ltcai 8.1.0 → 8.2.0

package/latticeai/services/model_loading.py

@@ -0,0 +1,482 @@
+"""Model loading, prepare and stream logic extracted from model_runtime for server decomp.
+
+This moves the large prepare_and_load_model and stream functions out of the monolith.
+Re-exports will be added in model_runtime for compat.
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import queue
+import subprocess
+import threading
+from pathlib import Path
+from typing import AsyncIterator, Dict, Optional
+
+from fastapi import HTTPException, Request
+
+# Late imports to avoid circulars during extraction
+def _get_model_runtime_deps():
+    from .model_runtime import (
+        _download_allowed,
+        _download_block,
+        _engine_install_block,
+        _friendly_model_runtime_error,
+        _ModelResolution,
+        _model_runtime_compatibility,
+        _smoke_test_loaded_model,
+        download_hf_model,
+        engine_installed,
+        ensure_engine_ready,
+        ensure_llamacpp_server,
+        ensure_lmstudio_model,
+        ensure_ollama_server,
+        ensure_vllm_server,
+        get_current_user,
+        get_lmstudio_models,
+        get_ollama_pulled_models,
+        get_user_api_key,
+        hf_model_dir,
+        hf_model_ready,
+        local_binary,
+        MODEL_ENGINE_ALIASES,
+        model_download_progress_payload,
+        normalize_local_model_request,
+        parse_model_ref,
+        pull_ollama_model_with_progress,
+        router,
+    )
+    return {
+        "_download_allowed": _download_allowed,
+        "_download_block": _download_block,
+        "_engine_install_block": _engine_install_block,
+        "_friendly_model_runtime_error": _friendly_model_runtime_error,
+        "_ModelResolution": _ModelResolution,
+        "_model_runtime_compatibility": _model_runtime_compatibility,
+        "_smoke_test_loaded_model": _smoke_test_loaded_model,
+        "download_hf_model": download_hf_model,
+        "engine_installed": engine_installed,
+        "ensure_engine_ready": ensure_engine_ready,
+        "ensure_llamacpp_server": ensure_llamacpp_server,
+        "ensure_lmstudio_model": ensure_lmstudio_model,
+        "ensure_ollama_server": ensure_ollama_server,
+        "ensure_vllm_server": ensure_vllm_server,
+        "get_current_user": get_current_user,
+        "get_lmstudio_models": get_lmstudio_models,
+        "get_ollama_pulled_models": get_ollama_pulled_models,
+        "get_user_api_key": get_user_api_key,
+        "hf_model_dir": hf_model_dir,
+        "hf_model_ready": hf_model_ready,
+        "local_binary": local_binary,
+        "MODEL_ENGINE_ALIASES": MODEL_ENGINE_ALIASES,
+        "model_download_progress_payload": model_download_progress_payload,
+        "normalize_local_model_request": normalize_local_model_request,
+        "parse_model_ref": parse_model_ref,
+        "pull_ollama_model_with_progress": pull_ollama_model_with_progress,
+        "router": router,
+    }
+
+
+async def prepare_and_load_model(
+    model_id: str,
+    request: Request,
+    engine: Optional[str] = None,
+    user_email: Optional[str] = None,
+    adapter_path: Optional[str] = None,
+    draft_model_id: Optional[str] = None,
+    allow_download: bool = False,
+) -> Dict[str, object]:
+    deps = _get_model_runtime_deps()
+    model_id = deps["normalize_local_model_request"](model_id, engine)
+    if not model_id:
+        raise HTTPException(status_code=400, detail="모델 식별자가 비어 있습니다.")
+
+    resolution = deps["_ModelResolution"].from_request(
+        model_id,
+        engine=engine,
+        user_email=user_email or deps["get_current_user"](request),
+        engine_aliases=deps["MODEL_ENGINE_ALIASES"],
+    )
+
+    parsed_provider, parsed_model = deps["parse_model_ref"](model_id)
+    if parsed_provider == "mlx":
+        parsed_provider = "local_mlx"
+    compatibility = deps["_model_runtime_compatibility"](parsed_model, engine=parsed_provider)
+    if compatibility.get("supported") is False:
+        raise HTTPException(status_code=400, detail=compatibility)
+
+    local_engines = {"local_mlx", "ollama", "vllm", "lmstudio", "llamacpp"}
+    install_result: Dict[str, object] = {}
+    download_result: Optional[Dict[str, object]] = None
+
+    if parsed_provider in local_engines:
+        if not deps["engine_installed"](parsed_provider) and not deps["_download_allowed"](allow_download):
+            deps["_engine_install_block"](parsed_provider)
+        install_result = deps["ensure_engine_ready"](parsed_provider)
+
+    if parsed_provider == "local_mlx":
+        explicit_path = Path(parsed_model).expanduser()
+        if not explicit_path.exists() and not deps["hf_model_ready"](parsed_model, "local_mlx"):
+            if not deps["_download_allowed"](allow_download):
+                deps["_download_block"](parsed_provider, parsed_model)
+            download_result = deps["download_hf_model"](parsed_model, "local_mlx")
+    elif parsed_provider == "ollama":
+        deps["ensure_ollama_server"]()
+        ollama = deps["local_binary"]("ollama")
+        if not ollama:
+            raise HTTPException(status_code=400, detail="Ollama가 설치되지 않았습니다.")
+        if parsed_model not in deps["get_ollama_pulled_models"]():
+            if not deps["_download_allowed"](allow_download):
+                deps["_download_block"](parsed_provider, parsed_model)
+            completed = subprocess.run(
+                [ollama, "pull", parsed_model],
+                capture_output=True,
+                text=True,
+                timeout=900,
+                check=False,
+            )
+            if completed.returncode != 0:
+                raise HTTPException(status_code=500, detail=completed.stderr[-2000:] or "Ollama 모델 다운로드 실패")
+            download_result = {"provider": "ollama", "model": parsed_model, "returncode": completed.returncode}
+    elif parsed_provider == "vllm":
+        if not deps["hf_model_ready"](parsed_model, "vllm") and not deps["_download_allowed"](allow_download):
+            deps["_download_block"](parsed_provider, parsed_model)
+        deps["ensure_vllm_server"](parsed_model)
+        download_result = {"provider": "vllm", "model": parsed_model, "server_ready": True}
+    elif parsed_provider == "llamacpp":
+        if not deps["hf_model_ready"](parsed_model, "llamacpp") and not deps["_download_allowed"](allow_download):
+            deps["_download_block"](parsed_provider, parsed_model)
+        deps["ensure_llamacpp_server"](parsed_model)
+        download_result = {"provider": "llamacpp", "model": parsed_model, "server_ready": True}
+    elif parsed_provider == "lmstudio":
+        downloaded = {
+            str(item.get("key") or "").strip()
+            for item in deps["get_lmstudio_models"]()
+            if isinstance(item, dict)
+        }
+        if parsed_model not in downloaded and not deps["_download_allowed"](allow_download):
+            deps["_download_block"](parsed_provider, parsed_model)
+        ensured = deps["ensure_lmstudio_model"](parsed_model)
+        resolved_model = str(
+            ensured.get("instance_id")
+            or ensured.get("resolved_model")
+            or parsed_model
+        ).strip()
+        parsed_model = resolved_model
+        model_id = f"lmstudio:{resolved_model}"
+        download_result = ensured
+
+    effective_email = (user_email or deps["get_current_user"](request) or "").strip()
+    user_api_key = deps["get_user_api_key"](effective_email, parsed_provider) if parsed_provider != "local_mlx" else None
+    msg = await deps["router"].load_model(
+        model_id,
+        adapter_path,
+        draft_model_id=draft_model_id,
+        api_key_override=user_api_key,
+        owner=effective_email or None,
+    )
+    resolution.update_after_load(actual_current=deps["router"].current_model_id)
+    smoke_result: Dict[str, object] = {}
+    ready_to_chat = True
+    compat_status = "ok"
+    try:
+        smoke_result = await deps["_smoke_test_loaded_model"](resolution, api_key_override=user_api_key)
+        ready_to_chat = bool(smoke_result.get("ok"))
+        compat_status = str(smoke_result.get("status") or ("ok" if ready_to_chat else "degraded"))
+    except Exception as exc:
+        logging.warning("smoke test failed for %s: %s", resolution.load_id, exc)
+        compat_status = "unknown"
+    return {
+        "status": "ok",
+        "message": msg,
+        "model": model_id,
+        "current": deps["router"].current_model_id,
+        "engine": parsed_provider,
+        "installed_now": bool(install_result.get("installed_now")),
+        "download": download_result,
+        "resolution": resolution.to_dict(),
+        "downloaded": bool(download_result and not (isinstance(download_result, dict) and download_result.get("cached"))),
+        "loaded": True,
+        "ready_to_chat": ready_to_chat,
+        "compatibility_status": compat_status,
+        "smoke_test": smoke_result,
+    }
+
+
+def sse_event(event: str, data: Dict[str, object]) -> str:
+    return f"event: {event}\ndata: {json.dumps(data, ensure_ascii=False)}\n\n"
+
+
+async def prepare_and_load_model_stream(
+    model_id: str,
+    request: Request,
+    engine: Optional[str] = None,
+    user_email: Optional[str] = None,
+    allow_download: bool = False,
+) -> AsyncIterator[str]:
+    deps = _get_model_runtime_deps()
+    model_id = deps["normalize_local_model_request"](model_id, engine)
+    if not model_id:
+        raise HTTPException(status_code=400, detail="모델 식별자가 비어 있습니다.")
