furu 0.0.4__tar.gz → 0.0.5__tar.gz

{furu-0.0.4 → furu-0.0.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: furu
-Version: 0.0.4
+Version: 0.0.5
 Summary: Cacheable, nested pipelines for Python. Define computations as configs; furu handles caching, state tracking, and result reuse across runs.
 Author: Herman Brunborg
 Author-email: Herman Brunborg <herman@brunborg.com>
@@ -459,8 +459,11 @@ The `/api/experiments` endpoint supports:
 |----------|---------|-------------|
 | `FURU_PATH` | `<project>/furu-data` | Base storage directory for non-versioned artifacts |
 | `FURU_VERSION_CONTROLLED_PATH` | `<project>/furu-data/artifacts` | Override version-controlled storage root |
+| `FURU_SUBMITIT_PATH` | `<FURU_PATH>/submitit` | Override submitit logs root |
 | `FURU_LOG_LEVEL` | `INFO` | Console verbosity (`DEBUG`, `INFO`, `WARNING`, `ERROR`) |
-| `FURU_IGNORE_DIFF` | `false` | Skip embedding git diff in metadata |
+| `FURU_RICH_UNCAUGHT_TRACEBACKS` | `true` | Use Rich for exception formatting (set `0` to disable) |
+| `FURU_RECORD_GIT` | `cached` | Git provenance capture: `ignore` skips git metadata, `cached` records once per process, `uncached` records every time |
+| `FURU_ALLOW_NO_GIT_ORIGIN` | `false` | Allow missing git `origin` when recording git metadata (invalid with `FURU_RECORD_GIT=ignore`) |
 | `FURU_ALWAYS_RERUN` | `""` | Comma-separated class qualnames to always rerun (use `ALL` to bypass cache globally; cannot combine with other entries; entries must be importable) |
 | `FURU_RETRY_FAILED` | `true` | Retry failed artifacts by default (set to `0` to keep failures sticky) |
 | `FURU_MAX_COMPUTE_RETRIES` | `3` | Maximum compute retries per node after the first failure |
@@ -469,12 +472,21 @@ The `/api/experiments` endpoint supports:
 | `FURU_WAIT_LOG_EVERY_SECS` | `10` | Interval between "waiting" log messages |
 | `FURU_STALE_AFTER_SECS` | `1800` | Consider running jobs stale after this duration |
 | `FURU_LEASE_SECS` | `120` | Compute lock lease duration |
-| `FURU_HEARTBEAT_SECS` | `lease/3` | Heartbeat interval for running jobs |
+| `FURU_HEARTBEAT_SECS` | `lease/3` | Heartbeat interval for running jobs (min 1s) |
 | `FURU_PREEMPT_MAX` | `5` | Maximum submitit requeues on preemption |
 | `FURU_CANCELLED_IS_PREEMPTED` | `false` | Treat SLURM CANCELLED as preempted |
-| `FURU_RICH_UNCAUGHT_TRACEBACKS` | `true` | Use Rich for exception formatting |
+| `SLURM_JOB_ID` | unset | Read-only; set by Slurm to record job id and enable submitit context |
 
-Local `.env` files are loaded automatically if `python-dotenv` is installed.
+Local `.env` files are not loaded automatically. Call `furu.load_env()` when you
+want to load `.env` values (requires `python-dotenv`).
+
+### Test and CI Environment Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `FURU_DASHBOARD_DEV_DATA_DIR` | unset | Override data dir for `make dashboard-dev` (defaults to a temp dir) |
+| `FURU_E2E_DATA_DIR` | unset | Required for Playwright e2e runs; used as the data root and to set `FURU_PATH` |
+| `CI` | unset | Enables CI-friendly Playwright settings (retries, single worker, traces, screenshots, video) |
 
 ### Programmatic Configuration
 
@@ -487,7 +499,7 @@ furu.set_furu_root(Path("/my/storage"))
 root = furu.get_furu_root()
 
 # Access config directly
-furu.FURU_CONFIG.ignore_git_diff = True
+furu.FURU_CONFIG.record_git = "uncached"
 furu.FURU_CONFIG.poll_interval = 5.0
 ```
 

{furu-0.0.4 → furu-0.0.5}/README.md

@@ -440,8 +440,11 @@ The `/api/experiments` endpoint supports:
 |----------|---------|-------------|
 | `FURU_PATH` | `<project>/furu-data` | Base storage directory for non-versioned artifacts |
 | `FURU_VERSION_CONTROLLED_PATH` | `<project>/furu-data/artifacts` | Override version-controlled storage root |
+| `FURU_SUBMITIT_PATH` | `<FURU_PATH>/submitit` | Override submitit logs root |
 | `FURU_LOG_LEVEL` | `INFO` | Console verbosity (`DEBUG`, `INFO`, `WARNING`, `ERROR`) |
-| `FURU_IGNORE_DIFF` | `false` | Skip embedding git diff in metadata |
+| `FURU_RICH_UNCAUGHT_TRACEBACKS` | `true` | Use Rich for exception formatting (set `0` to disable) |
+| `FURU_RECORD_GIT` | `cached` | Git provenance capture: `ignore` skips git metadata, `cached` records once per process, `uncached` records every time |
+| `FURU_ALLOW_NO_GIT_ORIGIN` | `false` | Allow missing git `origin` when recording git metadata (invalid with `FURU_RECORD_GIT=ignore`) |
 | `FURU_ALWAYS_RERUN` | `""` | Comma-separated class qualnames to always rerun (use `ALL` to bypass cache globally; cannot combine with other entries; entries must be importable) |
 | `FURU_RETRY_FAILED` | `true` | Retry failed artifacts by default (set to `0` to keep failures sticky) |
 | `FURU_MAX_COMPUTE_RETRIES` | `3` | Maximum compute retries per node after the first failure |
@@ -450,12 +453,21 @@ The `/api/experiments` endpoint supports:
 | `FURU_WAIT_LOG_EVERY_SECS` | `10` | Interval between "waiting" log messages |
 | `FURU_STALE_AFTER_SECS` | `1800` | Consider running jobs stale after this duration |
 | `FURU_LEASE_SECS` | `120` | Compute lock lease duration |
-| `FURU_HEARTBEAT_SECS` | `lease/3` | Heartbeat interval for running jobs |
+| `FURU_HEARTBEAT_SECS` | `lease/3` | Heartbeat interval for running jobs (min 1s) |
 | `FURU_PREEMPT_MAX` | `5` | Maximum submitit requeues on preemption |
 | `FURU_CANCELLED_IS_PREEMPTED` | `false` | Treat SLURM CANCELLED as preempted |
-| `FURU_RICH_UNCAUGHT_TRACEBACKS` | `true` | Use Rich for exception formatting |
+| `SLURM_JOB_ID` | unset | Read-only; set by Slurm to record job id and enable submitit context |
 
-Local `.env` files are loaded automatically if `python-dotenv` is installed.
+Local `.env` files are not loaded automatically. Call `furu.load_env()` when you
+want to load `.env` values (requires `python-dotenv`).
+
+### Test and CI Environment Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `FURU_DASHBOARD_DEV_DATA_DIR` | unset | Override data dir for `make dashboard-dev` (defaults to a temp dir) |
+| `FURU_E2E_DATA_DIR` | unset | Required for Playwright e2e runs; used as the data root and to set `FURU_PATH` |
+| `CI` | unset | Enables CI-friendly Playwright settings (retries, single worker, traces, screenshots, video) |
 
 ### Programmatic Configuration
 
@@ -468,7 +480,7 @@ furu.set_furu_root(Path("/my/storage"))
 root = furu.get_furu_root()
 
 # Access config directly
-furu.FURU_CONFIG.ignore_git_diff = True
+furu.FURU_CONFIG.record_git = "uncached"
 furu.FURU_CONFIG.poll_interval = 5.0
 ```
 

