrock_bearer_auth.rb
+- lib/legion/llm/claude_config_loader.rb
 - lib/legion/llm/compressor.rb
 - lib/legion/llm/discovery/ollama.rb
 - lib/legion/llm/discovery/system.rb