Resolves bad rebase for artifact pulling

ramalama/annotations.py

@@ -3,15 +3,15 @@
 # https://github.com/CloudNativeAI/model-spec
 
 # ArtifactTypeModelManifest specifies the media type for a model manifest.
-ArtifactTypeModelManifest = "application/vnd.cnai.model.manifest.v1+json"
+ArtifactTypeModelManifest = "application/vnd.cncf.model.manifest.v1+json"
 
 # ArtifactTypeModelLayer is the media type used for layers referenced by the
 # manifest.
-ArtifactTypeModelLayer = "application/vnd.cnai.model.layer.v1.tar"
+ArtifactTypeModelLayer = "application/vnd.cncf.model.layer.v1.tar"
 
 # ArtifactTypeModelLayerGzip is the media type used for gzipped layers
 # referenced by the manifest.
-ArtifactTypeModelLayerGzip = "application/vnd.cnai.model.layer.v1.tar+gzip"
+ArtifactTypeModelLayerGzip = "application/vnd.cncf.model.layer.v1.tar+gzip"
 
 # AnnotationCreated is the annotation key for the date and time on which the
 # model was built (date-time string as defined by RFC 3339).
@@ -76,7 +76,7 @@
 # DEPRECATED: Migrate to AnnotationFilepath
 # AnnotationModel is the annotation key for the layer is a model file (boolean),
 # such as `true` or `false`.
-AnnotationModel = "org.cnai.model.model"
+AnnotationModel = "org.cncf.model.model"
 
 # AnnotationFilepath is the annotation key for the file path of the layer.
-AnnotationFilepath = "org.cnai.model.filepath"
+AnnotationFilepath = "org.cncf.model.filepath"

ramalama/chat_providers/api_providers.py

@@ -13,7 +13,8 @@ def get_provider_api_key(scheme: str) -> str | None:
     """Return a configured API key for the given provider scheme, if any."""
 
     if resolver := PROVIDER_API_KEY_RESOLVERS.get(scheme):
-        return resolver()
+        if key := resolver():
+            return key
     return get_config().api_key
 
 
@@ -26,7 +27,7 @@ def get_provider_api_key(scheme: str) -> str | None:
 
 def get_chat_provider(scheme: str) -> ChatProvider:
     if (resolver := DEFAULT_PROVIDERS.get(scheme, None)) is None:
-        raise ValueError(f"No support chat providers for {scheme}")
+        raise ValueError(f"No supported chat provider for {scheme}")
     return resolver()
 
 

ramalama/chat_providers/base.py

@@ -185,7 +185,7 @@ def list_models(self) -> list[str]:
             if exc.code in (401, 403):
                 message = (
                     f"Could not authenticate with {self.provider}."
-                    "The provided API key was either missing or invalid.\n"
+                    " The provided API key was either missing or invalid.\n"
                     f"Set RAMALAMA_API_KEY or ramalama.provider.<provider_name>.api_key."
                 )
                 try:

ramalama/cli.py

@@ -713,7 +713,6 @@ def info_cli(args: DefaultArgsType) -> None:
             message = f"{name}={source} ({config_source})"
             print(message)
         return
-
     info: dict[str, Any] = {
         "Accelerator": get_accel(),
         "Config": load_file_config(),
@@ -950,6 +949,8 @@ def push_cli(args):
 
     if args.TARGET:
         shortnames = get_shortnames()
+        if source_model.type == "OCI":
+            raise ValueError(f"converting from an OCI based image {args.SOURCE} is not supported")
         target = shortnames.resolve(args.TARGET)
 
     target_model = New(target, args)
@@ -1642,12 +1643,7 @@ def _rm_model(models, args):
 
         try:
             m = New(model, args)
-            if m.remove(args):
-                continue
-            # Failed to remove and might be OCI so attempt to remove OCI
-            if args.ignore:
-                _rm_oci_model(model, args)
-                continue
+            m.remove(args)
         except (KeyError, subprocess.CalledProcessError) as e:
             for prefix in MODEL_TYPES:
                 if model.startswith(prefix + "://"):
@@ -1723,7 +1719,6 @@ def inspect_cli(args):
     if not args.MODEL:
         parser = get_parser()
         parser.error("inspect requires MODEL")
-
     args.pull = "never"
 
