m8flow 1.1.6 → 1.1.8

package/bundled/backend/api/routes/nodes.py

@@ -2,6 +2,8 @@ from fastapi import APIRouter, HTTPException, Request, UploadFile, File
 from pydantic import BaseModel
 import os
 import shutil
+import sys
+import subprocess
 
 from core.parser import parse_node_code
 from templates import TEMPLATES
@@ -13,6 +15,42 @@ class ParseRequest(BaseModel):
     code: str
 
 
+class InstallRequest(BaseModel):
+    package: str   # e.g. "lightgbm" or "xgboost==2.0.3"
+
+
+@router.post("/install")
+def install_package(req: InstallRequest):
+    """
+    Install a Python package into the currently running venv using pip.
+    Returns stdout/stderr so the frontend can display real output.
+    """
+    package = req.package.strip()
+
+    # Basic safety: reject obviously dangerous strings
+    forbidden = (";", "&", "|", "`", "$", ">", "<", "\n", "\r")
+    if not package or any(c in package for c in forbidden):
+        raise HTTPException(status_code=422, detail="Invalid package name.")
+
+    result = subprocess.run(
+        [sys.executable, "-m", "pip", "install", package,
+         "--disable-pip-version-check"],
+        capture_output=True,
+        text=True,
+        timeout=180,
+    )
+
+    output = (result.stdout + result.stderr).strip()
+
+    if result.returncode != 0:
+        raise HTTPException(
+            status_code=400,
+            detail=output or f"pip install failed (exit {result.returncode})",
+        )
+
+    return {"ok": True, "output": output, "package": package}
+
+
 @router.post("/parse")
 def parse(req: ParseRequest):
     """Parse Python code and return the derived NodeSchema (inputs, outputs, errors)."""
@@ -92,6 +130,22 @@ async def generate_node_code_route(http_request: Request, req: GenerateCodeReque
         raise HTTPException(status_code=500, detail=f"{type(exc).__name__}: {exc}")
 
 
+@router.get("/packages")
+def list_packages():
+    """Return all packages currently installed in the running Python environment."""
+    result = subprocess.run(
+        [sys.executable, "-m", "pip", "list", "--format=json", "--disable-pip-version-check"],
+        capture_output=True,
+        text=True,
+        timeout=30,
+    )
+    if result.returncode != 0:
+        raise HTTPException(status_code=500, detail="Failed to list packages.")
+    import json as _json
+    packages = _json.loads(result.stdout or "[]")
+    return {"packages": packages}
+
+
 @router.get("/templates")
 def list_templates():
     """Return every prebuilt template with its pre-parsed schema attached."""

package/bundled/backend/core/code_validator.py

@@ -17,14 +17,14 @@ from typing import NamedTuple
 # ── Blocklists ────────────────────────────────────────────────────────────────
 
 BLOCKED_IMPORTS = frozenset({
-    # System / process execution
-    "os", "sys", "subprocess", "shutil", "signal",
+    # System / process execution (os is allowed with restrictions below)
+    "sys", "subprocess", "shutil", "signal",
     # Networking
     "socket", "http", "urllib", "requests", "httpx", "ftplib", "smtplib",
     # Code injection
     "importlib", "builtins", "ctypes", "cffi",
-    # Filesystem I/O (allow pandas read_csv via allowlist only)
-    "pathlib", "glob", "tempfile",
+    # Filesystem wildcards
+    "glob",
     # Threading / multiprocessing
     "threading", "multiprocessing", "concurrent",
     # Package management
@@ -32,19 +32,25 @@ BLOCKED_IMPORTS = frozenset({
 })
 
 BLOCKED_BUILTINS = frozenset({
-    "eval", "exec", "compile", "__import__", "open", "input",
+    "eval", "exec", "compile", "__import__", "input",
     "breakpoint", "vars", "dir",
+    # os-level execution even if os is imported
+    "system", "popen", "execv", "execve", "execvp",
 })
 
 ALLOWED_IMPORTS = frozenset({
     # Scientific computing
     "numpy", "pandas", "scipy", "sklearn", "xgboost", "lightgbm",
     "statsmodels", "imblearn",
+    # Model persistence
+    "joblib", "pickle",
     # Plotting
     "matplotlib", "seaborn", "plotly", "mpl_toolkits",
     # Standard safe libs
     "math", "statistics", "itertools", "functools", "collections",
     "json", "re", "datetime", "typing",
+    # Filesystem (path operations, needed for model saving)
+    "os", "io", "pathlib", "tempfile",
 })
 
 

package/bundled/backend/templates.py

@@ -2769,6 +2769,56 @@ def run(
 '''
 
 
+MODEL_SAVER = '''
+def run(model, file_path: str = "saved_model.pkl") -> dict:
+    """Save a trained model to disk using joblib (faster + handles numpy arrays better than pickle).
+
+    file_path : path where the model file will be written.
+                Supports any extension (.pkl / .joblib / .model).
+    Returns the absolute path so downstream nodes can reference it.
+    """
+    import joblib
+    import os
+
+    abs_path = os.path.abspath(file_path)
+    parent   = os.path.dirname(abs_path)
+    if parent:
+        os.makedirs(parent, exist_ok=True)
+
+    joblib.dump(model, abs_path)
+    size_kb = round(os.path.getsize(abs_path) / 1024, 1)
+
+    return {
+        "path":    abs_path,
+        "size_kb": size_kb,
+        "status":  "saved",
+    }
+'''
+
+MODEL_LOADER = '''
+def run(file_path: str = "saved_model.pkl") -> dict:
+    """Load a previously saved model from disk using joblib.
+
+    file_path : path to the model file produced by Model Saver.
+    Returns the model object ready for inference or further training.
+    """
+    import joblib
+    import os
+
+    abs_path = os.path.abspath(file_path)
+    if not os.path.exists(abs_path):
+        raise FileNotFoundError(f"Model file not found: {abs_path}")
+
+    model = joblib.load(abs_path)
+
+    return {
+        "model":   model,
+        "path":    abs_path,
+        "status":  "loaded",
+    }
+'''
+
+
 TEMPLATES: list[dict] = [
     # Data
     {"id": "csv_loader",                  "label": "CSV Loader",              "category": "Data",          "code": CSV_LOADER},
@@ -2892,6 +2942,10 @@ TEMPLATES: list[dict] = [
     {"id": "calibration_curve_data",        "label": "Calibration Curve",       "category": "Evaluation",    "code": CALIBRATION_CURVE_DATA},
     {"id": "learning_curve_analyzer",       "label": "Learning Curve Analyzer", "category": "Evaluation",    "code": LEARNING_CURVE_ANALYZER},
     {"id": "cost_benefit_matrix",           "label": "Cost-Benefit Matrix",     "category": "Evaluation",    "code": COST_BENEFIT_MATRIX},
+
+    # ── Model Persistence ────────────────────────────────────────────────────
+    {"id": "model_saver",                   "label": "Model Saver",             "category": "Data",          "code": MODEL_SAVER},
+    {"id": "model_loader",                  "label": "Model Loader",            "category": "Data",          "code": MODEL_LOADER},
 ]