Skip to content

refactor(config,docs): inline max_tokens default and note task-budgets beta#32

Merged
hakula139 merged 2 commits intomainfrom
refactor/config-max-tokens-default
Apr 24, 2026
Merged

refactor(config,docs): inline max_tokens default and note task-budgets beta#32
hakula139 merged 2 commits intomainfrom
refactor/config-max-tokens-default

Conversation

@hakula139
Copy link
Copy Markdown
Owner

@hakula139 hakula139 commented Apr 24, 2026

Summary

Two small cleanups to the config + research-doc surface.

  • DEFAULT_MAX_TOKENS inlined as 16_000. The constant was used in exactly one place (the low / medium / no-effort arm of default_max_tokens) and obscured the tier table that already has 64_000 and 32_000 inline; rebased to 16_000 for round-number consistency with the upper tiers. Trimmed narrative references from the default_max_tokens doc and the affected test comment.
  • Docs swept for the value change. docs/guide/configuration.md and docs/research/anthropic-api.md now state the 16 000 floor and no longer pin the scaling to an external reference.
  • Added task-budgets-2026-03-13 to the anthropic-api research doc. New Opus 4.7-exclusive beta that ships alongside xhigh / adaptive thinking. oxide-code doesn't implement it today, but the descriptor table now notes its existence so the research notes stay comprehensive for readers tracking the 4.7 surface.

Test plan

  • cargo fmt --all --check
  • cargo build
  • cargo clippy --all-targets -- -D warnings
  • cargo test — 926 passed
  • pnpm lint — 0 errors

@hakula139 hakula139 added the enhancement New feature or request label Apr 24, 2026
@hakula139 hakula139 self-assigned this Apr 24, 2026
@hakula139 hakula139 added the enhancement New feature or request label Apr 24, 2026
@codecov
Copy link
Copy Markdown

codecov Bot commented Apr 24, 2026

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@hakula139 hakula139 force-pushed the refactor/config-max-tokens-default branch from a1bc64e to f3ffbca Compare April 24, 2026 09:45
Used in exactly one place; the named constant obscures the tier
table that already lives inline with 64_000 and 32_000. Drop the
constant, rebase the default to 16_000 for consistency with the
other tiers, and trim narrative references from the `default_max_tokens`
doc and a test comment.
New Opus 4.7-exclusive beta that ships alongside `xhigh` / adaptive
thinking. oxide-code doesn't implement it today — the beta is
opt-in and paired with an `output_config.task_budget` body field.
Document its existence in the additional-betas descriptor table so
the research notes stay comprehensive for readers tracking the 4.7
surface.
@hakula139 hakula139 force-pushed the refactor/config-max-tokens-default branch from f3ffbca to b1ca79a Compare April 24, 2026 09:53
@hakula139 hakula139 changed the title refactor(config): inline DEFAULT_MAX_TOKENS as 16_000 refactor(config,docs): inline max_tokens default and note task-budgets beta Apr 24, 2026
@hakula139 hakula139 merged commit 3f3da00 into main Apr 24, 2026
3 checks passed
@hakula139 hakula139 deleted the refactor/config-max-tokens-default branch April 24, 2026 09:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant