Publish expanded LumynaX product platform package

PYPI_RELEASE.md

@@ -1,7 +1,7 @@
 # PyPI Release: LumynaX MaramaRoute
 
 Package: `lumynax-marama-route`
-Version: `0.7.11`
+Version: `0.7.12`
 
 ## Local Build
 

marama_route/cli.py

@@ -89,11 +89,22 @@ MODEL_PICKER_COMMANDS = {
     "/search",
     "/switch",
     "/use",
+    "/vllm",
+    "/nim",
+    "/nem",
+    "/nemo",
     "/recommended",
     "/recommend",
     "?",
 }
 MODEL_PICKER_PAGE_SIZE = 12
+_DEPLOYMENT_USABLE_STATUSES = {"supported", "candidate", "experimental"}
+_DEPLOYMENT_PATHWAY_STATUSES = _DEPLOYMENT_USABLE_STATUSES | {"convert_required"}
+_DEPLOYMENT_TARGET_LABELS = {
+    "vllm": "vLLM",
+    "nvidia_nim": "NVIDIA NIM",
+    "nvidia_nemo": "NVIDIA NeMo/NEM",
+}
 
 
 class _ExitConversation(Exception):
@@ -961,6 +972,9 @@ def _conversation(args: argparse.Namespace, models: tuple[Any, ...]) -> int:
                 "/hardware",
                 "/recommended",
                 "/categories",
+                "/vllm",
+                "/nim",
+                "/nemo",
                 "/search <text>",
                 "/switch <text>",
                 "/pull [text]",
@@ -1062,6 +1076,17 @@ def _conversation(args: argparse.Namespace, models: tuple[Any, ...]) -> int:
                 return 0
             session = None
             continue
+        if command in {"/vllm", "/nim", "/nemo", "/nem"}:
+            try:
+                model = _prompt_for_model(
+                    models,
+                    args.cache_dir,
+                    compatibility_target=_deployment_target_from_command(command),
+                )
+            except _ExitConversation:
+                return 0
+            session = None
+            continue
         if command == "/catalog":
             choices = _conversation_choices(models, include_all=True)
             _print_model_picker_header(
@@ -1070,6 +1095,7 @@ def _conversation(args: argparse.Namespace, models: tuple[Any, ...]) -> int:
                 query="",
                 include_all=True,
                 hardware_only=False,
+                compatibility_target="",
                 offset=0,
                 shown=min(len(choices), MODEL_PICKER_PAGE_SIZE),
                 hardware=None,
@@ -1271,13 +1297,17 @@ def _conversation_choices(
     *,
     search: str = "",
     include_all: bool = False,
+    compatibility_target: str = "",
     limit: int | None = None,
 ) -> list[Any]:
     filtered: list[Any] = []
     query = search.strip().lower()
     for model in models:
         haystack = " ".join((model.model_id, model.repo_id, model.family, " ".join(model.tags))).lower()
-        if not include_all and not _is_chat_runnable(model):
+        if compatibility_target:
+            if not _matches_deployment_target(model, compatibility_target):
+                continue
+        elif not include_all and not _is_chat_runnable(model):
             continue
         if query and query not in haystack:
             continue
@@ -1326,16 +1356,17 @@ def _prompt_for_model(
     initial_query: str = "",
     include_all: bool = False,
     hardware_only: bool = False,
+    compatibility_target: str = "",
     show_menu: bool = True,
 ) -> Any:
     query = initial_query
-    menu_open = show_menu and not initial_query and not include_all and not hardware_only
+    menu_open = show_menu and not initial_query and not include_all and not hardware_only and not compatibility_target
     offset = 0
     last_hardware: dict[str, Any] | None = None
     while True:
         if menu_open:
             _print_model_picker_menu(models, cache_dir)
