Skip to content

probbit 0.5.0: rename pbit → probbit everywhere - #8

Merged
BitmapAsset merged 1 commit into
mainfrom
r29-rename-probbit
Oct 2, 2026
Merged

BitmapAsset merged 1 commit into
mainfrom
r29-rename-probbit

Conversation

@BitmapAsset

Copy link
Copy Markdown
Owner

The name is now probbit everywhere: crates, binary, env vars, MCP tools, Python module, npm package, installers, CI, playground, docs. The probabilistic bit is still a p-bit. Version 0.5.0. Nothing is published, tagged or released.

No behaviour changes. I built 0.4.0 (ee64b65) and this branch side by side and ran both on the 7 golden inputs, every file under examples/ (exact, sampled and --summary --pretty, fixed seeds), examples/node/decide.mjs, piped --help, a schema error, stats, and an MCP session (initialize, tools/list, demo, decide, evaluate, run). Method: apply sed -E 's/pbit/probbit/g; s/0\.4\.0/0.5.0/g' to the 0.4.0 stdout, normalise timings the way the golden test does, then compare with the 0.5.0 stdout byte for byte. Result: 30/30 identical, with the same exit codes. Two default-budget denoise runs were not counted because 0.4.0 doesn't reproduce its own output on them (wall-clock budget). Their fixed-work versions are counted and pass. After the proof, I regenerated the golden digests from 0.5.0 and renamed the test to stdout_matches_the_0_5_0_goldens.

Breaking for callers (0 users, 0 releases):

  • The binary is now probbit.
  • Env vars are now PROBBIT_*.
  • MCP tools are now probbit_*.
  • IR programs use "probbit_ir": 1.
  • The evaluate request's pbit block is now probbit.
  • engine reads "probbit 0.5.0".

Migration note in the CHANGELOG.

Two changes outside a plain rename:

  • Hero screen column. The hero screen padded its command column with 45 - cmd.len(). With the longer name that subtraction underflowed and the screen panicked. The column is now 6 wider (the rename's length), so the layout matches 0.4.0.
  • Hero mark. The block-letter mark now spells probbit.

Also in this PR:

  • The IR schema $id now names BitmapAsset/probbit (it named a wrong owner).
  • The hero and social preview use the v2 rabbit + probbit lockup.
  • New docs/brand/ with logo.svg, lockup-probbit.png and a short README.

Checks run locally (Apple M4):

  • cargo build --release --workspace: 0 warnings.
  • cargo test --release --workspace: 159 passed, 1 ignored (same as 0.4.0).
  • Clippy: same warning set as 0.4.0, nothing new.
  • Python 3.9.6: 16 + 7 tests pass.
  • node playground/check.mjs: passes.
  • npm pack --dry-run: probbit-0.5.0.tgz.
  • bash -n install.sh: OK.
  • cargo publish --dry-run: probbit-core OK. The other three crates fail only because probbit-core is not on crates.io yet.

The product, crates, binary, module and tool name is probbit; the
probabilistic bit stays a p-bit. No behaviour changes.

- crates probbit-core, probbit-ir, probbit-decide, probbit-cli (binary
  probbit), probbit-wasm; directories and files moved with git mv
- env vars PROBBIT_*, config file probbit.json, cache dir probbit,
  MCP server probbit and tools probbit_*, the IR key probbit_ir and the
  evaluate request's probbit block, the engine field "probbit 0.5.0"
- Python probbit.py, npm package and bin probbit, installers, release
  assets probbit-<tag>-<target>, CI, the playground (probbit.wasm, C ABI
  probbit_*)
- the terminal hero screen spells probbit; its command column is 6
  characters wider so the longer name fits (it underflowed otherwise)
- golden test renamed stdout_matches_the_0_5_0_goldens, digests from
  0.5.0 after a byte-for-byte equivalence proof against 0.4.0 on the
  7 golden inputs and every example
- README H1 probbit with the p-bit subheading, docs/probbit-ir-json.md
  (probbit IR JSON), the schema $id names BitmapAsset/probbit
- brand: v2 lockup on the hero and social preview, docs/brand/
- version 0.5.0; CHANGELOG entry with a migration note
@BitmapAsset
BitmapAsset merged commit ad94a9d into main Oct 2, 2026
8 checks passed
@BitmapAsset
BitmapAsset deleted the r29-rename-probbit branch October 2, 2026 23:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant