ask-llm-providers 0.1.15 → 0.1.16

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d05af6c6de7917048a5b86d1fea1c0acfd3ef820c71c7672f19375a5024fe138
-  data.tar.gz: 3cf21a18360b52e5fd44d01359c0312e05163a234407f07b515f325abf9ebee7
+  metadata.gz: 93b16f8655292838315d314b3a57dd94368182372a421f88b3911d6a9b171488
+  data.tar.gz: 59fe887341f72aa42c765da7bda10e93454aa8b8cc25bfd1ebb85e316d401a17
 SHA512:
-  metadata.gz: 7e91f0e672b159d2d8c7af3de889d0155ede9fc60191b38a78c4f34e675892f0c867ef30dfd6d7feac055d04ef0be347f06fa9888b8d3282b48d34d26dcd59c5
-  data.tar.gz: 813cc7dc3ca391fca712dae4c59a356019ef363d34a1b2a19376faf1a0b0ac2f2dddb065bd842b64adb7909045936d048bceadc47d2f07de3377115e748f2f08
+  metadata.gz: 8627f783b0e7848373fd9358c13e48db978fe728b043c7c5e9f23fd4425c631203c19eae980d870e1f683cdd16d803199cf7b8cba7b8bb0b5f4bfe7b09ffc1ad
+  data.tar.gz: 13cbee9ed69bca82917cbfd060e6f8f8a5130c717b975b886f3851203dada60da1ac081200c2613efa53b7625e744baedc639e06e2ee60bc990faa91c14bccfa

data/lib/ask/llm/sse_buffer.rb

@@ -0,0 +1,43 @@
+# frozen_string_literal: true
+
+module Ask
+  module LLM
+    module SSEBuffer
+      def init_sse_buffer
+        @_sse_buffer = +""
+      end
+
+      def each_sse_event(raw)
+        @_sse_buffer ||= +""
+        @_sse_buffer << raw
+
+        while (event_end = @_sse_buffer.index("\n\n"))
+          event_data = @_sse_buffer.slice!(0, event_end + 2).strip
+          next if event_data.empty?
+
+          data_content = extract_data(event_data)
+          next if data_content.empty?
+          break if data_content == "[DONE]"
+
+          yield data_content
+        end
+      end
+
+      private
+
+      def extract_data(event_data)
+        content = +""
+        event_data.each_line do |line|
+          line = line.strip
+          next if line.empty? || line.start_with?(":")
+          if line.start_with?("data: ")
+            content << line[6..]
+          elsif line.start_with?("data:")
+            content << line[5..]
+          end
+        end
+        content
+      end
+    end
+  end
+end

data/lib/ask/llm/version.rb

@@ -2,6 +2,6 @@
 
 module Ask
   module LLM
-    VERSION = "0.1.15"
+    VERSION = "0.1.16"
   end
 end

data/lib/ask/provider/anthropic.rb

@@ -4,6 +4,7 @@ module Ask
   module Providers
     # Anthropic Claude API provider.
     class Anthropic < Ask::Provider
+      include Ask::LLM::SSEBuffer
       def initialize(config = {})
         config = normalize_config(config)
         super(config)
@@ -179,6 +180,7 @@ module Ask
 
       def chat_stream(payload, model, &block)
         stream = Ask::Stream.new
+        init_sse_buffer
         response = @http.post("v1/messages") do |req|
           req.body = payload.merge(stream: true)
           req.options.on_data = proc { |data, _bytes, _env| process_anthropic_chunk(data, stream, model, &block) }
@@ -189,39 +191,24 @@ module Ask
       end
 
       def process_anthropic_chunk(raw, stream, model)
-        raw.each_line do |line|
-          line = line.strip
-          next if line.empty? || line.start_with?(":")
-          next unless line.start_with?("event:") || line.start_with?("data:")
-
-          if line.start_with?("data: ")
-            data = line[6..]
-            begin
-              parsed = JSON.parse(data)
-            rescue JSON::ParserError
-              next
-            end
-
-            case parsed["type"]
-            when "content_block_delta"
-              delta = parsed.dig("delta")
-              next unless delta
-              chunk = Ask::Chunk.new(
-                content: delta["text"],
-                finish_reason: delta["type"] == "thinking_delta" ? nil : nil
-              )
+        each_sse_event(raw) do |data|
+          parsed = JSON.parse(data) rescue next
+
+          case parsed["type"]
+          when "content_block_delta"
+            delta = parsed.dig("delta")
+            next unless delta
+            chunk = Ask::Chunk.new(content: delta["text"])
+            stream.add(chunk)
+            yield chunk if block_given?
+          when "message_stop"
+            usage = parsed["usage"] || parsed["message"]&.dig("usage")
+            if usage
+              chunk = Ask::Chunk.new(finish_reason: "stop", usage: usage)
               stream.add(chunk)
               yield chunk if block_given?
-            when "message_stop"
-              usage = parsed["usage"] || parsed["message"]&.dig("usage")
-              if usage
-                chunk = Ask::Chunk.new(finish_reason: "stop", usage: usage)
-                stream.add(chunk)
-                yield chunk if block_given?
-              end
-            when "message_start"
-              # Message started — no content yet
             end
+          when "message_start"
           end
         end
       end