+
+    parsed_provider, parsed_model = deps["parse_model_ref"](model_id)
+    if parsed_provider == "mlx":
+        parsed_provider = "local_mlx"
+    compatibility = deps["_model_runtime_compatibility"](parsed_model, engine=parsed_provider)
+    if compatibility.get("supported") is False:
+        raise HTTPException(status_code=400, detail=compatibility)
+
+    work_queue: "queue.Queue[Dict[str, object]]" = queue.Queue()
+    work_result: Dict[str, object] = {}
+
+    def emit_progress(payload: Dict[str, object]) -> None:
+        work_queue.put({"kind": "progress", "data": payload})
+
+    def blocking_prepare() -> None:
+        try:
+            local_engines = {"local_mlx", "ollama", "vllm", "lmstudio", "llamacpp"}
+            install_result: Dict[str, object] = {}
+            download_result: Optional[Dict[str, object]] = None
+            prepared_model_id = model_id
+            prepared_model_name = parsed_model
+
+            if parsed_provider in local_engines:
+                emit_progress(deps["model_download_progress_payload"](
+                    "engine",
+                    "실행 엔진을 확인하는 중입니다.",
+                    percent=2,
+                    indeterminate=True,
+                ))
+                if not deps["engine_installed"](parsed_provider) and not deps["_download_allowed"](allow_download):
+                    deps["_engine_install_block"](parsed_provider)
+                install_result = deps["ensure_engine_ready"](parsed_provider)
+                emit_progress(deps["model_download_progress_payload"](
+                    "engine",
+                    "실행 엔진 준비가 완료되었습니다.",
+                    percent=10,
+                    indeterminate=False,
+                ))
+
+            if parsed_provider == "local_mlx":
+                explicit_path = Path(parsed_model).expanduser()
+                if explicit_path.exists():
+                    download_result = {"model": parsed_model, "path": str(explicit_path), "cached": True}
+                    emit_progress(deps["model_download_progress_payload"](
+                        "download",
+                        "로컬 모델 경로를 확인했습니다.",
+                        percent=100,
+                        detail=str(explicit_path),
+                        eta_seconds=0,
+                    ))
+                elif not deps["hf_model_ready"](parsed_model, "local_mlx"):
+                    if not deps["_download_allowed"](allow_download):
+                        deps["_download_block"](parsed_provider, parsed_model)
+                    download_result = deps["download_hf_model"](parsed_model, "local_mlx", progress_emit=emit_progress)
+                else:
+                    download_result = {"model": parsed_model, "path": str(deps["hf_model_dir"](parsed_model)), "cached": True}
+                    emit_progress(deps["model_download_progress_payload"](
+                        "download",
+                        "이미 다운로드된 모델을 확인했습니다.",
+                        percent=100,
+                        eta_seconds=0,
+                    ))
+            elif parsed_provider == "ollama":
+                emit_progress(deps["model_download_progress_payload"](
+                    "engine",
+                    "Ollama 서버를 확인하는 중입니다.",
+                    percent=12,
+                    indeterminate=True,
+                ))
+                deps["ensure_ollama_server"]()
+                if parsed_model not in deps["get_ollama_pulled_models"]():
+                    if not deps["_download_allowed"](allow_download):
+                        deps["_download_block"](parsed_provider, parsed_model)
+                    download_result = deps["pull_ollama_model_with_progress"](parsed_model, progress_emit=emit_progress)
+                else:
+                    download_result = {"provider": "ollama", "model": parsed_model, "cached": True}
+                    emit_progress(deps["model_download_progress_payload"](
+                        "download",
+                        "이미 다운로드된 Ollama 모델을 확인했습니다.",
+                        percent=100,
+                        detail=parsed_model,
+                        eta_seconds=0,
+                    ))
+            elif parsed_provider == "vllm":
