Add finish reason to Generation for usage downstream (#526)

Add `finish_reason` to `Generation` as well as extend
`BaseOpenAI._generate` to include it in the output. This can be useful
for usage in downstream tasks when we need to filter for only
generations that finished because of `"stop"` for example. Maybe we
should add this to `LLMChain` as well?

For more details, see
https://beta.openai.com/docs/guides/completion/best-practices

Signed-off-by: Diwank Singh Tomer <diwank.singh@gmail.com>

langchain/llms/openai.py

@@ -164,7 +164,16 @@ class BaseOpenAI(BaseLLM, BaseModel):
         for i, prompt in enumerate(prompts):
             sub_choices = choices[i * self.n : (i + 1) * self.n]
             generations.append(
-                [Generation(text=choice["text"]) for choice in sub_choices]
+                [
+                    Generation(
+                        text=choice["text"],
+                        generation_info=dict(
+                            finish_reason=choice["finish_reason"],
+                            logprobs=choice["logprobs"],
+                        ),
+                    )
+                    for choice in sub_choices
+                ]
             )
         return LLMResult(
             generations=generations, llm_output={"token_usage": token_usage}

langchain/schema.py

@@ -1,6 +1,6 @@
 """Common schema objects."""
 
-from typing import List, NamedTuple, Optional
+from typing import Any, Dict, List, NamedTuple, Optional
 
 
 class AgentAction(NamedTuple):
@@ -23,6 +23,10 @@ class Generation(NamedTuple):
 
     text: str
     """Generated text output."""
+
+    generation_info: Optional[Dict[str, Any]] = None
+    """Raw generation info response from the provider"""
+    """May include things like reason for finishing (e.g. in OpenAI)"""
     # TODO: add log probs