Learning fixes for v1.8.1 by MahdiAll99 · Pull Request #550 · MEDomicsLab/MEDomics

MahdiAll99 · 2026-04-09T15:54:11Z

No description provided.

…into learningFixes

…ndled

Copilot

Pull request overview

This PR bundles a set of “learning” fixes intended for the v1.8.1 release, spanning frontend UI tweaks, improved error propagation from the Go/Python execution layer, and updated evaluation/metrics handling in the Python ML pipeline.

Changes:

Update model-result rendering to incorporate overall_metrics (e.g., optimized-threshold metrics) and enhance dataset tooltip copy.
Improve CodeEditor error handling to accept structured backend errors (message + stack trace).
Enhance Go server execution error reporting when Python exits before emitting a response, and adjust Python evaluation output column naming.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
renderer/components/mainPages/settings.jsx	Re-run initial settings/status checks when `pythonEmbedded` changes.
renderer/components/learning/results/node/modelsResults.jsx	Incorporate `overall_metrics` into displayed model metrics.
renderer/components/learning/nodesTypes/datasetNode.jsx	Tooltip copy/formatting improvements for class-imbalance guidance.
renderer/components/flow/codeEditor.jsx	Support structured error objects and render stack traces in the UI.
pythonCode/modules/evaluation/predict_test.py	Rename `prediction_score` to `confidence_score` in prediction output.
pythonCode/med_libs/MEDml/nodes/ModelHandler.py	Threshold-optimization metric handling + recalculation of reported metrics.
go_server/src/utils.go	Capture stderr for early Python crashes and return structured error payloads.
electron-builder.yml	Package `baseFiles/emptyNotebook.ipynb` into app resources.

Comments suppressed due to low confidence (2)

renderer/components/flow/codeEditor.jsx:227

When a file loads successfully, error / stackTrace are never cleared. If the user previously hit an error, the component will keep rendering the error view (and/or stale stack trace) even after a subsequent successful load. Clear error and stackTrace when starting a load and on the success path; also clear stackTrace when response.error is a string or when the object error lacks stack_trace.

        if (response.error) {
          if (typeof response.error === "object") {
            setError(response.error.message || "An error occurred while loading the file.")
            toast.error("Error loading file: " + (response.error.message || "Unknown error"))
            if (response.error.stack_trace) {
              setStackTrace(response.error.stack_trace)
            }
          } else if (typeof response.error === "string") {
            setError(response.error)
            toast.error("Error opening file: " + response.error)
          }
        }
        else {
          setContent(response.content)
          setMode(response.language)
          toast.success("Opened file successfully")
        }

pythonCode/med_libs/MEDml/nodes/ModelHandler.py:752

There are two set_model methods; the second definition overwrites the first, making the try/except handling and the tuning settings update in the first method unreachable. Remove the duplicate and keep a single implementation (or rename them if both behaviors are needed).

    def set_model(self, model_id: str) -> None:
        try:
            model_obj = self.global_config_json['nodes'][model_id]
            self.config_json['data']['estimator'] = {
                "type": model_obj['data']['selection'],
                "settings": model_obj['data']['settings']
            }
            if self.isTuningEnabled:
                self.config_json['data']['internal']['settingsTuning'] = model_obj['data']['internal'].get('settingsTuning', {})
        except KeyError as e:
            print(f"ERROR: Failed to set model {model_id}. Missing key: {e}")
            raise ValueError(f"Model configuration for {model_id} not found in the global config.")
        except Exception as e:
            print(f"ERROR: An error occurred while setting the model: {e}")
            raise ValueError(f"An error occurred while setting model {model_id}.")

    def set_model(self, model_id: str) -> None:
        model_obj = self.global_config_json['nodes'][model_id]
        self.config_json['data']['estimator'] = {
            "type": model_obj['data']['selection'],
            "settings": model_obj['data']['settings']
        }

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-09T15:59:20Z

renderer/components/learning/results/node/modelsResults.jsx

  useEffect(() => {
    let allModelsData = []
    if (models.length > 0) {
-      models.forEach((model) => {
-        if (!model.metrics) return
-        let modifiedRow = { ...model.metrics }
-        modifiedRow["Parameters"] = model.params
-        modifiedRow["OverallMetrics"] = selectedResults?.data?.overall_metrics
-        modifiedRow = Object.assign({ Name: model.name }, modifiedRow)
-        allModelsData.push(modifiedRow)
-      })
+     models.forEach((model) => {
+      if (!model.metrics) return
+      let modifiedRow = { ...model.metrics }
+
+      // FIX: if overall_metrics exists, use these new metrics (optimal threshold)
+      const overallMetrics = selectedResults?.data?.overall_metrics
+      if (overallMetrics) {
+        Object.keys(overallMetrics).forEach((key) => {
+          modifiedRow[key] = overallMetrics[key].mean
+        })
+      }


This effect reads selectedResults (for overall_metrics) but it isn’t included in the dependency array. That can lead to stale metrics if selectedResults changes without triggering a models state change, and it will also trip the React hooks exhaustive-deps rule. Include selectedResults (or selectedResults.data.overall_metrics) in the dependency list, or derive overallMetrics earlier and pass it in.

Copilot · 2026-04-09T15:59:21Z

pythonCode/med_libs/MEDml/nodes/ModelHandler.py

 from sklearn.metrics import (accuracy_score, confusion_matrix, f1_score,
                             matthews_corrcoef, precision_score, recall_score,
-                             roc_auc_score)
+                             roc_auc_score, roc_curve)



roc_curve is imported but never used. This will fail linting/CI in setups that treat unused imports as errors. Remove the unused import (or add the missing usage if intended).