data/lib/ask/provider/cloudflare.rb

@@ -4,6 +4,7 @@ module Ask
   module Providers
     # Cloudflare Workers AI provider. Supports both direct Workers AI and AI Gateway.
     class Cloudflare < Ask::Provider
+      include Ask::LLM::SSEBuffer
       def initialize(config = {})
         config = normalize_config(config)
         super(config)
@@ -96,6 +97,7 @@ module Ask
 
       def chat_stream_gateway(endpoint, payload, model, &block)
         stream = Ask::Stream.new
+        init_sse_buffer
         response = @http.post(endpoint) do |req|
           req.body = payload.merge(stream: true)
           req.options.on_data = proc { |data, _bytes, _env| process_stream_chunk(data, stream, model, &block) }
@@ -106,11 +108,7 @@ module Ask
       end
 
       def process_stream_chunk(raw, stream, model)
-        raw.each_line do |line|
-          line = line.strip
-          next unless line.start_with?("data: ")
-          data = line[6..]
-          next if data == "[DONE]"
+        each_sse_event(raw) do |data|
           parsed = JSON.parse(data) rescue next
           delta = parsed.dig("choices", 0, "delta") || {}
           chunk = Ask::Chunk.new(content: delta["content"])

data/lib/ask/provider/google.rb

@@ -4,6 +4,7 @@ module Ask
   module Providers
     # Google Gemini API provider. Also supports Vertex AI via GCP service account auth.
     class Google < Ask::Provider
+      include Ask::LLM::SSEBuffer
       def initialize(config = {})
         config = normalize_config(config)
         super(config)
@@ -187,6 +188,7 @@ module Ask
 
       def chat_stream(path, payload, model, &block)
         stream = Ask::Stream.new
+        init_sse_buffer
         response = @http.post(path) do |req|
           req.body = payload
           req.params["key"] = @config.api_key if @config.api_key
@@ -198,10 +200,7 @@ module Ask
       end
 
       def process_google_chunk(raw, stream, model)
-        raw.each_line do |line|
-          next unless line.start_with?("data: ")
-          data = line[6..]
-          next if data.strip == "[DONE]"
+        each_sse_event(raw) do |data|
           parsed = JSON.parse(data) rescue next
           candidate = parsed.dig("candidates", 0) or next
           part = candidate.dig("content", "parts", 0)

data/lib/ask/provider/ollama.rb

@@ -79,6 +79,7 @@ module Ask
 
       def chat_stream(payload, model, &block)
         stream = Ask::Stream.new
+        @_sse_buffer = +""
         response = @http.post("api/chat") do |req|
           req.body = payload.merge(stream: true)
           req.options.on_data = proc { |data, _bytes, _env| process_ollama_chunk(data, stream, model, &block) }
@@ -89,7 +90,13 @@ module Ask
       end
 
       def process_ollama_chunk(raw, stream, model)
-        raw.each_line do |line|
+        @_sse_buffer ||= +""
+        @_sse_buffer << raw
+
+        while (line_end = @_sse_buffer.index("\n"))
+          line = @_sse_buffer.slice!(0, line_end + 1).strip
+          next if line.empty?
+
           parsed = JSON.parse(line) rescue next
           msg = parsed["message"] || {}
           chunk = Ask::Chunk.new(content: msg["content"])

data/lib/ask/provider/openai.rb

@@ -6,6 +6,7 @@ module Ask
     # (OpenRouter, DeepSeek, Azure, XAI, Perplexity, GPUStack, etc.) via
     # +base_url+ override.
     class OpenAI < Ask::Provider
+      include Ask::LLM::SSEBuffer
       def initialize(config = {})
         @provider_keys = extract_provider_keys(config)
         config = normalize_config(config)
@@ -148,6 +149,7 @@ module Ask
 
       def chat_stream(payload, model, &block)
         stream = Ask::Stream.new
+        init_sse_buffer
         @http.post("chat/completions") do |req|
           req.body = payload.merge(stream: true)
           req.options.on_data = proc { |data, _bytes, _env| process_chunk(data, stream, model, &block) }
@@ -167,12 +169,8 @@ module Ask
         stream
       end
 
-
       def process_chunk(raw, stream, model)
-        raw.each_line do |line|
-          line = line.strip
-          next if line.empty? || line.start_with?(":") || !line.start_with?("data: ")
-          data = line[6..]; next if data == "[DONE]"
+        each_sse_event(raw) do |data|
           parsed = JSON.parse(data) rescue next
           choice = parsed.dig("choices", 0) or next
           delta = choice["delta"] || {}

data/lib/ask-llm-providers.rb

@@ -10,6 +10,7 @@ require "base64"
 # Common infrastructure
 require_relative "ask/llm/config"
 require_relative "ask/llm/http"
+require_relative "ask/llm/sse_buffer"
 require_relative "ask/llm/models/openai"
 
 # Load providers

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: ask-llm-providers
 version: !ruby/object:Gem::Version
-  version: 0.1.15
+  version: 0.1.16
 platform: ruby
 authors:
 - Kaka Ruto
@@ -178,6 +178,7 @@ files:
 - lib/ask/llm/config.rb
 - lib/ask/llm/http.rb
 - lib/ask/llm/models/openai.rb
+- lib/ask/llm/sse_buffer.rb
 - lib/ask/llm/version.rb
 - lib/ask/provider/anthropic.rb
 - lib/ask/provider/bedrock.rb