     model = New(args.MODEL, args)

ramalama/common.py

@@ -12,7 +12,9 @@
 import subprocess
 import sys
 from collections.abc import Callable, Sequence
+from dataclasses import dataclass
 from functools import lru_cache
+from pathlib import Path
 from typing import IO, TYPE_CHECKING, Any, Literal, Optional, Protocol, TypeAlias, TypedDict, cast, get_args
 
 import yaml
@@ -294,10 +296,11 @@ def genname():
     return "ramalama-" + "".join(random.choices(string.ascii_letters + string.digits, k=10))
 
 
-def engine_version(engine: SUPPORTED_ENGINES) -> str:
+@lru_cache
+def engine_version(engine: SUPPORTED_ENGINES | Path | str) -> SemVer:
     # Create manifest list for target with imageid
     cmd_args = [str(engine), "version", "--format", "{{ .Client.Version }}"]
-    return run_cmd(cmd_args, encoding="utf-8").stdout.strip()
+    return SemVer.parse(run_cmd(cmd_args, encoding="utf-8").stdout.strip())
 
 
 class CDI_DEVICE(TypedDict):
@@ -712,3 +715,36 @@ def __str__(self):
 
     def __repr__(self):
         return repr(self.entrypoint)
+
+
+SEMVER_RE = re.compile(
+    r"^(?P<major>0|[1-9]\d*)\."
+    r"(?P<minor>0|[1-9]\d*)\."
+    r"(?P<patch>0|[1-9]\d*)"
+    r"(?:-(?P<prerelease>"
+    r"(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*)"
+    r"(?:\.(?:0|[1-9]\d*|\d*[a-zA-Z-][0-9a-zA-Z-]*))*"
+    r"))?"
+    r"(?:\+(?P<buildmetadata>[0-9a-zA-Z-]+(?:\.[0-9a-zA-Z-]+)*))?$"
+)
+
+
+def parse_semver(s: str) -> SemVer:
+    m = SEMVER_RE.fullmatch(s)
+    if not m:
+        raise ValueError(f"Not a valid SemVer 2.0.0: {s!r}")
+    major = int(m.group("major"))
+    minor = int(m.group("minor"))
+    patch = int(m.group("patch"))
+    return SemVer(major, minor, patch)
+
+
+@dataclass(frozen=True, order=True)
+class SemVer:
+    major: int
+    minor: int
+    patch: int
+
+    @classmethod
+    def parse(cls, s: str) -> "SemVer":
+        return parse_semver(s)

ramalama/compose.py

@@ -2,7 +2,7 @@
 
 import os
 import shlex
-from typing import Optional, Tuple
+from typing import Optional
 
 from ramalama.common import RAG_DIR, get_accel_env_vars
 from ramalama.file import PlainFile
@@ -13,9 +13,9 @@ class Compose:
     def __init__(
         self,
         model_name: str,
-        model_paths: Tuple[str, str],
-        chat_template_paths: Optional[Tuple[str, str]],
-        mmproj_paths: Optional[Tuple[str, str]],
+        model_paths: tuple[str, str],
+        chat_template_paths: Optional[tuple[str, str]],
+        mmproj_paths: Optional[tuple[str, str]],
         args,
         exec_args,
     ):

ramalama/oci_tools.py

@@ -1,11 +1,13 @@
 import json
 import subprocess
+from dataclasses import dataclass
 from datetime import datetime
+from itertools import chain
 from typing import TypedDict
 
 import ramalama.annotations as annotations
 from ramalama.arg_types import EngineArgType
-from ramalama.common import engine_version, run_cmd
+from ramalama.common import SemVer, engine_version, run_cmd
 from ramalama.logger import logger
 
 ocilabeltype = "org.containers.type"
@@ -69,21 +71,33 @@ def list_artifacts(args: EngineArgType):
     except subprocess.CalledProcessError as e:
         logger.debug(e)
         return []
+    if output == "":
+        return []
 
