Release 0.1.6

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code

[Shashankss1205]

Cuts the Unreleased section, which had been accumulating since 0.1.5, and adds the five changes that landed after it.

What ships

• grapharc[mcp] was broken on arrival (grapharc[mcp] is broken on arrival: mcp>=1.2 resolves to 2.0, where mcp.server.fastmcp no longer exists #101) — mcp>=1.2 resolved to 2.0, where mcp.server.fastmcp no longer exists. Pinned <2; grapharc/mcp/__init__.py resolves the SDK-backed names lazily so the subpackage imports without the extra.
• go <dir> silently re-executed an already-executed plan (go <dir> re-executes an already-executed plan silently, and plan.json forgets the earlier runs #100, go <dir> refuses a plan that already ran, and the record stops forgetting #104) — refused with exit 2 unless --again; the record accumulates executed_run_ids. An approval binds to a fingerprint that does not change between runs, so this was one human yes authorising N executions of a mutating plan.
• The deep dive's Verified this pass figures rotted (docs: the README's 'Verified this pass' numbers rot silently — nothing re-derives them #42, The deep dive's verified figures are re-derived by a test, or they are not on the line #105) — now re-derived by tests/test_deep_dive.py, with guards against a figure being deleted or an unowned one being added.
• Five max_seconds timing tests failed only under load (tests: the max_seconds timing tests flake under CPU contention, not on an idle machine #40, Timing bounds sit between the deadline and the body, not close to either #106) — every wall-clock bound now sits between the deadline and the uninterrupted body.
• Novita joins the model gateway (Add Novita as a model gateway backend #102, thanks @jax-novita) — a fifth OpenAI-wire backend behind a novita extra.

The version moves in six places

Two that CI compares (pyproject.toml, grapharc/__init__.py) and four where prose states it: the README status line, the deep dive's "on PyPI is that wheel", the basics cookbook's "verified against" line and its grapharc --version transcript, and the serving cookbook's two /health transcripts. All but the README are asserted by tests, which is how they were found — the bump reddened test_cookbook_serving.py and test_deep_dive.py immediately.

One test changed shape

test_version_matches_the_page was assert __version__ == "0.1.5" — a third place the version was written down, and the only one of the three that CI's version check does not compare. Its practical effect was to redden on every release while saying nothing about the page it is named after. It now reads the page and holds every version stated there against __version__, so a bump either moves the page or fails.

Verification

uv run pytest green · uv run ruff check . clean.

Note for the merge

0.1.6 is a patch bump on a 0.x line, consistent with 0.1.1→0.1.5, but it carries one behaviour change: a second grapharc go <run-dir> on an executed plan now exits 2 instead of 0. That is the fix, not a side effect, and --again restores the old behaviour explicitly.

🤖 Generated with Claude Code