add streamUsage

libs/langchain-cohere/src/chat_models.ts

@@ -47,6 +47,14 @@ export interface ChatCohereInput extends BaseChatModelParams {
    * @default {false}
    */
   streaming?: boolean;
+  /**
+   * Whether or not to include token usage when streaming.
+   * This will include an extra chunk at the end of the stream
+   * with `eventType: "stream-end"` and the token usage in
+   * `usage_metadata`.
+   * @default {true}
+   */
+  streamUsage?: boolean;
 }
 
 interface TokenUsage {
@@ -58,7 +66,8 @@ interface TokenUsage {
 export interface CohereChatCallOptions
   extends BaseLanguageModelCallOptions,
     Partial<Omit<Cohere.ChatRequest, "message">>,
-    Partial<Omit<Cohere.ChatStreamRequest, "message">> {}
+    Partial<Omit<Cohere.ChatStreamRequest, "message">>,
+    Pick<ChatCohereInput, "streamUsage"> {}
 
 function convertMessagesToCohereMessages(
   messages: Array<BaseMessage>
@@ -130,6 +139,8 @@ export class ChatCohere<
 
   streaming = false;
 
+  streamUsage: boolean = true;
+
   constructor(fields?: ChatCohereInput) {
     super(fields ?? {});
 
@@ -144,6 +155,7 @@ export class ChatCohere<
     this.model = fields?.model ?? this.model;
     this.temperature = fields?.temperature ?? this.temperature;
     this.streaming = fields?.streaming ?? this.streaming;
+    this.streamUsage = fields?.streamUsage ?? this.streamUsage;
   }
 
   getLsParams(options: this["ParsedCallOptions"]): LangSmithParams {
@@ -331,7 +343,9 @@ export class ChatCohere<
       if (chunk.eventType === "text-generation") {
         yield new ChatGenerationChunk({
           text: chunk.text,
-          message: new AIMessageChunk({ content: chunk.text }),
+          message: new AIMessageChunk({
+            content: chunk.text,
+          }),
         });
         await runManager?.handleLLMNewToken(chunk.text);
       } else if (chunk.eventType !== "stream-end") {
@@ -351,13 +365,11 @@ export class ChatCohere<
         });
       } else if (
         chunk.eventType === "stream-end" &&
-        chunk.response.meta?.tokens &&
-        (chunk.response.meta.tokens.inputTokens ||
-          chunk.response.meta.tokens.outputTokens)
+        (this.streamUsage || options.streamUsage)
       ) {
         // stream-end events contain the final token count
-        const input_tokens = chunk.response.meta.tokens.inputTokens ?? 0;
-        const output_tokens = chunk.response.meta.tokens.outputTokens ?? 0;
+        const input_tokens = chunk.response.meta?.tokens?.inputTokens ?? 0;
+        const output_tokens = chunk.response.meta?.tokens?.outputTokens ?? 0;
         yield new ChatGenerationChunk({
           text: "",
           message: new AIMessageChunk({

libs/langchain-cohere/src/tests/chat_models.int.test.ts

@@ -59,12 +59,13 @@ test("should abort the request", async () => {
   }).rejects.toThrow("AbortError");
 });
 
-test("Stream token count usage_metadata", async () => {
+test.only("Stream token count usage_metadata", async () => {
   const model = new ChatCohere({
     model: "command-light",
     temperature: 0,
   });
   let res: AIMessageChunk | null = null;
+  let lastRes: AIMessageChunk | null = null;
   for await (const chunk of await model.stream(
     "Why is the sky blue? Be concise."
   )) {
@@ -73,6 +74,7 @@ test("Stream token count usage_metadata", async () => {
     } else {
       res = res.concat(chunk);
     }
+    lastRes = chunk;
   }
   console.log(res);
   expect(res?.usage_metadata).toBeDefined();
@@ -84,6 +86,41 @@ test("Stream token count usage_metadata", async () => {
   expect(res.usage_metadata.total_tokens).toBe(
     res.usage_metadata.input_tokens + res.usage_metadata.output_tokens
   );
+  expect(lastRes?.additional_kwargs).toBeDefined();
+  if (!lastRes?.additional_kwargs) {
+    return;
+  }
+  expect(lastRes.additional_kwargs.eventType).toBe("stream-end");
+});
+
+test.only("streamUsage excludes token usage", async () => {
+  const model = new ChatCohere({
+    model: "command-light",
+    temperature: 0,
+    streamUsage: false,
+  });
+  let res: AIMessageChunk | null = null;
+  let lastRes: AIMessageChunk | null = null;
+  for await (const chunk of await model.stream(
+    "Why is the sky blue? Be concise."
+  )) {
+    if (!res) {
+      res = chunk;
+    } else {
+      res = res.concat(chunk);
+    }
+    lastRes = chunk;
+  }
+  console.log(res);
+  expect(res?.usage_metadata).not.toBeDefined();
+  if (res?.usage_metadata) {
+    return;
+  }
+  expect(lastRes?.additional_kwargs).toBeDefined();
+  if (!lastRes?.additional_kwargs) {
+    return;
+  }
+  expect(lastRes.additional_kwargs.eventType).not.toBe("stream-end");
 });
 
 test("Invoke token count usage_metadata", async () => {