legion-llm 0.5.13 → 0.5.14

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: fa7ba49d17c9d418f172c576bc3d1b2d5174757c489928cecf52a403c43c4544
-  data.tar.gz: a1339abda9a424179721b7b7af22e5ec972c5847976876a967547b5080f48c29
+  metadata.gz: ceb7fa0519b3985439d577579889de688b42a9e8d3c5fdbc6a1be5b22c7fb2ba
+  data.tar.gz: 80fd2cc36a19cf49783f9429e4aca545e26c3b8ca130fa649800955ce52001d2
 SHA512:
-  metadata.gz: 3283898e616549b10b78d722210c538e9f30135a5b20eed49fea1f326c300ca9be4cfbb269ab698080eaee9f8f3d0ccee2bf7350474b195b59130cc55ec5cbe6
-  data.tar.gz: 76c4137add776f2c40634daf05baaf459d99a865147b6021e24f22c442564cca569ac6ff63d6bc3b7c7bb43aafa01690f02a2ec948e8bbf840ca7a43bc89c34f
+  metadata.gz: 82f00569a04406cc64983f447e3fbd5fbb4c9765f9caa59543a7db0dc612ca659368dae3441f4fa77ae27aaf74347dac997cccb842dc3a2e3e99184ac52f591e
+  data.tar.gz: 40d6ec150d5832f0a4d0ac8f9f9e25754c58e6d4d5a528eb13c2c1420a62a8e6e579c3dda417fdd4dcff8f9d6ba888d7613baba266596c442e787f041a1a50a0

data/.rubocop.yml

@@ -54,3 +54,7 @@ Naming/PredicateMethod:
 
 Metrics/ParameterLists:
   Max: 9
+
+Style/RedundantConstantBase:
+  Exclude:
+    - 'spec/**/*'

data/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Legion LLM Changelog
 
+## [0.5.14] - 2026-03-27
+
+### Added
+- `DaemonClient.inference` method for conversation-level routing — accepts a full `messages:` array and optional `tools:`, `model:`, `provider:`, and `timeout:` keyword args, posts to `POST /api/llm/inference`, and returns a structured `{ status: :ok, data: { content:, tool_calls:, stop_reason:, model:, input_tokens:, output_tokens: } }` hash on success
+- `http_post` now accepts an optional `timeout:` keyword argument (default `DEFAULT_TIMEOUT = 60`) so callers like `inference` can pass a longer timeout (120s) without affecting existing `chat` calls
+- `interpret_inference_response` private helper that maps the `/api/llm/inference` HTTP response — 200 returns `:ok` with structured fields, 4xx/5xx follow the same error handling as `interpret_response`
+
 ## [0.5.13] - 2026-03-27
 
 ### Changed

data/lib/legion/llm/daemon_client.rb

@@ -102,13 +102,29 @@ module Legion
         http.request(request)
       end
 
+      # POSTs a conversation-level inference request to the daemon REST API.
+      # Accepts a full messages array and optional tool schemas.
+      # Returns a status hash with structured inference fields on success.
+      def inference(messages:, tools: [], model: nil, provider: nil, timeout: 120)
+        body = { messages: messages, tools: tools }
+        body[:model]    = model    if model
+        body[:provider] = provider if provider
+
+        response = http_post('/api/llm/inference', body, timeout: timeout)
+        interpret_inference_response(response)
+      rescue StandardError => e
+        mark_unhealthy
+        { status: :unavailable, error: e.message }
+      end
+
       # Builds and sends a POST request with a JSON body.
       # Returns Net::HTTPResponse.
-      def http_post(path, body)
+      # The optional timeout: keyword overrides the default read timeout.
+      def http_post(path, body, timeout: DEFAULT_TIMEOUT)
         uri     = URI.parse("#{daemon_url}#{path}")
         http    = Net::HTTP.new(uri.host, uri.port)
         http.open_timeout = 5
-        http.read_timeout = DEFAULT_TIMEOUT
+        http.read_timeout = timeout
         request = Net::HTTP::Post.new(uri.request_uri)
         request['Content-Type'] = 'application/json'
         request.body = ::JSON.dump(body)
@@ -180,7 +196,44 @@ module Legion
         0
       end
 
-      private_class_method :fetch_daemon_url, :safe_parse, :extract_retry_after
+      # Maps an HTTP response from /api/llm/inference to a structured status hash.
+      # On 200 returns :ok with structured inference fields extracted from the body.
+      # All other codes follow the same error handling as interpret_response.
+      def interpret_inference_response(response)
+        code   = response.code.to_i
+        parsed = safe_parse(response.body)
+
+        if code == 200
+          data = parsed.fetch(:data, parsed)
+          return {
+            status: :ok,
+            data:   {
+              content:       data[:content],
+              tool_calls:    data[:tool_calls] || [],
+              stop_reason:   data[:stop_reason],
+              model:         data[:model],
+              input_tokens:  data[:input_tokens],
+              output_tokens: data[:output_tokens]
+            }
+          }
+        end
+
+        case code
+        when 403
+          Legion::Logging.warn("Daemon returned 403 Denied url=#{daemon_url}") if defined?(Legion::Logging)
+          { status: :denied, error: parsed.fetch(:error, parsed) }
+        when 429
+          retry_after = extract_retry_after(response, parsed)
+          Legion::Logging.warn("Daemon returned 429 RateLimited url=#{daemon_url} retry_after=#{retry_after}") if defined?(Legion::Logging)
+          { status: :rate_limited, retry_after: retry_after }
+        when 503
+          { status: :unavailable }
+        else
+          { status: :error, code: code, body: parsed }
+        end
+      end
+
+      private_class_method :fetch_daemon_url, :safe_parse, :extract_retry_after, :interpret_inference_response
     end
   end
 end

data/lib/legion/llm/version.rb

@@ -2,6 +2,6 @@
 
 module Legion
   module LLM
-    VERSION = '0.5.13'
+    VERSION = '0.5.14'
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: legion-llm
 version: !ruby/object:Gem::Version
-  version: 0.5.13
+  version: 0.5.14
 platform: ruby
 authors:
 - Esity