Fireworks - deepseek r1 example with think token filtration

Author: peterbanda

openai-examples/src/main/scala/io/cequence/openaiscala/examples/nonopenai/FireworksAICreateChatCompletion.scala

@@ -4,7 +4,6 @@ import io.cequence.openaiscala.domain._
 import io.cequence.openaiscala.domain.settings.CreateChatCompletionSettings
 import io.cequence.openaiscala.examples.ExampleBase
 import io.cequence.openaiscala.service.OpenAIChatCompletionService
-import io.cequence.openaiscala.service.adapter.{MessageConversions, OpenAIServiceAdapters}
 
 import scala.concurrent.Future
 
@@ -16,18 +15,7 @@ import scala.concurrent.Future
  */
 object FireworksAICreateChatCompletion extends ExampleBase[OpenAIChatCompletionService] {
 
-  // thinking process ends with </think>
-  private val omitThinkingOutput = true
-
-  override val service: OpenAIChatCompletionService = {
-    val adapters = OpenAIServiceAdapters.forChatCompletionService
-    val vanillaService = ChatCompletionProvider.fireworks
-
-    if (omitThinkingOutput)
-      adapters.chatCompletionOutput(MessageConversions.filterOutToThinkEnd)(vanillaService)
-    else
-      vanillaService
-  }
+  override val service: OpenAIChatCompletionService = ChatCompletionProvider.fireworks
 
   private val fireworksModelPrefix = "accounts/fireworks/models/"
 
@@ -36,7 +24,7 @@ object FireworksAICreateChatCompletion extends ExampleBase[OpenAIChatCompletionS
     UserMessage("What is the weather like in Norway?")
   )
 
-  private val modelId = NonOpenAIModelId.deepseek_r1 // llama_v3p1_405b_instruct
+  private val modelId = NonOpenAIModelId.llama_v3p1_405b_instruct
 
   override protected def run: Future[_] =
     service

openai-examples/src/main/scala/io/cequence/openaiscala/examples/nonopenai/FireworksAICreateChatCompletionStreamed.scala

@@ -4,30 +4,15 @@ import akka.stream.scaladsl.Sink
 import io.cequence.openaiscala.domain._
 import io.cequence.openaiscala.domain.settings.CreateChatCompletionSettings
 import io.cequence.openaiscala.examples.ExampleBase
-import io.cequence.openaiscala.service.OpenAIChatCompletionIOConversionAdapter
 import io.cequence.openaiscala.service.StreamedServiceTypes.OpenAIChatCompletionStreamedService
-import io.cequence.openaiscala.service.adapter.MessageConversions
 
 import scala.concurrent.Future
 
 // requires `openai-scala-client-stream` as a dependency and `FIREWORKS_API_KEY` environment variable to be set
 object FireworksAICreateChatCompletionStreamed
     extends ExampleBase[OpenAIChatCompletionStreamedService] {
 
-  // thinking process ends with </think>
-  private val omitThinkingOutput = true
-
-  override val service: OpenAIChatCompletionStreamedService = {
-    val vanillaService = ChatCompletionProvider.fireworks
-
-    if (omitThinkingOutput)
-      OpenAIChatCompletionIOConversionAdapter(
-        vanillaService,
-        outputChunkMessageConversion = Some(MessageConversions.filterOutToThinkEndFlow)
-      )
-    else
-      vanillaService
-  }
+  override val service: OpenAIChatCompletionStreamedService = ChatCompletionProvider.fireworks
 
   private val fireworksModelPrefix = "accounts/fireworks/models/"
 
@@ -36,7 +21,7 @@ object FireworksAICreateChatCompletionStreamed
     UserMessage("What is the weather like in Norway?")
   )
 
-  private val modelId = NonOpenAIModelId.deepseek_r1 // drbx_instruct
+  private val modelId = NonOpenAIModelId.llama_v3p3_70b_instruct
 
   override protected def run: Future[_] =
     service

openai-examples/src/main/scala/io/cequence/openaiscala/examples/nonopenai/FireworksAICreateChatCompletionStreamedWithDeepseek.scala

