legion-llm 0.8.27 → 0.8.28

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2cccf9351fd9f4db59b1548197bf7b78c5947e85183535e86ede5c3359d71b89
-  data.tar.gz: 14e3a1b5b6648bea618941f84473e63aeccee7edc9520366d09dde8d27b00a7b
+  metadata.gz: 523afac32d76644a92db4f6af5228c9ff9856521ccffb7cff0a0e8194570a432
+  data.tar.gz: b58073ec104d42eb18436fd708a33bea881931386dfab1e578f0570d891a6f55
 SHA512:
-  metadata.gz: 31ec279fcb498e5cc3308bcefcb6adc94915c36867967fd08aa3f0422d4c583f83bc0ace1db9a47ecd45371a0ce0c82542fea7015c1474f5b7386409d789e5e0
-  data.tar.gz: '083bd8e581399a574b313a31784eacca4424fadb131f82cd17a5a7e840420da14ec3e5f589efa88b7be6be8a76916439c88bbf936a6d13c9a9435bd8fd04245c'
+  metadata.gz: 205d3a1ef6f1c9e8712bc61e2d88382b88a91560343ab6be7e5c863f2b839ea3d384f5a5642240f7a62fed73ed96aff7b653855c15c1cb5517a43d342306a54b
+  data.tar.gz: 8847c3be8580a5c1c62bd61b83c72ef53db974b19a8567fd102827b748e9113bfc3dc0af7aa14e23d25b524453b2465dd8053e55f64ee2798f9ae9b387cac264

data/CHANGELOG.md

@@ -1,5 +1,14 @@
 # Legion LLM Changelog
 
+## [0.8.28] - 2026-04-24
+
+### Fixed
+- Model/provider mismatch when clients send a model name (e.g., `qwen3.5:latest`) without an explicit provider. The fallback paths blindly paired it with `default_provider` (typically `bedrock`), causing `RubyLLM::ModelNotFoundError`. Now infers the correct provider from model naming patterns before falling back to the global default.
+- `arbitrage_fallback` hardcoded `:cloud` tier and `:bedrock` provider when inference failed. Now uses `PROVIDER_TIER` to resolve the correct tier for the inferred provider.
+
+### Added
+- `Router.infer_provider_for_model(model)` — public method that maps model naming patterns to providers. Recognizes Ollama-style models (`:` or `/` in name), Bedrock (`us.*`), OpenAI (`gpt-*`, `o1-*`/`o3-*`/`o4-*`), Anthropic (`claude-*`), and Gemini (`gemini-*`).
+
 ## [0.8.27] - 2026-04-24
 
 ### Fixed

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

@@ -328,7 +328,9 @@ module Legion
             end
           end
 
-          @resolved_provider = provider || Legion::LLM.settings[:default_provider]
+          @resolved_provider = provider ||
+                               (model && Router.infer_provider_for_model(model)) ||
+                               Legion::LLM.settings[:default_provider]
           @resolved_model = model || Legion::LLM.settings[:default_model]
 
           log.info "[llm][inference] resolved provider=#{@resolved_provider} model=#{@resolved_model}"
@@ -846,6 +848,8 @@ module Legion
           duration_ms = started_at ? ((finished_at - started_at) * 1000).round : nil
 
           result_str = (raw.is_a?(String) ? raw : raw.to_s)
+          result_str = result_str.encode('UTF-8', invalid: :replace, undef: :replace, replace: '�') unless result_str.valid_encoding?
+          result_str = result_str.delete("\x00")
           is_error = raw.is_a?(Hash) && (raw[:error] || raw['error']) ? true : false
 
           @pending_tool_history_mutex.synchronize do

data/lib/legion/llm/inference.rb

@@ -496,7 +496,8 @@ module Legion
         end
 
         model ||= Legion::LLM.settings[:default_model]
-        provider ||= Legion::LLM.settings[:default_provider]
+        provider ||= (model && Router.infer_provider_for_model(model)) ||
+                     Legion::LLM.settings[:default_provider]
 
         opts = {}
         opts[:model] = model if model

data/lib/legion/llm/router.rb

@@ -18,7 +18,22 @@ module Legion
                         gemini: :cloud, azure: :cloud, ollama: :local, vllm: :local }.freeze
       PROVIDER_ORDER = %i[ollama vllm bedrock azure gemini anthropic openai].freeze
 
+      OLLAMA_MODEL_PATTERN = %r{[:/]}
+
       class << self
+        def infer_provider_for_model(model)
+          return nil if model.nil? || model.to_s.empty?
+
+          model_s = model.to_s
+          return :bedrock if model_s.start_with?('us.')
+          return :openai if model_s.match?(/\Agpt-|\Ao[134]-/)
+          return :anthropic if model_s.start_with?('claude-')
+          return :gemini if model_s.start_with?('gemini-')
+          return :ollama if model_s.match?(OLLAMA_MODEL_PATTERN)
+
+          nil
+        end
+
         # Resolve an LLM routing intent to a tier/provider/model decision.
         #
         # @param intent   [Hash, nil] routing intent (capability, privacy, etc.)
@@ -95,18 +110,12 @@ module Legion
           model = Arbitrage.cheapest_for(capability: capability)
           return nil unless model
 
-          provider = Arbitrage.cost_table[model] ? infer_provider(model) : nil
-          log.debug("Router: arbitrage fallback selected model=#{model}")
-          Resolution.new(tier: :cloud, provider: provider || :bedrock, model: model, rule: 'arbitrage_fallback')
-        end
-
-        def infer_provider(model)
-          return :ollama if model.include?('llama')
-          return :bedrock if model.start_with?('us.')
-          return :openai if model.start_with?('gpt')
-          return :google if model.start_with?('gemini')
+          provider = infer_provider_for_model(model)
+          return nil unless provider
 
-          :anthropic if model.start_with?('claude')
+          tier = PROVIDER_TIER.fetch(provider, :cloud)
+          log.debug("Router: arbitrage fallback selected model=#{model} provider=#{provider} tier=#{tier}")
+          Resolution.new(tier: tier, provider: provider, model: model, rule: 'arbitrage_fallback')
         end
 
         def explicit_resolution(tier, provider, model)

data/lib/legion/llm/version.rb

@@ -2,6 +2,6 @@
 
 module Legion
   module LLM
-    VERSION = '0.8.27'
+    VERSION = '0.8.28'
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: legion-llm
 version: !ruby/object:Gem::Version
-  version: 0.8.27
+  version: 0.8.28
 platform: ruby
 authors:
 - Esity