FIX: making paras manual dependency

Author: davidmeijer

README.md

@@ -25,6 +25,16 @@ We recommend installing BioCracker in a virtual conda environment, based on the
 conda env create -f environment.yml
 ```
 
+### Installing PARAS
+
+PARAS is used by BioCracker to predict substrate specificities of NRPS adenylation domains.
+
+PARAS has no PyPI package ans must be installed from source manually:
+
+```bash
+pip install "paras @ git+https://github.com/bthedragonmaster/parasect.git@v2.0.0"
+```
+
 ### Installing HMMER2 on macOS Arm64
 
 Use Rosetta to install the x86_64 version of HMMER2:

environment.dev.yml

@@ -15,4 +15,5 @@ dependencies:
   - pip
   - pip:
       - -e .
+      - "paras @ git+https://github.com/bthedragonmaster/parasect.git@v2.0.0"
       - hatch

environment.yml

@@ -14,4 +14,5 @@ dependencies:
   # Extra PyPI deps
   - pip
   - pip:
-      - "biocracker @ git+https://github.com/moltools/biocracker.git"
+      - biocracker
+      - "paras @ git+https://github.com/bthedragonmaster/parasect.git@v2.0.0"

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "biocracker"
-version = "1.0.0"
+version = "1.0.1"
 description = "BioCracker is a parser for antiSMASH output GenBank files"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -15,8 +15,7 @@ keywords = ["parser", "antismash"]
 
 dependencies = [
     "biopython",
-    "tqdm",
-    "paras @ git+https://github.com/bthedragonmaster/parasect.git@v2.0.0"
+    "tqdm"
 ]
 
 classifiers = [

src/biocracker/paras.py

@@ -5,7 +5,14 @@
 from pathlib import Path
 
 import joblib
-from parasect.api import run_paras  # module is called parasect, but we are using the paras model
+
+try:
+    from parasect.api import run_paras  # module is called parasect, but we are using the paras model
+
+    _HAS_PARAS = True
+except ImportError:
+    run_paras = None
+    _HAS_PARAS = False
 
 from biocracker.antismash import DomainRec
 from biocracker.config import LOGGER_NAME, PARAS_CACHE_DIR_NAME, PARAS_MODEL_DOWNLOAD_URL
@@ -14,13 +21,25 @@
 _PARAS_MODEL_CACHE: dict[str, object] = {}
 
 
+def has_parasect() -> bool:
+    """
+    Check if parasect is installed.
+
+    :return: True if parasect is installed, False otherwise
+    """
+    return _HAS_PARAS
+
+
 def _load_paras_model(cache_dir: Path) -> object:
     """
     Load the paras model from disk (cached in memory for reuse).
 
     :param cache_dir: Path to the cache directory
     :return: loaded paras model
     """
+    if not _HAS_PARAS:
+        raise ImportError("paras is not installed, cannot load paras model")
+
     global _PARAS_MODEL_CACHE
 
     # If model already loaded, return it immediately
@@ -64,6 +83,11 @@ def predict_amp_domain_substrate(
     if domain.kind != "AMP-binding":
         return None
 
+    # If parasect is missing, log and return None
+    if not has_parasect():
+        logger.warning("parasect not installed — skipping substrate prediction.")
+        return None
+
     # Define cache directory
     cache_dir = (
         Path(cache_dir_override)