refactor: use project-named state file and store prompt/params directly
All checks were successful
Continuous Integration / Build Package (push) Successful in 48s
Continuous Integration / Lint, Check & Test (push) Successful in 1m1s

- State filename now derives from config: cards.bulkgen.yaml produces
  .cards.bulkgen-state.yaml instead of .bulkgen.state.yaml
- Store resolved prompt text and extra params directly in state file
  instead of hashing them, making state files human-readable
- Only file input contents remain hashed (SHA-256)
- Thread project_name through builder and CLI
- Remove hash_string() and _extra_hash() helpers
- Update .gitignore pattern to .*.bulkgen-state.yaml
This commit is contained in:
Konstantin Fickel 2026-02-15 13:56:12 +01:00
parent 870023865d
commit 0ecf1f0f9e
Signed by: kfickel
GPG key ID: A793722F9933C1A5
7 changed files with 98 additions and 82 deletions

2
.gitignore vendored
View file

@ -15,7 +15,7 @@ wheels/
.pre-commit-config.yaml
# bulkgen state
.bulkgen.state.yaml
.*.bulkgen-state.yaml
# Nix
result

View file

@ -147,6 +147,7 @@ async def _build_single_target(
async def run_build(
config: ProjectConfig,
project_dir: Path,
project_name: str,
target: str | None = None,
on_progress: ProgressCallback = _noop_callback,
) -> BuildResult:
@ -171,7 +172,7 @@ async def run_build(
raise ValueError(msg)
graph = get_subgraph_for_target(graph, target)
state = load_state(project_dir)
state = load_state(project_dir, project_name)
generations = get_build_order(graph)
target_names = set(config.targets)
@ -209,7 +210,7 @@ async def run_build(
)
_process_outcomes(outcomes, config, project_dir, state, result, on_progress)
save_state(state, project_dir)
save_state(state, project_dir, project_name)
return result

View file

@ -14,9 +14,21 @@ from bulkgen.builder import BuildEvent, BuildResult, run_build
from bulkgen.config import ProjectConfig, load_config
from bulkgen.graph import build_graph, get_build_order
from bulkgen.providers.registry import get_all_models
from bulkgen.state import state_filename
app = typer.Typer(name="bulkgen", help="AI artifact build tool.")
_CONFIG_SUFFIX = ".bulkgen.yaml"
def _project_name(config_path: Path) -> str:
"""Derive the project name from a config path.
``cards.bulkgen.yaml`` ``cards``
"""
name = config_path.name
return name.removesuffix(_CONFIG_SUFFIX)
def _find_config(directory: Path) -> Path:
"""Find the single ``*.bulkgen.yaml`` file in *directory*."""
@ -49,7 +61,7 @@ def _format_elapsed(seconds: float) -> str:
def _run_build(
config: ProjectConfig, project_dir: Path, target: str | None
config: ProjectConfig, project_dir: Path, project_name: str, target: str | None
) -> tuple[BuildResult, float]:
"""Run the async build with click-styled progress output and timing."""
@ -85,7 +97,7 @@ def _run_build(
start = time.monotonic()
result = asyncio.run(
run_build(config, project_dir, target, on_progress=on_progress)
run_build(config, project_dir, project_name, target, on_progress=on_progress)
)
elapsed = time.monotonic() - start
@ -102,10 +114,11 @@ def build(
project_dir = Path.cwd()
config_path = _find_config(project_dir)
config = load_config(config_path)
name = _project_name(config_path)
click.echo(click.style("bulkgen", fg="cyan", bold=True) + " building targets...\n")
result, elapsed = _run_build(config, project_dir, target)
result, elapsed = _run_build(config, project_dir, name, target)
# Summary
click.echo("")
@ -131,6 +144,7 @@ def clean() -> None:
project_dir = Path.cwd()
config_path = _find_config(project_dir)
config = load_config(config_path)
state_name = state_filename(_project_name(config_path))
removed = 0
for target_name in config.targets:
@ -140,10 +154,10 @@ def clean() -> None:
click.echo(click.style(" rm ", fg="red") + target_name)
removed += 1
state_path = project_dir / ".bulkgen.state.yaml"
state_path = project_dir / state_name
if state_path.exists():
state_path.unlink()
click.echo(click.style(" rm ", fg="red") + ".bulkgen.state.yaml")
click.echo(click.style(" rm ", fg="red") + state_name)
click.echo(click.style(f"\nCleaned {removed} artifact(s)", bold=True))

