diff --git a/.github/workflows/integration-eval.yml b/.github/workflows/integration-eval.yml index 518aafdf..c1d5fda2 100644 --- a/.github/workflows/integration-eval.yml +++ b/.github/workflows/integration-eval.yml @@ -41,9 +41,10 @@ concurrency: jobs: eval-and-judge: runs-on: ubuntu-latest - # Provider credentials are stored as environment secrets so they can be - # rotated without requiring repo-admin access to Actions secrets. - environment: pypi-internal-preview + # Provider credentials are stored in a dedicated CI environment so they can + # be rotated without repo-admin access and without polluting PyPI preview + # deployment status in the GitHub Deployments sidebar. + environment: integration-eval-secrets # Only run where integration secrets are available. Fork PRs lack secrets, # while internal PRs should fail visibly when the real eval path breaks. if: >- diff --git a/docs/release.md b/docs/release.md index a130059e..829ea8e8 100644 --- a/docs/release.md +++ b/docs/release.md @@ -12,9 +12,9 @@ Current release state: - `0.6.2` is the latest stable release **published on PyPI** (tag `v0.6.2`). Use the public install commands below. -- `main` currently carries `0.6.2`. To resume internal-preview builds for the - next line, bump `main` to `0.6.3.dev0` — internal previews then publish as - `0.6.3.dev` automatically after the `test` workflow passes on `main`. +- `main` tracks the next preview line, currently `0.6.3.dev0`. Internal + previews publish as `0.6.3.dev` automatically after the `test` workflow + and integration gate pass on `main`. ## Install and Upgrade Commands @@ -143,6 +143,10 @@ Create matching GitHub environments: - `pypi-internal-preview`: used for automatic preview publishing from `main`. - `pypi-public`: used for tag-driven public releases. +- `integration-eval-secrets`: used by the integration-eval gate for provider + credentials. Do not reuse `pypi-internal-preview` here, because every Actions + job environment creates a GitHub deployment record and can make the PyPI + preview environment appear stale or inactive. The workflows build with `uv build --no-sources`, check distributions with `twine check`, and publish with `uv publish`. diff --git a/pyproject.toml b/pyproject.toml index a6e485c0..d84a4fa4 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [project] name = "benchflow" -version = "0.6.2" +version = "0.6.3.dev0" description = "Multi-turn agent benchmarking with ACP — run any agent, any model, any provider." readme = "README.md" requires-python = ">=3.12" diff --git a/uv.lock b/uv.lock index 7493a47a..97a5e6d8 100644 --- a/uv.lock +++ b/uv.lock @@ -288,7 +288,7 @@ wheels = [ [[package]] name = "benchflow" -version = "0.6.2" +version = "0.6.3.dev0" source = { editable = "." } dependencies = [ { name = "agent-client-protocol" },