Track usage on OpenAI stream requests, closes #591

llm/default_plugins/openai_models.py

@@ -346,7 +346,7 @@ def execute(self, prompt, stream, response, conversation=None):
                 )
             messages.append({"role": "user", "content": attachment_message})
 
-        kwargs = self.build_kwargs(prompt)
+        kwargs = self.build_kwargs(prompt, stream)
         client = self.get_client()
         if stream:
             completion = client.chat.completions.create(
@@ -358,7 +358,10 @@ def execute(self, prompt, stream, response, conversation=None):
             chunks = []
             for chunk in completion:
                 chunks.append(chunk)
-                content = chunk.choices[0].delta.content
+                try:
+                    content = chunk.choices[0].delta.content
+                except IndexError:
+                    content = None
                 if content is not None:
                     yield content
             response.response_json = remove_dict_none_values(combine_chunks(chunks))
@@ -395,13 +398,15 @@ def get_client(self):
             kwargs["http_client"] = logging_client()
         return openai.OpenAI(**kwargs)
 
-    def build_kwargs(self, prompt):
+    def build_kwargs(self, prompt, stream):
         kwargs = dict(not_nulls(prompt.options))
         json_object = kwargs.pop("json_object", None)
         if "max_tokens" not in kwargs and self.default_max_tokens is not None:
             kwargs["max_tokens"] = self.default_max_tokens
         if json_object:
             kwargs["response_format"] = {"type": "json_object"}
+        if stream:
+            kwargs["stream_options"] = {"include_usage": True}
         return kwargs
 
 
@@ -431,7 +436,7 @@ def execute(self, prompt, stream, response, conversation=None):
                 messages.append(prev_response.prompt.prompt)
                 messages.append(prev_response.text())
         messages.append(prompt.prompt)
-        kwargs = self.build_kwargs(prompt)
+        kwargs = self.build_kwargs(prompt, stream)
         client = self.get_client()
         if stream:
             completion = client.completions.create(
@@ -443,7 +448,10 @@ def execute(self, prompt, stream, response, conversation=None):
             chunks = []
             for chunk in completion:
                 chunks.append(chunk)
-                content = chunk.choices[0].text
+                try:
+                    content = chunk.choices[0].text
+                except IndexError:
+                    content = None
                 if content is not None:
                     yield content
             combined = combine_chunks(chunks)
@@ -472,8 +480,11 @@ def combine_chunks(chunks: List) -> dict:
     # If any of them have log probability, we're going to persist
     # those later on
     logprobs = []
+    usage = {}
 
     for item in chunks:
+        if item.usage:
+            usage = dict(item.usage)
         for choice in item.choices:
             if choice.logprobs and hasattr(choice.logprobs, "top_logprobs"):
                 logprobs.append(
@@ -497,6 +508,7 @@ def combine_chunks(chunks: List) -> dict:
         "content": content,
         "role": role,
         "finish_reason": finish_reason,
+        "usage": usage,
     }
     if logprobs:
         combined["logprobs"] = logprobs