name: "VisionAtomicFlow"
description: "A flow that, given a textual input, and a set of images and/or videos, generates a textual output."

enable_cache: True

n_api_retries: 6
wait_time_between_retries: 20

system_name: system
user_name: user
assistant_name: assistant

backend:
  _target_: aiflows.backends.llm_lite.LiteLLMBackend
  api_infos: ???
  model_name: "gpt-4-vision-preview"
  n: 1
  max_tokens: 2000
  temperature: 0.3
  top_p: 0.2
  frequency_penalty: 0
  presence_penalty: 0

system_message_prompt_template:
  _target_: aiflows.prompt_template.JinjaPrompt

init_human_message_prompt_template:
  _target_: aiflows.prompt_template.JinjaPrompt

human_message_prompt_template:
  _target_: aiflows.prompt_template.JinjaPrompt
  template: "{{query}}"
  input_variables:
    - "query"
    
input_interface_initialized:
  - "query"
  - "data"

previous_messages:
  first_k: null  # Note that the first message is the system prompt
  last_k: null

input_interface:
  - "query"
  - "data"

input_interface_non_initialized:
  - "question"
  - "data"

output_interface:
  - "api_output"