docs(rfd): Add D66, provider-native token counting - #1147
Open
JeanMertz wants to merge 1 commit into
Open
Conversation
Drafts a design for `count_input_tokens` on the `Provider` trait, backed by each provider's own counting endpoint, and uses it to tell a user which turn range would fit when `jp conversation compact --summary` is rejected for size. The motivating session took eight commands over seven minutes to find a range that fit, each one uploading a multi-megabyte request to learn a single number. The draft records the measurements behind the design: a 4,220,150-byte Anthropic body counts in 0.63 seconds, the count endpoint rejects any top-level field outside its accepted set, and `create_request` emits eight such fields of which five depend on user config. Signed-off-by: Jean Mertz <git@jeanmertz.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Drafts a design for
count_input_tokenson theProvidertrait, backed by each provider's own counting endpoint, and uses it to tell a user which turn range would fit whenjp conversation compact --summaryis rejected for size.The motivating session took eight commands over seven minutes to find a range that fit, each one uploading a multi-megabyte request to learn a single number. The draft records the measurements behind the design: a 4,220,150-byte Anthropic body counts in 0.63 seconds, the count endpoint rejects any top-level field outside its accepted set, and
create_requestemits eight such fields of which five depend on user config.