The following objects can be passed to a step

​
ChatGeneration

@dataclass
class ChatGeneration(BaseGeneration):
    messages: List[GenerationMessage] = Field(default_factory=list)
    type = GenerationType.CHAT
​
messages
List[Dict]
List of messages sent to the LLM. Following the openai chat format.
​
provider
str
The provider of the LLM, like openai.
​
model
str
The model used for the generation, like gpt-4.
​
error
str
The error message if the generation failed.
​
settings
Dict
The settings of the LLM, like temperature…
​
variables
Dict
The variables used to format the prompt.
​
tags
List[str]
Optional tags to add to the generation.
​
prompt
str
The prompt used for the generation.
​
tools
Dict
The tools used to generate the completion, following the openai format.
​
message_completion
Dict
The message returned by the LLM, following the openai chat format.
​
token_count
int
The token count of the completion.
​
input_token_count
int
The token count of the input.
​
output_token_count
int
The token count of the output.
​
tt_first_token
float
The time it took to generate the first token. Only available when streaming.
​
token_throughput_in_s
float
The token throughput in tokens per second. Only available when streaming.
​
duration
float
The duration of the generation in ms.

​
CompletionGeneration

​
provider
str
The provider of the LLM, like openai.
​
model
str
The model used for the generation, like gpt-4.
​
error
str
The error message if the generation failed.
​
settings
Dict
The settings of the LLM, like temperature…
​
variables
Dict
The variables used to format the prompt.
​
tags
List[str]
Optional tags to add to the generation.
​
prompt
str
The prompt used for the generation.
​
completion
str
The completion returned by the LLM.
​
token_count
int
The token count of the completion.
​
input_token_count
int
The token count of the input.
​
output_token_count
int
The token count of the output.
​
tt_first_token
float
The time it took to generate the first token. Only available when streaming.
​
token_throughput_in_s
float
The token throughput in tokens per second. Only available when streaming.
​
duration
float
The duration of the generation in ms.