-            raw = input("Choose option 1-6, type search text, or /exit: ").strip()
+            raw = input("Choose option 1-9, type search text, or /exit: ").strip()
             if not raw:
                 raw = "1"
             lowered = raw.lower()
@@ -1347,6 +1378,7 @@ def _prompt_for_model(
             if raw == "1" or lowered in {"/hardware", "/recommended-hardware"}:
                 hardware_only = True
                 include_all = False
+                compatibility_target = ""
                 query = ""
                 offset = 0
                 menu_open = False
@@ -1354,6 +1386,7 @@ def _prompt_for_model(
             if raw == "2" or lowered in {"/recommended", "/recommend", "/best"}:
                 hardware_only = False
                 include_all = False
+                compatibility_target = ""
                 query = ""
                 offset = 0
                 menu_open = False
@@ -1362,12 +1395,14 @@ def _prompt_for_model(
                 query = input("Search text: ").strip()
                 include_all = True
                 hardware_only = False
+                compatibility_target = ""
                 offset = 0
                 menu_open = False
                 continue
             if raw == "4" or lowered in {"/models", "/switch", "/runnable"}:
                 hardware_only = False
                 include_all = False
+                compatibility_target = ""
                 query = ""
                 offset = 0
                 menu_open = False
@@ -1375,6 +1410,7 @@ def _prompt_for_model(
             if raw == "5" or lowered in {"/all", "/catalog"}:
                 hardware_only = False
                 include_all = True
+                compatibility_target = ""
                 query = ""
                 offset = 0
                 menu_open = False
@@ -1382,22 +1418,56 @@ def _prompt_for_model(
             if raw == "6" or lowered == "/local":
                 _print_local_models(cache_dir)
                 continue
+            if raw == "7" or lowered == "/vllm":
+                hardware_only = False
+                include_all = True
+                compatibility_target = "vllm"
+                query = ""
+                offset = 0
+                menu_open = False
+                continue
+            if raw == "8" or lowered == "/nim":
+                hardware_only = False
+                include_all = True
+                compatibility_target = "nvidia_nim"
+                query = ""
+                offset = 0
+                menu_open = False
+                continue
+            if raw == "9" or lowered in {"/nemo", "/nem"}:
+                hardware_only = False
+                include_all = True
+                compatibility_target = "nvidia_nemo"
+                query = ""
+                offset = 0
+                menu_open = False
+                continue
             handled = _handle_model_picker_command(raw, models)
             if handled is not None:
                 query = handled["query"]
                 include_all = handled["include_all"]
                 hardware_only = handled.get("hardware_only", False)
+                compatibility_target = handled.get("compatibility_target", "")
                 menu_open = handled.get("menu", False)
                 offset = 0
             else:
                 query = raw
                 include_all = True
                 hardware_only = False
+                compatibility_target = ""
                 offset = 0
                 menu_open = False
             continue
 