{furu-0.0.4 → furu-0.0.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "furu"
-version = "0.0.4"
+version = "0.0.5"
 description = "Cacheable, nested pipelines for Python. Define computations as configs; furu handles caching, state tracking, and result reuse across runs."
 readme = "README.md"
 authors = [

{furu-0.0.4 → furu-0.0.5}/src/furu/config.py

@@ -1,6 +1,10 @@
 import os
 from importlib import import_module
 from pathlib import Path
+from typing import Literal, cast
+
+
+RecordGitMode = Literal["ignore", "cached", "uncached"]
 
 
 class FuruConfig:
@@ -41,21 +45,14 @@ class FuruConfig:
             "true",
             "yes",
         }
-        self.ignore_git_diff = os.getenv("FURU_IGNORE_DIFF", "0").lower() in {
-            "1",
-            "true",
-            "yes",
-        }
-        self.require_git = os.getenv("FURU_REQUIRE_GIT", "1").lower() in {
-            "1",
-            "true",
-            "yes",
-        }
-        self.require_git_remote = os.getenv("FURU_REQUIRE_GIT_REMOTE", "1").lower() in {
-            "1",
-            "true",
-            "yes",
-        }
+        self.record_git = self._parse_record_git(os.getenv("FURU_RECORD_GIT", "cached"))
+        self.allow_no_git_origin = self._parse_bool(
+            os.getenv("FURU_ALLOW_NO_GIT_ORIGIN", "0")
+        )
+        if self.allow_no_git_origin and self.record_git == "ignore":
+            raise ValueError(
+                "FURU_ALLOW_NO_GIT_ORIGIN cannot be enabled when FURU_RECORD_GIT=ignore"
+            )
         always_rerun_items = {
             item.strip()
             for item in os.getenv("FURU_ALWAYS_RERUN", "").split(",")
@@ -77,35 +74,25 @@ class FuruConfig:
             "FURU_CANCELLED_IS_PREEMPTED", "false"
         ).lower() in {"1", "true", "yes"}
 
-        # Parse FURU_CACHE_METADATA: "never", "forever", or duration like "5m", "1h"
-        # Default: "5m" (5 minutes) - balances performance with freshness
-        self.cache_metadata_ttl_sec: float | None = self._parse_cache_duration(
-            os.getenv("FURU_CACHE_METADATA", "5m")
-        )
-
     @staticmethod
-    def _parse_cache_duration(value: str) -> float | None:
-        """Parse cache duration string into seconds. Returns None for 'never', float('inf') for 'forever'."""
-        value = value.strip().lower()
-        if value in {"never", "0", "false", "no"}:
-            return None  # No caching
-        if value in {"forever", "inf", "true", "yes", "1"}:
-            return float("inf")  # Cache forever
-
-        # Parse duration like "5m", "1h", "30s"
-        import re
-
-        match = re.match(r"^(\d+(?:\.\d+)?)\s*([smh]?)$", value)
-        if not match:
+    def _parse_bool(value: str) -> bool:
+        return value.strip().lower() in {"1", "true", "yes"}
+
+    @classmethod
+    def _parse_record_git(cls, value: str) -> RecordGitMode:
+        normalized = value.strip().lower()
+        allowed = {"ignore", "cached", "uncached"}
+        if normalized not in allowed:
             raise ValueError(
-                f"Invalid FURU_CACHE_METADATA value: {value!r}. "
-                "Use 'never', 'forever', or duration like '5m', '1h', '30s'"
+                "FURU_RECORD_GIT must be one of 'ignore', 'cached', or 'uncached'"
             )
+        return cast(RecordGitMode, normalized)
 
-        num = float(match.group(1))
-        unit = match.group(2) or "s"
-        multipliers = {"s": 1, "m": 60, "h": 3600}
-        return num * multipliers[unit]
+    @property
+    def cache_metadata_ttl_sec(self) -> float | None:
+        if self.record_git == "cached":
+            return float("inf")
+        return None
 
     def get_root(self, version_controlled: bool = False) -> Path:
         """Get root directory for storage (version_controlled uses its own root)."""