-    artifacts = json.loads(f"[{output[:-1]}]")
-    models: list[ListModelResponse] = []
+    try:
+        artifacts = json.loads(f"[{output[:-1]}]")
+    except json.JSONDecodeError:
+        return []
+
+    models = []
     for artifact in artifacts:
         conman_args = [
             args.engine,
             "artifact",
             "inspect",
             artifact["ID"],
         ]
-        output = run_cmd(conman_args).stdout.decode("utf-8").strip()
+        try:
+            output = run_cmd(conman_args, ignore_stderr=True).stdout.decode("utf-8").strip()
+        except Exception:
+            continue
 
         if output == "":
             continue
-        inspect = json.loads(output)
+        try:
+            inspect = json.loads(output)
+        except json.JSONDecodeError:
+            continue
         if "Manifest" not in inspect:
             continue
         if "artifactType" not in inspect["Manifest"]:
@@ -103,8 +117,12 @@ def list_artifacts(args: EngineArgType):
 def engine_supports_manifest_attributes(engine) -> bool:
     if not engine or engine == "" or engine == "docker":
         return False
-    if engine == "podman" and engine_version(engine) < "5":
-        return False
+    if engine == "podman":
+        try:
+            if engine_version(engine) < SemVer(5, 0, 0):
+                return False
+        except Exception:
+            return False
     return True
 
 
@@ -227,12 +245,67 @@ def list_images(args: EngineArgType) -> list[ListModelResponse]:
 
 
 def list_models(args: EngineArgType) -> list[ListModelResponse]:
-    conman = args.engine
-    if conman is None:
+    if args.engine is None:
         return []
 
-    models = list_images(args)
-    models.extend(list_manifests(args))
-    models.extend(list_artifacts(args))
+    model_gen = chain(list_images(args), list_manifests(args), list_artifacts(args))
 
+    seen: set[str] = set()
+    models: list[ListModelResponse] = []
+    for m in model_gen:
+        if (name := m["name"]) in seen:
+            continue
+        seen.add(name)
+        models.append(m)
     return models
+
+
+@dataclass(frozen=True)
+class OciRef:
+    registry: str
+    repository: str
+    specifier: str  # Either the digest or the tag
+    tag: str | None = None
+    digest: str | None = None
+
+    def __str__(self) -> str:
+        if self.digest:
+            return f"{self.registry}/{self.repository}@{self.digest}"
+        return f"{self.registry}/{self.repository}:{self.tag or self.specifier}"
+
+    @staticmethod
+    def from_ref_string(ref: str) -> "OciRef":
+        return split_oci_reference(ref)
+
+
+def split_oci_reference(ref: str, default_registry: str = "docker.io") -> OciRef:
+    ref = ref.strip()
+
+    name, digest = ref.split("@", 1) if "@" in ref else (ref, None)
+
+    slash = name.rfind("/")
+    colon = name.rfind(":")
+    if colon > slash:
+        name, tag = name[:colon], name[colon + 1 :]
+    else:
+        tag = None
+
+    parts = name.split("/", 1)
+    if len(parts) == 1:
+        registry = default_registry
+        repository = parts[0]
+    else:
+        first, rest = parts[0], parts[1]
+        if first == "localhost" or "." in first or ":" in first:
+            registry = first
+            repository = rest
+        else:
+            registry = default_registry
+            repository = name  # keep full path
+
+    specifier = digest or tag
+    if specifier is None:
+        tag = "latest"
+        specifier = tag
+
+    return OciRef(registry=registry, repository=repository, tag=tag, digest=digest, specifier=specifier)

ramalama/rag.py

@@ -15,7 +15,7 @@
 from ramalama.engine import BuildEngine, Engine, is_healthy, stop_container, wait_for_healthy
 from ramalama.path_utils import get_container_mount_path
 from ramalama.transports.base import Transport
-from ramalama.transports.oci import OCI
+from ramalama.transports.oci.oci import OCI
 
 INPUT_DIR = "/docs"
 
@@ -197,7 +197,6 @@ def serve(self, args: RagArgsType, cmd: list[str]):
                 stop_container(args.model_args, args.model_args.name, remove=True)
 
     def run(self, args: RagArgsType, cmd: list[str]):
-
         args.model_args.name = self.imodel.get_container_name(args.model_args)
         process = self.imodel.serve_nonblocking(args.model_args, self.model_cmd)
         rag_process = self.serve_nonblocking(args, cmd)