@@ -0,0 +1,59 @@
+package io.cequence.openaiscala.examples.nonopenai
+
+import akka.stream.scaladsl.Sink
+import io.cequence.openaiscala.domain._
+import io.cequence.openaiscala.domain.settings.CreateChatCompletionSettings
+import io.cequence.openaiscala.examples.ExampleBase
+import io.cequence.openaiscala.service.OpenAIChatCompletionIOConversionAdapter
+import io.cequence.openaiscala.service.StreamedServiceTypes.OpenAIChatCompletionStreamedService
+import io.cequence.openaiscala.service.adapter.MessageConversions
+
+import scala.concurrent.Future
+
+// requires `openai-scala-client-stream` as a dependency and `FIREWORKS_API_KEY` environment variable to be set
+object FireworksAICreateChatCompletionStreamedWithDeepseek
+    extends ExampleBase[OpenAIChatCompletionStreamedService] {
+
+  // thinking process ends with </think>
+  private val omitThinkingOutput = true
+
+  override val service: OpenAIChatCompletionStreamedService = {
+    val vanillaService = ChatCompletionProvider.fireworks
+
+    if (omitThinkingOutput)
+      OpenAIChatCompletionIOConversionAdapter(
+        vanillaService,
+        outputChunkMessageConversion = Some(MessageConversions.filterOutToThinkEndFlow)
+      )
+    else
+      vanillaService
+  }
+
+  private val fireworksModelPrefix = "accounts/fireworks/models/"
+
+  private val messages = Seq(
+    SystemMessage("You are a helpful assistant. Be short."),
+    UserMessage("What is the weather like in Norway?")
+  )
+
+  private val modelId = NonOpenAIModelId.deepseek_r1
+
+  override protected def run: Future[_] =
+    service
+      .createChatCompletionStreamed(
+        messages = messages,
+        settings = CreateChatCompletionSettings(
+          model = fireworksModelPrefix + modelId,
+          temperature = Some(0.01),
+          max_tokens = Some(2048),
+          top_p = Some(0.9),
+          presence_penalty = Some(0)
+        )
+      )
+      .runWith(
+        Sink.foreach { completion =>
+          val content = completion.choices.headOption.flatMap(_.delta.content)
+          print(content.getOrElse(""))
+        }
+      )
+}

openai-examples/src/main/scala/io/cequence/openaiscala/examples/nonopenai/FireworksAICreateChatCompletionWithDeepseek.scala

@@ -0,0 +1,54 @@
+package io.cequence.openaiscala.examples.nonopenai
+
+import io.cequence.openaiscala.domain._
+import io.cequence.openaiscala.domain.settings.CreateChatCompletionSettings
+import io.cequence.openaiscala.examples.ExampleBase
+import io.cequence.openaiscala.service.OpenAIChatCompletionService
+import io.cequence.openaiscala.service.adapter.{MessageConversions, OpenAIServiceAdapters}
+
+import scala.concurrent.Future
+
+/**
+ * Requires `FIREWORKS_API_KEY` environment variable to be set.
+ *
+ * Check out [[ChatCompletionInputAdapterForFireworksAI]] for a more complex example with an
+ * input adapter
+ */
+object FireworksAICreateChatCompletionWithDeepseek extends ExampleBase[OpenAIChatCompletionService] {
+
+  // thinking process ends with </think>
+  private val omitThinkingOutput = true
+
+  override val service: OpenAIChatCompletionService = {
+    val adapters = OpenAIServiceAdapters.forChatCompletionService
+    val vanillaService = ChatCompletionProvider.fireworks
+
+    if (omitThinkingOutput)
+      adapters.chatCompletionOutput(MessageConversions.filterOutToThinkEnd)(vanillaService)
+    else
+      vanillaService
+  }
+
+  private val fireworksModelPrefix = "accounts/fireworks/models/"
+
+  private val messages = Seq(
+    SystemMessage("You are a helpful assistant."),
+    UserMessage("What is the weather like in Norway?")
+  )
+
+  private val modelId = NonOpenAIModelId.deepseek_r1 // llama_v3p1_405b_instruct
+
+  override protected def run: Future[_] =
+    service
+      .createChatCompletion(
+        messages = messages,
+        settings = CreateChatCompletionSettings(
+          model = fireworksModelPrefix + modelId,
+          temperature = Some(0.1),
+          max_tokens = Some(2048),
+          top_p = Some(0.9),
+          presence_penalty = Some(0)
+        )
+      )
+      .map(printMessageContent)
+}