+                if not deps["hf_model_ready"](parsed_model, "vllm"):
+                    if not deps["_download_allowed"](allow_download):
+                        deps["_download_block"](parsed_provider, parsed_model)
+                    download_result = deps["download_hf_model"](parsed_model, "vllm", progress_emit=emit_progress)
+                else:
+                    download_result = {"provider": "vllm", "model": parsed_model, "cached": True}
+                    emit_progress(deps["model_download_progress_payload"](
+                        "download",
+                        "이미 다운로드된 모델을 확인했습니다.",
+                        percent=100,
+                        detail=parsed_model,
+                        eta_seconds=0,
+                    ))
+                emit_progress(deps["model_download_progress_payload"](
+                    "server",
+                    "vLLM 서버를 시작하는 중입니다.",
+                    percent=92,
+                    indeterminate=True,
+                ))
+                deps["ensure_vllm_server"](parsed_model)
+                download_result = {**(download_result or {}), "provider": "vllm", "model": parsed_model, "server_ready": True}
+            elif parsed_provider == "llamacpp":
+                if not deps["hf_model_ready"](parsed_model, "llamacpp"):
+                    if not deps["_download_allowed"](allow_download):
+                        deps["_download_block"](parsed_provider, parsed_model)
+                    download_result = deps["download_hf_model"](parsed_model, "llamacpp", progress_emit=emit_progress)
+                else:
+                    download_result = {"provider": "llamacpp", "model": parsed_model, "cached": True}
+                    emit_progress(deps["model_download_progress_payload"](
+                        "download",
+                        "이미 다운로드된 GGUF 모델을 확인했습니다.",
+                        percent=100,
+                        detail=parsed_model,
+                        eta_seconds=0,
+                    ))
+                emit_progress(deps["model_download_progress_payload"](
+                    "server",
+                    "llama.cpp 서버를 시작하는 중입니다.",
+                    percent=92,
+                    indeterminate=True,
+                ))
+                deps["ensure_llamacpp_server"](parsed_model)
+                download_result = {**(download_result or {}), "provider": "llamacpp", "model": parsed_model, "server_ready": True}
+            elif parsed_provider == "lmstudio":
+                downloaded = {
+                    str(item.get("key") or "").strip()
+                    for item in deps["get_lmstudio_models"]()
+                    if isinstance(item, dict)
+                }
+                if parsed_model not in downloaded and not deps["_download_allowed"](allow_download):
+                    deps["_download_block"](parsed_provider, parsed_model)
+                emit_progress(deps["model_download_progress_payload"](
+                    "download",
+                    "LM Studio 모델을 확인하는 중입니다.",
+                    percent=35,
+                    indeterminate=True,
+                ))
+                ensured = deps["ensure_lmstudio_model"](parsed_model)
+                resolved_model = str(
+                    ensured.get("instance_id")
+                    or ensured.get("resolved_model")
+                    or parsed_model
+                ).strip()
+                prepared_model_name = resolved_model
+                prepared_model_id = f"lmstudio:{resolved_model}"
+                download_result = ensured
+            else:
+                emit_progress(deps["model_download_progress_payload"](
+                    "engine",
+                    "모델 연결을 준비하는 중입니다.",
+                    percent=30,
+                    indeterminate=True,
+                ))
+
+            work_result.update({
+                "model_id": prepared_model_id,
+                "parsed_provider": parsed_provider,
+                "parsed_model": prepared_model_name,
+                "install_result": install_result,
+                "download_result": download_result,
+            })
+            work_queue.put({"kind": "done"})