-        if hardware_only:
+        if compatibility_target:
+            choices = _conversation_choices(
+                models,
+                search=query,
+                include_all=True,
+                compatibility_target=compatibility_target,
+            )
+            last_hardware = None
+        elif hardware_only:
             choices, last_hardware = _hardware_choices(models, cache_dir, search=query, include_all=include_all)
         else:
             choices = _conversation_choices(models, search=query, include_all=include_all)
@@ -1409,6 +1479,7 @@ def _prompt_for_model(
                 query = handled["query"]
                 include_all = handled["include_all"]
                 hardware_only = handled.get("hardware_only", False)
+                compatibility_target = handled.get("compatibility_target", "")
                 menu_open = handled.get("menu", False)
                 offset = 0
             else:
@@ -1423,6 +1494,7 @@ def _prompt_for_model(
             query=query,
             include_all=include_all,
             hardware_only=hardware_only,
+            compatibility_target=compatibility_target,
             offset=offset,
             shown=len(page),
             hardware=last_hardware,
@@ -1457,6 +1529,7 @@ def _prompt_for_model(
             query = handled["query"]
             include_all = handled["include_all"]
             hardware_only = handled.get("hardware_only", False)
+            compatibility_target = handled.get("compatibility_target", "")
             menu_open = handled.get("menu", False)
             offset = 0
             continue
@@ -1482,42 +1555,49 @@ def _handle_model_picker_command(raw: str, models: tuple[Any, ...]) -> dict[str,
     if command in CHAT_COMMANDS:
         raise _ExitConversation
     if command in {"/models", "/switch", "/runnable"}:
-        return {"query": "", "include_all": False, "hardware_only": False}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": ""}
     if command in {"/hardware", "/recommended-hardware"}:
-        return {"query": "", "include_all": False, "hardware_only": True}
+        return {"query": "", "include_all": False, "hardware_only": True, "compatibility_target": ""}
     if command in {"/recommended", "/recommend", "/best"}:
-        return {"query": "", "include_all": False, "hardware_only": False}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": ""}
     if command == "/menu":
-        return {"query": "", "include_all": False, "hardware_only": False, "menu": True}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": "", "menu": True}
     if command == "/all":
-        return {"query": "", "include_all": True, "hardware_only": False}
+        return {"query": "", "include_all": True, "hardware_only": False, "compatibility_target": ""}
     if command == "/catalog":
-        return {"query": "", "include_all": True, "hardware_only": False}
+        return {"query": "", "include_all": True, "hardware_only": False, "compatibility_target": ""}
+    if command in {"/vllm", "/nim", "/nemo", "/nem"}:
+        return {
+            "query": "",
+            "include_all": True,
+            "hardware_only": False,
+            "compatibility_target": _deployment_target_from_command(command),
+        }
     if command.startswith("/search "):
         query = command.removeprefix("/search ").strip()
-        return {"query": query, "include_all": True, "hardware_only": False}
+        return {"query": query, "include_all": True, "hardware_only": False, "compatibility_target": ""}
     if command.startswith("/use "):
         query = command.removeprefix("/use ").strip()
-        return {"query": query, "include_all": False, "hardware_only": False}
+        return {"query": query, "include_all": False, "hardware_only": False, "compatibility_target": ""}
     if command.startswith("/family "):
         family = command.removeprefix("/family ").strip()
-        return {"query": family, "include_all": True, "hardware_only": False}
+        return {"query": family, "include_all": True, "hardware_only": False, "compatibility_target": ""}
     if command in {"/families", "/categories"}:
         _print_model_categories(models)
-        return {"query": "", "include_all": True, "hardware_only": False}
+        return {"query": "", "include_all": True, "hardware_only": False, "compatibility_target": ""}
     if command == "/local":
         print("Local models are shown after a model is selected. Use `MaramaRoute local` for full JSON.")
-        return {"query": "", "include_all": False, "hardware_only": False, "menu": True}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": "", "menu": True}
     if command in {"/help", "/h", "?"}:
         _print_model_picker_help()
-        return {"query": "", "include_all": False, "hardware_only": False, "menu": True}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": "", "menu": True}
     if command == "/pull":
         print("Choose a model first; then /pull downloads the selected model.")
-        return {"query": "", "include_all": False, "hardware_only": False}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": ""}
     if command.startswith("/"):
         known = ", ".join(sorted(CHAT_COMMANDS | MODEL_PICKER_COMMANDS))
         print(f"Unknown command {raw!r}. Known commands: {known}.")
-        return {"query": "", "include_all": False, "hardware_only": False, "menu": True}
+        return {"query": "", "include_all": False, "hardware_only": False, "compatibility_target": "", "menu": True}
     return None
 
