Parse https://ollama.com/library/ syntax

People search for ollama models using the web ui, this change
allows one to copy the url from the browser and for it to be
compatible with ramalama run.

Also pull smaller models "smollm" to accelerate the builds.

Signed-off-by: Eric Curtin <[email protected]>

ramalama/cli.py

@@ -967,7 +967,7 @@ def rm_cli(args):
 def New(model, args):
     if model.startswith("huggingface://") or model.startswith("hf://") or model.startswith("hf.co/"):
         return Huggingface(model)
-    if model.startswith("ollama"):
+    if model.startswith("ollama://") or "ollama.com/library/" in model:
         return Ollama(model)
     if model.startswith("oci://") or model.startswith("docker://"):
         return OCI(model, args.engine)

ramalama/huggingface.py

@@ -2,7 +2,7 @@
 import pathlib
 import urllib.request
 from ramalama.common import available, run_cmd, exec_cmd, download_file, verify_checksum, perror
-from ramalama.model import Model
+from ramalama.model import Model, rm_until_substring
 
 missing_huggingface = """
 Optional: Huggingface models require the huggingface-cli module.
@@ -33,9 +33,8 @@ def fetch_checksum_from_api(url):
 
 class Huggingface(Model):
     def __init__(self, model):
-        model = model.removeprefix("huggingface://")
-        model = model.removeprefix("hf://")
-        model = model.removeprefix("hf.co/")
+        model = rm_until_substring(model, "hf.co/")
+        model = rm_until_substring(model, "://")
         super().__init__(model)
         self.type = "huggingface"
         split = self.model.rsplit("/", 1)

ramalama/model.py

@@ -486,3 +486,12 @@ def distinfo_volume():
         return ""
 
     return f"-v{path}:/usr/share/ramalama/{dist_info}:ro"
+
+
+def rm_until_substring(model, substring):
+    pos = model.find(substring)
+    if pos == -1:
+        return model
+
+    # Create a new string starting after the found substring
+    return ''.join(model[i] for i in range(pos + len(substring), len(model)))

ramalama/ollama.py

@@ -2,7 +2,7 @@
 import urllib.request
 import json
 from ramalama.common import run_cmd, verify_checksum, download_file
-from ramalama.model import Model
+from ramalama.model import Model, rm_until_substring
 
 
 def fetch_manifest_data(registry_head, model_tag, accept):
@@ -60,7 +60,9 @@ def init_pull(repos, accept, registry_head, model_name, model_tag, models, model
 
 class Ollama(Model):
     def __init__(self, model):
-        super().__init__(model.removeprefix("ollama://"))
+        model = rm_until_substring(model, "ollama.com/library/")
+        model = rm_until_substring(model, "://")
+        super().__init__(model)
         self.type = "Ollama"
 
     def _local(self, args):

ramalama/url.py

@@ -1,17 +1,13 @@
 import os
 from ramalama.common import download_file
-from ramalama.model import Model
+from ramalama.model import Model, rm_until_substring
+from urllib.parse import urlparse
 
 
 class URL(Model):
     def __init__(self, model):
-        self.type = ""
-        for prefix in ["file", "http", "https"]:
-            if model.startswith(f"{prefix}://"):
-                self.type = prefix
-                model = model.removeprefix(f"{prefix}://")
-                break
-
+        self.type = urlparse(model).scheme
+        model = rm_until_substring(model, "://")
         super().__init__(model)
         split = self.model.rsplit("/", 1)
         self.directory = split[0].removeprefix("/") if len(split) > 1 else ""

test/system/050-pull.bats

@@ -14,15 +14,15 @@ load setup_suite
 @test "ramalama pull ollama" {
     run_ramalama pull tiny
     run_ramalama rm tiny
-    run_ramalama pull ollama://tinyllama
+    run_ramalama pull https://ollama.com/library/smollm:135m
     run_ramalama list
-    is "$output" ".*ollama://tinyllama" "image was actually pulled locally"
+    is "$output" ".*ollama://smollm:135m" "image was actually pulled locally"
 
-    RAMALAMA_TRANSPORT=ollama run_ramalama pull tinyllama:1.1b
-    run_ramalama pull ollama://tinyllama:1.1b
+    RAMALAMA_TRANSPORT=ollama run_ramalama pull smollm:360m
+    run_ramalama pull ollama://smollm:360m
     run_ramalama list
-    is "$output" ".*ollama://tinyllama:1.1b" "image was actually pulled locally"
-    run_ramalama rm ollama://tinyllama ollama://tinyllama:1.1b
+    is "$output" ".*ollama://smollm:360m" "image was actually pulled locally"
+    run_ramalama rm ollama://smollm:135m ollama://smollm:360m
 
     random_image_name=i_$(safename)
     run_ramalama 1 pull ${random_image_name}