View file

@ -1,4 +1,4 @@
"""Incremental build state tracking via ``.bulkgen.state.yaml``."""
"""Incremental build state tracking via ``.<project>.bulkgen-state.yaml``."""
from __future__ import annotations
@ -8,16 +8,23 @@ from pathlib import Path
import yaml
from pydantic import BaseModel
STATE_FILENAME = ".bulkgen.state.yaml"
def state_filename(project_name: str) -> str:
"""Return the state filename for a given project name.
For a config file named ``cards.bulkgen.yaml`` the project name is
``cards`` and the state file is ``.cards.bulkgen-state.yaml``.
"""
return f".{project_name}.bulkgen-state.yaml"
class TargetState(BaseModel):
"""Recorded state of a single target from its last successful build."""
input_hashes: dict[str, str]
prompt_hash: str
prompt: str
model: str
extra_hash: str = ""
extra_params: dict[str, object] = {}
class BuildState(BaseModel):
@ -35,14 +42,9 @@ def hash_file(path: Path) -> str:
return h.hexdigest()
def hash_string(value: str) -> str:
"""Compute the SHA-256 hex digest of a string."""
return hashlib.sha256(value.encode("utf-8")).hexdigest()
def load_state(project_dir: Path) -> BuildState:
def load_state(project_dir: Path, project_name: str) -> BuildState:
"""Load build state from disk, returning empty state if the file is missing."""
state_path = project_dir / STATE_FILENAME
state_path = project_dir / state_filename(project_name)
if not state_path.exists():
return BuildState()
with state_path.open() as f:
@ -52,20 +54,13 @@ def load_state(project_dir: Path) -> BuildState:
return BuildState.model_validate(raw)
def save_state(state: BuildState, project_dir: Path) -> None:
def save_state(state: BuildState, project_dir: Path, project_name: str) -> None:
"""Persist build state to disk."""
state_path = project_dir / STATE_FILENAME
state_path = project_dir / state_filename(project_name)
with state_path.open("w") as f:
yaml.dump(state.model_dump(), f, default_flow_style=False, sort_keys=False)
def _extra_hash(params: dict[str, object]) -> str:
"""Hash extra target parameters (width, height, etc.) for change detection."""
if not params:
return ""
return hash_string(str(sorted(params.items())))
def is_target_dirty(
target_name: str,
*,
@ -98,10 +93,10 @@ def is_target_dirty(
if prev.model != model:
return True
if prev.prompt_hash != hash_string(resolved_prompt):
if prev.prompt != resolved_prompt:
return True
if prev.extra_hash != _extra_hash(extra_params):
if prev.extra_params != extra_params:
return True
for dep_path in dep_files:
@ -131,7 +126,7 @@ def record_target_state(
state.targets[target_name] = TargetState(
input_hashes=input_hashes,
prompt_hash=hash_string(resolved_prompt),
prompt=resolved_prompt,
model=model,
extra_hash=_extra_hash(extra_params),
extra_params=extra_params,
)