 
@@ -1538,6 +1618,86 @@ def _offline_capability_label(model: Any) -> str:
     return "offline-task"
 
 
+def _deployment_target_from_command(command: str) -> str:
+    normalized = command.strip().lower().lstrip("/").replace("-", "_")
+    if normalized == "vllm":
+        return "vllm"
+    if normalized == "nim":
+        return "nvidia_nim"
+    if normalized in {"nem", "nemo"}:
+        return "nvidia_nemo"
+    raise ValueError(f"Unknown deployment target command: {command}")
+
+
+def _deployment_statuses_for_target(target: str) -> set[str]:
+    if target == "nvidia_nemo":
+        return set(_DEPLOYMENT_PATHWAY_STATUSES)
+    return set(_DEPLOYMENT_USABLE_STATUSES)
+
+
+def _matches_deployment_target(model: Any, target: str) -> bool:
+    compatibility = model_runtime_compatibility(model)
+    entry = compatibility.get(target, {})
+    status = str(entry.get("status") or "").lower() if isinstance(entry, dict) else ""
+    return status in _deployment_statuses_for_target(target)
+
+
+def _deployment_summary(models: tuple[Any, ...]) -> dict[str, Any]:
+    summary: dict[str, Any] = {}
+    for target, label in _DEPLOYMENT_TARGET_LABELS.items():
+        statuses: dict[str, int] = {}
+        usable = 0
+        pathway = 0
+        for model in models:
+            entry = model_runtime_compatibility(model).get(target, {})
+            status = str(entry.get("status") or "unknown").lower() if isinstance(entry, dict) else "unknown"
+            statuses[status] = statuses.get(status, 0) + 1
+            if status in _DEPLOYMENT_USABLE_STATUSES:
+                usable += 1
+            if status in _DEPLOYMENT_PATHWAY_STATUSES:
+                pathway += 1
+        summary[target] = {
+            "label": label,
+            "usable": usable,
+            "pathway": pathway,
+            "statuses": dict(_top_counts(statuses, limit=max(1, len(statuses)))),
+            "browse_command": (
+                f"MaramaRoute compat --target {_deployment_cli_target(target)} "
+                f"--status {_deployment_cli_status(target)}"
+            ),
+            "picker_command": f"/{_deployment_cli_target(target)}",
+        }
+    return summary
+
+
+def _deployment_cli_target(target: str) -> str:
+    if target == "nvidia_nim":
+        return "nim"
+    if target == "nvidia_nemo":
+        return "nemo"
+    return target
+
+
+def _deployment_cli_status(target: str) -> str:
+    if target == "nvidia_nemo":
+        return "pathway"
+    return "usable"
+
+
+def _format_deployment_counts(deployment: dict[str, Any]) -> list[str]:
+    lines: list[str] = []
+    for target in ("vllm", "nvidia_nim", "nvidia_nemo"):
+        entry = deployment.get(target, {})
+        label = str(entry.get("label") or target)
+        statuses = entry.get("statuses") if isinstance(entry.get("statuses"), dict) else {}
+        status_text = _format_counts(statuses)
+        if target == "nvidia_nemo":
+            lines.append(f"{label}: direct {entry.get('usable', 0)}, pathway {entry.get('pathway', 0)} ({status_text})")
+        else:
+            lines.append(f"{label}: usable {entry.get('usable', 0)} ({status_text})")
+    return lines
+
+
 def _is_transformers_text_generation(model: Any) -> bool:
     runtime = model.runtime.lower()
     if "transformers" not in runtime or "multimodal" in runtime:
@@ -1608,13 +1768,16 @@ def _print_model_picker_header(
     query: str,
     include_all: bool,
     hardware_only: bool,
+    compatibility_target: str,
     offset: int,
     shown: int,
     hardware: dict[str, Any] | None,
 ) -> None:
     total = len(models)
     runnable = sum(1 for model in models if _is_chat_runnable(model))
-    if hardware_only:
+    if compatibility_target:
+        scope = f"{_DEPLOYMENT_TARGET_LABELS.get(compatibility_target, compatibility_target)} deployment-path models"
+    elif hardware_only:
         scope = "hardware-suitable local GGUF chat models"
     else:
         scope = "all AbteeXAILab HF registry entries" if include_all else "recommended local GGUF chat models"
@@ -1628,10 +1791,12 @@ def _print_model_picker_header(
         print("Use /next for more results.")
     if offset > 0:
         print("Use /prev for previous results.")
-    if not include_all and not hardware_only:
-        print("Use /hardware for machine-suitable models, /all for the full registry, or /help for commands.")
+    if compatibility_target:
+        print("Use /vllm, /nim, /nemo, /all, or /models to change category.")
+    elif not include_all and not hardware_only:
+        print("Use /hardware for machine-suitable models, /all, /vllm, /nim, /nemo, or /help.")
     else:
-        print("Use /models to return to recommended chat models, /family <name> to filter, or /help for commands.")
+        print("Use /models to return to recommended chat models, /family <name>, /vllm, /nim, /nemo, or /help.")
 