+        except HTTPException as exc:
+            work_queue.put({"kind": "error", "status_code": exc.status_code, "detail": exc.detail})
+        except Exception as exc:
+            logging.exception("model prepare stream worker failed")
+            work_queue.put({
+                "kind": "error",
+                "status_code": 500,
+                "detail": deps["_friendly_model_runtime_error"](exc, model_id=model_id, engine=parsed_provider),
+            })
+
+    worker = threading.Thread(target=blocking_prepare, daemon=True)
+    worker.start()
+
+    while True:
+        item = await asyncio.to_thread(work_queue.get)
+        kind = item.get("kind")
+        if kind == "progress":
+            yield sse_event("progress", item["data"])
+        elif kind == "error":
+            raise HTTPException(
+                status_code=int(item.get("status_code") or 500),
+                detail=item.get("detail") or "모델 준비에 실패했습니다.",
+            )
+        elif kind == "done":
+            break
+
+    prepared_model_id = str(work_result.get("model_id") or model_id)
+    prepared_provider = str(work_result.get("parsed_provider") or parsed_provider)
+    install_result = work_result.get("install_result") or {}
+    download_result = work_result.get("download_result")
+
+    yield sse_event("progress", deps["model_download_progress_payload"](
+        "load",
+        "모델을 메모리에 로드하는 중입니다.",
+        percent=96,
+        indeterminate=True,
+    ))
+
+    effective_email = (user_email or deps["get_current_user"](request) or "").strip()
+    user_api_key = deps["get_user_api_key"](effective_email, prepared_provider) if prepared_provider != "local_mlx" else None
+    msg = await deps["router"].load_model(
+        prepared_model_id,
+        None,
+        draft_model_id=None,
+        api_key_override=user_api_key,
+        owner=effective_email or None,
+    )
+    resolution_stream = deps["_ModelResolution"].from_request(
+        prepared_model_id,
+        engine=prepared_provider,
+        user_email=effective_email or None,
+        engine_aliases=deps["MODEL_ENGINE_ALIASES"],
+    )
+    resolution_stream.update_after_load(actual_current=deps["router"].current_model_id)
+    yield sse_event("progress", deps["model_download_progress_payload"](
+        "smoke_test",
+        "채팅 호환성 테스트 중입니다.",
+        percent=98,
+        indeterminate=True,
+    ))
+    smoke_result: Dict[str, object] = {}
+    ready_to_chat = True
+    compat_status = "ok"
+    try:
+        smoke_result = await deps["_smoke_test_loaded_model"](resolution_stream, api_key_override=user_api_key)
+        ready_to_chat = bool(smoke_result.get("ok"))
+        compat_status = str(smoke_result.get("status") or ("ok" if ready_to_chat else "degraded"))
+    except Exception as exc:
+        logging.warning("smoke test (stream) failed for %s: %s", resolution_stream.load_id, exc)
+        compat_status = "unknown"
+    result = {
+        "status": "ok",
+        "message": msg,
+        "model": prepared_model_id,
+        "current": deps["router"].current_model_id,
+        "engine": prepared_provider,
+        "installed_now": bool(isinstance(install_result, dict) and install_result.get("installed_now")),
+        "download": download_result,
+        "resolution": resolution_stream.to_dict(),
+        "downloaded": bool(download_result and not (isinstance(download_result, dict) and download_result.get("cached"))),
+        "loaded": True,
+        "ready_to_chat": ready_to_chat,
+        "compatibility_status": compat_status,
+        "smoke_test": smoke_result,
+    }
+    yield sse_event("progress", deps["model_download_progress_payload"](
+        "done",
+        "모델 준비가 완료되었습니다.",
+        percent=100,
+        eta_seconds=0,
+    ))
+    yield sse_event("done", result)
+
+
+# To maintain exact public API, model_runtime will re-export these.

package/latticeai/services/model_recommendation.py

@@ -34,7 +34,12 @@ _APPLE_ONLY_ENGINES = {"local_mlx"}
 _FAMILY_ORDER = [
     "Gemma 4",
     "Qwen3-VL",
+    "Qwen2.5-VL",
     "Llama 4",
+    "Llama 3.2 Vision",
+    "Phi Vision",
+    "Moondream",
+    "Pixtral",
 ]
 
 _SIZE_RE = re.compile(r"([\d.]+)\s*(TB|GB|MB)", re.IGNORECASE)