View file

@ -23,6 +23,8 @@ from bulkgen.state import load_state
WriteConfig = Callable[[dict[str, object]], ProjectConfig]
_PROJECT = "project"
_FAKE_TEXT_MODELS = [
ModelInfo(
@ -196,7 +198,7 @@ class TestRunBuild:
self, project_dir: Path, simple_text_config: ProjectConfig
) -> None:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result = await run_build(simple_text_config, project_dir)
result = await run_build(simple_text_config, project_dir, _PROJECT)
assert result.built == ["output.txt"]
assert result.skipped == []
@ -207,7 +209,7 @@ class TestRunBuild:
self, project_dir: Path, multi_target_config: ProjectConfig
) -> None:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result = await run_build(multi_target_config, project_dir)
result = await run_build(multi_target_config, project_dir, _PROJECT)
assert "summary.md" in result.built
assert "final.txt" in result.built
@ -222,10 +224,10 @@ class TestRunBuild:
self, project_dir: Path, simple_text_config: ProjectConfig
) -> None:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result1 = await run_build(simple_text_config, project_dir)
result1 = await run_build(simple_text_config, project_dir, _PROJECT)
assert result1.built == ["output.txt"]
result2 = await run_build(simple_text_config, project_dir)
result2 = await run_build(simple_text_config, project_dir, _PROJECT)
assert result2.skipped == ["output.txt"]
assert result2.built == []
@ -234,11 +236,11 @@ class TestRunBuild:
) -> None:
config1 = write_config({"targets": {"out.txt": {"prompt": "version 1"}}})
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
r1 = await run_build(config1, project_dir)
r1 = await run_build(config1, project_dir, _PROJECT)
assert r1.built == ["out.txt"]
config2 = write_config({"targets": {"out.txt": {"prompt": "version 2"}}})
r2 = await run_build(config2, project_dir)
r2 = await run_build(config2, project_dir, _PROJECT)
assert r2.built == ["out.txt"]
async def test_rebuild_after_input_change(
@ -249,11 +251,11 @@ class TestRunBuild:
{"targets": {"out.md": {"prompt": "x", "inputs": ["data.txt"]}}}
)
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
r1 = await run_build(config, project_dir)
r1 = await run_build(config, project_dir, _PROJECT)
assert r1.built == ["out.md"]
_ = (project_dir / "data.txt").write_text("modified")
r2 = await run_build(config, project_dir)
r2 = await run_build(config, project_dir, _PROJECT)
assert r2.built == ["out.md"]
async def test_selective_build_single_target(
@ -261,7 +263,7 @@ class TestRunBuild:
) -> None:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result = await run_build(
multi_target_config, project_dir, target="summary.md"
multi_target_config, project_dir, _PROJECT, target="summary.md"
)
assert "summary.md" in result.built
@ -274,7 +276,7 @@ class TestRunBuild:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
with pytest.raises(ValueError, match="Unknown target"):
_ = await run_build(
simple_text_config, project_dir, target="nonexistent.txt"
simple_text_config, project_dir, _PROJECT, target="nonexistent.txt"
)
async def test_failed_target_isolates_independent(
@ -322,7 +324,7 @@ class TestRunBuild:
"bulkgen.builder._create_providers",
return_value=[routing_provider, FakeImageProvider()],
):
result = await run_build(config, project_dir)
result = await run_build(config, project_dir, _PROJECT)
assert "fail.txt" in result.failed
assert "ok.txt" in result.built
@ -343,7 +345,7 @@ class TestRunBuild:
"bulkgen.builder._create_providers",
return_value=[FailingTextProvider(), FakeImageProvider()],
):
result = await run_build(config, project_dir)
result = await run_build(config, project_dir, _PROJECT)
assert "base.txt" in result.failed
assert "child.txt" in result.failed
@ -356,7 +358,7 @@ class TestRunBuild:
"bulkgen.builder._create_providers",
return_value=[],
):
result = await run_build(simple_text_config, project_dir)
result = await run_build(simple_text_config, project_dir, _PROJECT)
assert "output.txt" in result.failed
assert "No provider available" in result.failed["output.txt"]
@ -373,9 +375,9 @@ class TestRunBuild:
}
)
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
_ = await run_build(config, project_dir)
_ = await run_build(config, project_dir, _PROJECT)
state = load_state(project_dir)
state = load_state(project_dir, _PROJECT)
assert "a.txt" in state.targets
assert "b.txt" in state.targets
@ -385,7 +387,7 @@ class TestRunBuild:
config = write_config({"targets": {"out.txt": {"prompt": prompt_file.name}}})
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result = await run_build(config, project_dir)
result = await run_build(config, project_dir, _PROJECT)
assert result.built == ["out.txt"]
content = (project_dir / "out.txt").read_text()
@ -395,12 +397,12 @@ class TestRunBuild:
self, project_dir: Path, simple_text_config: ProjectConfig
) -> None:
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
r1 = await run_build(simple_text_config, project_dir)
r1 = await run_build(simple_text_config, project_dir, _PROJECT)
assert r1.built == ["output.txt"]
(project_dir / "output.txt").unlink()
r2 = await run_build(simple_text_config, project_dir)
r2 = await run_build(simple_text_config, project_dir, _PROJECT)
assert r2.built == ["output.txt"]
async def test_diamond_dependency_all_built(
@ -420,7 +422,7 @@ class TestRunBuild:
}
)
with patch("bulkgen.builder._create_providers", return_value=_fake_providers()):
result = await run_build(config, project_dir)
result = await run_build(config, project_dir, _PROJECT)
assert set(result.built) == {"left.md", "right.md", "merge.txt"}
assert result.failed == {}

View file

@ -132,7 +132,8 @@ class TestBuildCommand:
assert result.exit_code == 0
call_args = mock_run.call_args
assert call_args[0][2] == "output.txt"
# positional args: (config, project_dir, project_name, target)
assert call_args[0][3] == "output.txt"
class TestCleanCommand:
@ -141,7 +142,8 @@ class TestCleanCommand:
def test_clean_removes_targets(self, cli_project: Path) -> None:
_ = (cli_project / "output.txt").write_text("generated")
_ = (cli_project / "image.png").write_bytes(b"\x89PNG")
_ = (cli_project / ".bulkgen.state.yaml").write_text("targets: {}")
state_file = ".project.bulkgen-state.yaml"
_ = (cli_project / state_file).write_text("targets: {}")
with patch("bulkgen.cli.Path") as mock_path_cls:
mock_path_cls.cwd.return_value = cli_project
@ -150,7 +152,7 @@ class TestCleanCommand:
assert result.exit_code == 0
assert not (cli_project / "output.txt").exists()
assert not (cli_project / "image.png").exists()
assert not (cli_project / ".bulkgen.state.yaml").exists()
assert not (cli_project / state_file).exists()
assert "Cleaned 2 artifact(s)" in result.output
def test_clean_no_artifacts(self, cli_project: Path) -> None:

View file

@ -10,13 +10,15 @@ from bulkgen.state import (
BuildState,
TargetState,
hash_file,
hash_string,
is_target_dirty,
load_state,
record_target_state,
save_state,
state_filename,
)
_PROJECT = "test"
class TestHashFunctions:
"""Test hashing helpers."""
@ -34,18 +36,22 @@ class TestHashFunctions:
h2 = hash_file(f)
assert h1 != h2
def test_hash_string_deterministic(self) -> None:
assert hash_string("abc") == hash_string("abc")
def test_hash_string_differs(self) -> None:
assert hash_string("abc") != hash_string("xyz")
class TestStateFilename:
"""Test state filename derivation."""
def test_state_filename(self) -> None:
assert state_filename("cards") == ".cards.bulkgen-state.yaml"
def test_state_filename_simple(self) -> None:
assert state_filename("project") == ".project.bulkgen-state.yaml"
class TestStatePersistence:
"""Test save/load round-trip of build state."""
def test_load_missing_file_returns_empty(self, project_dir: Path) -> None:
state = load_state(project_dir)
state = load_state(project_dir, _PROJECT)
assert state.targets == {}
def test_save_and_load_round_trip(self, project_dir: Path) -> None:
@ -53,40 +59,36 @@ class TestStatePersistence:
targets={
"out.txt": TargetState(
input_hashes={"dep.txt": "abc123"},
prompt_hash="prompt_hash_val",
prompt="Generate something",
model="mistral-large-latest",
extra_hash="",
extra_params={},
)
}
)
save_state(state, project_dir)
loaded = load_state(project_dir)
save_state(state, project_dir, _PROJECT)
loaded = load_state(project_dir, _PROJECT)
assert loaded.targets["out.txt"].model == "mistral-large-latest"
assert loaded.targets["out.txt"].input_hashes == {"dep.txt": "abc123"}
assert loaded.targets["out.txt"].prompt_hash == "prompt_hash_val"
assert loaded.targets["out.txt"].prompt == "Generate something"
def test_load_empty_yaml(self, project_dir: Path) -> None:
_ = (project_dir / ".bulkgen.state.yaml").write_text("")
state = load_state(project_dir)
_ = (project_dir / state_filename(_PROJECT)).write_text("")
state = load_state(project_dir, _PROJECT)
assert state.targets == {}
def test_save_overwrites_existing(self, project_dir: Path) -> None:
state1 = BuildState(
targets={
"a.txt": TargetState(input_hashes={}, prompt_hash="h1", model="m1")
}
targets={"a.txt": TargetState(input_hashes={}, prompt="p1", model="m1")}
)
save_state(state1, project_dir)
save_state(state1, project_dir, _PROJECT)
state2 = BuildState(
targets={
"b.txt": TargetState(input_hashes={}, prompt_hash="h2", model="m2")
}
targets={"b.txt": TargetState(input_hashes={}, prompt="p2", model="m2")}
)
save_state(state2, project_dir)
save_state(state2, project_dir, _PROJECT)
loaded = load_state(project_dir)
loaded = load_state(project_dir, _PROJECT)
assert "b.txt" in loaded.targets
assert "a.txt" not in loaded.targets
@ -95,16 +97,16 @@ class TestStatePersistence:
targets={
"out.txt": TargetState(
input_hashes={"f.txt": "hash"},
prompt_hash="ph",
prompt="do something",
model="m",
extra_hash="eh",
extra_params={"width": 512},
)
}
)
save_state(state, project_dir)
save_state(state, project_dir, _PROJECT)
raw: object = yaml.safe_load( # pyright: ignore[reportAny]
(project_dir / ".bulkgen.state.yaml").read_text()
(project_dir / state_filename(_PROJECT)).read_text()
)
assert isinstance(raw, dict)
assert "targets" in raw
@ -296,9 +298,9 @@ class TestRecordAndDirtyRoundTrip:
state=state,
project_dir=project_dir,
)
save_state(state, project_dir)
save_state(state, project_dir, _PROJECT)
loaded_state = load_state(project_dir)
loaded_state = load_state(project_dir, _PROJECT)
assert not is_target_dirty(
"result.md",
resolved_prompt="do the thing",