 
 def _print_model_picker_menu(models: tuple[Any, ...], cache_dir: Path | None) -> None:
@@ -1647,6 +1812,9 @@ def _print_model_picker_menu(models: tuple[Any, ...], cache_dir: Path | None) ->
     print(" 4. Browse local GGUF chat models")
     print(" 5. Browse full registry")
     print(" 6. Show pulled local models")
+    print(" 7. Browse vLLM deployment-path models")
+    print(" 8. Browse NVIDIA NIM deployment-path models")
+    print(" 9. Browse NVIDIA NeMo/NEM deployment-path models")
 
 
 def _print_model_picker_help() -> None:
@@ -1658,9 +1826,13 @@ def _print_model_picker_help() -> None:
                 "  /recommended       show recommended local chat models",
                 "  /models            browse local GGUF chat-capable models",
                 "  /all               browse every bundled AbteeXAILab registry entry",
+                "  /vllm              browse vLLM supported/candidate/experimental models",
+                "  /nim               browse NVIDIA NIM supported/candidate/experimental models",
+                "  /nemo              browse NVIDIA NeMo candidates and conversion paths",
+                "  /nem               alias for /nemo",
                 "  /search <text>     search model id, repo, family, or tags",
                 "  /family <name>     filter by family or tag",
-                "  /categories        show category/family/runtime/tag counts",
+                "  /categories        show category/family/runtime/tag/deployment counts",
                 "  /families          alias for /categories",
                 "  /next              next page of results",
                 "  /prev              previous page of results",
@@ -1674,7 +1846,7 @@ def _print_model_picker_help() -> None:
 
 def _print_model_categories(models: tuple[Any, ...]) -> None:
     _print_category_summary(_category_summary(models, limit=18))
-    print("Type a family/tag/search term, or use /family qwen, /family vision, /all, /runnable.")
+    print("Type a family/tag/search term, or use /vllm, /nim, /nemo, /family qwen, /all, /runnable.")
 
 
 def _category_summary(models: tuple[Any, ...], *, limit: int) -> dict[str, Any]:
@@ -1702,6 +1874,7 @@ def _category_summary(models: tuple[Any, ...], *, limit: int) -> dict[str, Any]:
         "tags": dict(_top_counts(tags, limit=safe_limit)),
         "modalities": dict(_top_counts(modalities, limit=safe_limit)),
         "capabilities": dict(_top_counts(capabilities, limit=safe_limit)),
+        "deployment_compatibility": _deployment_summary(models),
         "commands": {
             "browse_recommended": "MaramaRoute chat",
             "browse_all": "MaramaRoute catalog",
@@ -1709,6 +1882,9 @@ def _category_summary(models: tuple[Any, ...], *, limit: int) -> dict[str, Any]:
             "filter_family": "MaramaRoute catalog --family qwen",
             "pull_family": "MaramaRoute pull --family qwen --limit 3 --dry-run",
             "compatibility": "MaramaRoute compat",
+            "browse_vllm": "MaramaRoute compat --target vllm --status usable",
+            "browse_nim": "MaramaRoute compat --target nim --status usable",
+            "browse_nem_nemo": "MaramaRoute compat --target nemo --status pathway",
         },
     }
 
@@ -1722,7 +1898,13 @@ def _print_category_summary(summary: dict[str, Any]) -> None:
     print(f"Runtimes: {_format_counts(summary['runtimes'])}")
     print(f"Modalities: {_format_counts(summary['modalities'])}")
     print(f"Tags: {_format_counts(summary['tags'])}")
-    print("Next: MaramaRoute chat  |  MaramaRoute catalog --family qwen  |  MaramaRoute pull --family qwen --dry-run")
+    print("Deployment compatibility:")
+    for line in _format_deployment_counts(summary.get("deployment_compatibility", {})):
+        print(f"  {line}")
+    print(
+        "Next: MaramaRoute chat  |  MaramaRoute compat --target vllm --status usable  |  "
+        "MaramaRoute compat --target nemo --status pathway",
+    )
 
