paratext-cli 0.2.0__tar.gz → 0.2.1__tar.gz

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: paratext-cli
-Version: 0.2.0
+Version: 0.2.1
 Summary: Modular, project-based metadata extraction from digitised library/archive collections with a multimodal model
 License-Expression: Apache-2.0
 License-File: LICENSE
@@ -159,7 +159,9 @@ prompt version, keyed on the prompt's hash:
   `-r3`, …). The UI shows the two most recent rounds side by side and highlights
   what changed. The flag is needed because a run resumes on sample id: without
   it the existing extractions are already there, so the model is never called.
-  paratext stops and says so rather than resuming into a stale file.
+  paratext stops and says so rather than resuming into a stale file. On a small
+  collection, `re-extract = true` in `paratext.toml` makes it the default and
+  the loop needs no flag.
 - **Re-run the same prompt** (a resume, or a bigger `--limit`) → the current round
   is updated in place, keeping the annotations you've already made.
 

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/README.md

@@ -134,7 +134,9 @@ prompt version, keyed on the prompt's hash:
   `-r3`, …). The UI shows the two most recent rounds side by side and highlights
   what changed. The flag is needed because a run resumes on sample id: without
   it the existing extractions are already there, so the model is never called.
-  paratext stops and says so rather than resuming into a stale file.
+  paratext stops and says so rather than resuming into a stale file. On a small
+  collection, `re-extract = true` in `paratext.toml` makes it the default and
+  the loop needs no flag.
 - **Re-run the same prompt** (a resume, or a bigger `--limit`) → the current round
   is updated in place, keeping the annotations you've already made.
 

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/paratext.example.toml

@@ -19,6 +19,11 @@ output     = "output/cards.jsonl"
 # prompt no longer opens a new round to compare against. Leave it unset unless
 # that's what you want.
 # review-out = "review/cards"
+# Editing the prompt makes existing extractions stale, and paratext stops rather
+# than resuming into them. On a small collection where a re-run is cheap, this
+# discards them and calls the model again instead — the run/edit/re-run loop
+# then needs no flag. Leave it off for anything big or metered.
+# re-extract = true
 
 # Verbatim passthrough into the request body — for provider parameters paratext
 # doesn't model. Reasoning controls live here, and every provider spells them

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/pyproject.toml

@@ -3,7 +3,7 @@
 # the plain name on PyPI belongs to an unrelated project, so the distribution
 # carries the `-cli` suffix.
 name = "paratext-cli"
-version = "0.2.0"
+version = "0.2.1"
 description = "Modular, project-based metadata extraction from digitised library/archive collections with a multimodal model"
 readme = "README.md"
 # Core is pure-Python + numpy/pillow/pydantic. The [detector] extra needs

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/src/paratext/__init__.py

@@ -1,3 +1,3 @@
 """paratext — metadata extraction from digitised collections with a multimodal model."""
 
-__version__ = "0.2.0"
+__version__ = "0.2.1"

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/src/paratext/cli.py

@@ -61,6 +61,7 @@ HARDCODED_DEFAULTS: dict = {
     "review_port": DEFAULT_PORT,
     "no_structured": False,
     "skip_preflight": False,
+    "re_extract": False,
     "max_tokens": None,  # None → runner.DEFAULT_MAX_TOKENS, or the project's own
 }
 
@@ -214,7 +215,7 @@ def _do_extract(args: argparse.Namespace) -> None:
         energy=energy,
         max_tokens=getattr(args, "max_tokens", None),
         extra_body=config_extra_body(args.project),
-        re_extract=getattr(args, "re_extract", False),
+        re_extract=args.re_extract,
     )
 
 
@@ -675,9 +676,9 @@ def _add_extract_args(p: argparse.ArgumentParser) -> None:
     p.add_argument("--base-url", default=None, help="model endpoint base URL (OpenAI-compatible)")
     p.add_argument("--api-key", default=None, help="API key for the endpoint (often unused)")
     p.add_argument("--limit", type=int, default=None, help="Process at most N inputs")
-    p.add_argument("--re-extract", action="store_true",
-                   help="Discard existing extractions and call the model again "
-                        "(needed when the prompt or model has changed)")
+    p.add_argument("--re-extract", action="store_true", default=None,
+                   help="Discard existing extractions and call the model again when the "
+                        "prompt or model has changed (config re-extract, else stop and say so)")
     p.add_argument("--max-tokens", type=int, default=None, metavar="N",
                    help="Output-token ceiling per call (default: 8192; raise it for "
                         "models that reason before answering)")

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/src/paratext/config.py

@@ -62,6 +62,7 @@ RECOGNISED = (
     "limit",
     "no_structured",
     "skip_preflight",
+    "re_extract",
     "max_tokens",
 )
 
@@ -232,7 +233,7 @@ def coerce_paths(d: dict) -> dict:
     for key in ("limit", "review_port", "max_tokens"):
         if key in d and isinstance(d[key], str):
             d[key] = int(d[key])
-    for key in ("no_structured", "skip_preflight"):
+    for key in ("no_structured", "skip_preflight", "re_extract"):
         if key in d and isinstance(d[key], str):
             d[key] = d[key].lower() in ("1", "true", "yes", "on")
     return d

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/src/paratext/extract.py

@@ -78,7 +78,9 @@ def _guard_stale_output(output: Path, header: dict, project: str, re_extract: bo
         f"Those records answer a different question, and resume would skip every\n"
         f"sample and call the model zero times. Either:\n"
         f"  paratext run -p {project} --re-extract           # redo them\n"
-        f"  paratext run -p {project} --output <new>.jsonl   # keep both"
+        f"  paratext run -p {project} --output <new>.jsonl   # keep both\n"
+        f"On a small, cheap collection you can make the first the default:\n"
+        f"  re-extract = true   under [project.{project}] in paratext.toml"
     )
 
 

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/tests/test_config.py

@@ -162,3 +162,11 @@ def test_max_tokens_from_the_environment(tmp_path, monkeypatch):
     monkeypatch.chdir(tmp_path)
     monkeypatch.setenv("PARATEXT_MAX_TOKENS", "16384")
     assert cfg.coerce_paths(cfg.load_defaults(None))["max_tokens"] == 16384
+
+
+def test_re_extract_is_recognised_and_coerced():
+    from paratext.config import RECOGNISED, coerce_paths
+
+    assert "re_extract" in RECOGNISED
+    assert coerce_paths({"re_extract": "true"})["re_extract"] is True
+    assert coerce_paths({"re_extract": "no"})["re_extract"] is False

{paratext_cli-0.2.0 → paratext_cli-0.2.1}/uv.lock

@@ -803,7 +803,7 @@ wheels = [
 
 [[package]]
 name = "paratext-cli"
-version = "0.2.0"
+version = "0.2.1"
 source = { editable = "." }
 dependencies = [
     { name = "huggingface-hub" },