Result Data & Objects

Agently keeps raw responses, parsed structured data, and metadata so you can consume results at different levels.

get_text: final text

get_text() returns the final done text (even when output is JSON).

text = response.get_text()
print(text)

get_data: parsed / original / all

get_data() supports:

• parsed: structured result (default)
• original: raw provider response
• all: full result snapshot

parsed = response.get_data(type="parsed")
original = response.get_data(type="original")
all_data = response.get_data(type="all")

Key fields inside all:

• meta: usage, finish_reason, etc.
• original_delta / original_done: raw stream chunks and final payload
• text_result: final text
• cleaned_result / parsed_result: cleaned JSON and parsed data
• result_object: Pydantic object (if available)
• errors / extra: errors and extra fields

get_data_object: typed output

When you use Agently Output Format (JSON), you can get a typed object:

result_obj = response.get_data_object()
print(result_obj)

ensure_keys: field guarantees

get_data() supports retries with ensure_keys:

data = response.get_data(
  type="parsed",
  ensure_keys=["intro"],
  key_style="dot",
  max_retries=3,
  raise_ensure_failure=True,
)

key_style supports dot or slash paths.