 
 def _format_counts(counts: dict[str, int]) -> str:
@@ -1988,7 +2170,11 @@ def build_parser() -> argparse.ArgumentParser:
     compat.add_argument("model", nargs="?", help="Optional model id, repo id, or unique search fragment.")
     compat.add_argument("--registry", type=Path, default=None, help="MaramaRoute model registry JSON.")
     compat.add_argument("--target", default="", help="Runtime target: vllm, nim, nem, nemo, llama-cpp, or all.")
-    compat.add_argument("--status", default="", help="Filter by status such as candidate, experimental, unsupported.")
+    compat.add_argument(
+        "--status",
+        default="",
+        help="Filter by status such as candidate, experimental, unsupported, usable, or pathway.",
+    )
     compat.add_argument("--limit", type=int, default=0, help="Maximum rows; 0 means all models.")
     compat.add_argument("--format", choices=["json", "table"], default="table")
     compat.set_defaults(handler=_compat)
@@ -1996,7 +2182,7 @@ def build_parser() -> argparse.ArgumentParser:
     categories = subparsers.add_parser(
         "categories",
         aliases=["families"],
-        help="Show model families, runtimes, tags, modalities, and local capability categories.",
+        help="Show model families, runtimes, tags, modalities, local capability, and deployment categories.",
     )
     categories.add_argument("--registry", type=Path, default=None, help="MaramaRoute model registry JSON.")
     categories.add_argument("--limit", type=int, default=18)

marama_route/compat.py

@@ -6,6 +6,15 @@ from typing import Any
 from .registry import ModelEndpoint
 
 COMPATIBILITY_TARGETS = ("llama_cpp", "vllm", "nvidia_nim", "nvidia_nemo")
+_USABLE_STATUSES = {"supported", "candidate", "experimental"}
+_PATHWAY_STATUSES = _USABLE_STATUSES | {"convert_required"}
+_STATUS_GROUPS = {
+    "usable": _USABLE_STATUSES,
+    "compatible": _USABLE_STATUSES,
+    "pathway": _PATHWAY_STATUSES,
+    "deployment_path": _PATHWAY_STATUSES,
+    "deployment-path": _PATHWAY_STATUSES,
+}
 
 _TASK_TAGS = {
     "asr",
@@ -104,12 +113,12 @@ def build_compatibility_matrix(
             summary[name][str(compatibility[name]["status"])] += 1
         if target:
             entry = compatibility[target]
-            if wanted_status and str(entry["status"]).lower() != wanted_status:
+            if wanted_status and not _status_matches(str(entry["status"]), wanted_status):
                 continue
             rows.append(_compatibility_row(model, {target: entry}))
         else:
             if wanted_status and not any(
-                str(entry["status"]).lower() == wanted_status for entry in compatibility.values()
+                _status_matches(str(entry["status"]), wanted_status) for entry in compatibility.values()
             ):
                 continue
             rows.append(_compatibility_row(model, compatibility))
@@ -294,6 +303,15 @@ def _entry(status: str, reason: str, *, commands: list[str] | None = None) -> di
     return payload
 
 
+def _status_matches(actual: str, wanted: str) -> bool:
+    actual_value = actual.strip().lower()
+    wanted_value = wanted.strip().lower()
+    status_group = _STATUS_GROUPS.get(wanted_value)
+    if status_group is not None:
+        return actual_value in status_group
+    return actual_value == wanted_value
+
+
 def _normalize_target(target: str) -> str:
     value = target.strip().lower().replace("-", "_")
     aliases = {
@@ -303,7 +321,7 @@ def _normalize_target(target: str) -> str:
         "llamacpp": "llama_cpp",
         "nim": "nvidia_nim",
         "nvidia_nim": "nvidia_nim",
-        "nem": "nvidia_nim",
+        "nem": "nvidia_nemo",
         "nemo": "nvidia_nemo",
         "nvidia_nemo": "nvidia_nemo",
     }

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "lumynax-marama-route"
-version = "0.7.11"
+version = "0.7.12"
 description = "LumynaX MaramaRoute: conversational AbteeX AI Labs model CLI, downloader, and local runtime for LumynaX releases."
 readme = "README.md"
 requires-python = ">=3.11"