Configure context budget behavior
Usage
context_budget(
max_input_tokens = NULL,
offload_tokens = 20000,
summarize_at = 0.85,
keep_recent = 0.1,
token_estimator = NULL
)Arguments
- max_input_tokens
Optional model input budget.
- offload_tokens
Approximate token threshold for offloading values.
- summarize_at
Fraction of budget at which summarization should occur.
- keep_recent
Fraction of turns to keep during summarization.
- token_estimator
Function used to estimate tokens from text.