Copilot · 2026-04-09T15:59:21Z

pythonCode/med_libs/MEDml/nodes/ModelHandler.py

            y_proba = model.predict_proba(X_test_fold)[:, 1] if hasattr(model, 'predict_proba') else None

-            # Calculate all metrics manually
+            if y_proba is not None:
+                # FIX 2: read probability_threshold with fallback to threshold
+                threshold = getattr(model, 'probability_threshold', getattr(model, 'threshold', 0.5))
+                y_pred = (y_proba >= threshold).astype(int)
+            else:
+                y_pred = model.predict(X_test_fold)
+


__custom_train_and_evaluate now derives y_pred by thresholding predict_proba()[:, 1] for any classifier that exposes predict_proba. This is incorrect for multiclass classification (it will force binary predictions based on the second column) and can corrupt all fold metrics. Gate the thresholding logic to binary problems only (e.g., len(np.unique(y_test_fold)) == 2 and/or only when optimize_threshold was applied / the model exposes probability_threshold).

Copilot · 2026-04-09T15:59:21Z

go_server/src/utils.go

 	stdout, err := Scripts[id].Cmd.StdoutPipe()
 	Mu.Unlock()
 	if err != nil {
-		log.Println("Error getting stdout pipe")
-		log.Println(err.Error())
+		log.Println("Error getting stdout pipe: " + err.Error())
 	}
 	Mu.Lock()
 	stderr, err := Scripts[id].Cmd.StderrPipe()
 	Mu.Unlock()
 	if err != nil {
-		log.Println("Error getting stderr pipe")
-		log.Println(err.Error())
+		log.Println("Error getting stderr pipe: " + err.Error())
 	}


If StdoutPipe() / StderrPipe() fails, the function logs the error but continues. That can pass a nil reader into copyOutput / copyOutputErr and lead to panics or hangs later. Return early with the pipe error (and consider cleaning up Scripts[id]) instead of continuing.

Copilot · 2026-04-09T15:59:21Z

go_server/src/utils.go

 	response := ""
-	go copyOutput(stdout, &response)
-	go copyOutput(stderr, &response)
-	err = Scripts[id].Cmd.Wait()
-	if err != nil {
-		log.Println("Error waiting for command to finish")
-		return "", err
+	// stderrBuf collects every line from stderr so we can surface it on early
+	// crashes (e.g. ModuleNotFoundError) that never reach send_response().
+	var stderrBuf strings.Builder
+	var stderrMu sync.Mutex
+
+	var wg sync.WaitGroup
+	wg.Add(2)
+	go func() {
+		defer wg.Done()
+		copyOutput(stdout, &response)
+	}()
+	go func() {
+		defer wg.Done()
+		copyOutputErr(stderr, &response, &stderrBuf, &stderrMu)
+	}()
+	wg.Wait()


response is written from both stdout and stderr goroutines without synchronization (*response = ...). This introduces a data race and nondeterministic behavior if both streams emit a response-ready signal or if a crash path writes while stdout is still running. Protect response with a mutex/atomic or send the response via a channel and only set it from a single goroutine.

…ual pycaret exp and not fold exp)

Copilot

Pull request overview

Copilot reviewed 10 out of 10 changed files in this pull request and generated 3 comments.

Comments suppressed due to low confidence (2)

renderer/components/flow/codeEditor.jsx:227

When a file load succeeds (or when an error without stack_trace occurs), the component doesn’t clear error/stackTrace. This can leave the UI stuck in the error view or show a stale stack trace from a previous failure. Clear error and stackTrace at the start of loadFileContent and/or in the success path, and ensure the string-error path also resets stackTrace.

        if (response.error) {
          if (typeof response.error === "object") {
            setError(response.error.message || "An error occurred while loading the file.")
            toast.error("Error loading file: " + (response.error.message || "Unknown error"))
            if (response.error.stack_trace) {
              setStackTrace(response.error.stack_trace)
            }
          } else if (typeof response.error === "string") {
            setError(response.error)
            toast.error("Error opening file: " + response.error)
          }
        }
        else {
          setContent(response.content)
          setMode(response.language)
          toast.success("Opened file successfully")
        }

pythonCode/med_libs/MEDml/nodes/ModelHandler.py:855

ModelHandler defines set_model twice. In Python the second definition overwrites the first, so the error handling and tuning-update logic in the first implementation becomes dead code. Remove one of the definitions (or merge their behavior) so there is a single authoritative set_model implementation.

    def set_model(self, model_id: str) -> None:
        try:
            model_obj = self.global_config_json['nodes'][model_id]
            self.config_json['data']['estimator'] = {
                "type": model_obj['data']['selection'],
                "settings": model_obj['data']['settings']
            }
            if self.isTuningEnabled:
                self.config_json['data']['internal']['settingsTuning'] = model_obj['data']['internal'].get('settingsTuning', {})
        except KeyError as e:
            print(f"ERROR: Failed to set model {model_id}. Missing key: {e}")
            raise ValueError(f"Model configuration for {model_id} not found in the global config.")
        except Exception as e:
            print(f"ERROR: An error occurred while setting the model: {e}")
            raise ValueError(f"An error occurred while setting model {model_id}.")

    def set_model(self, model_id: str) -> None:
        model_obj = self.global_config_json['nodes'][model_id]
        self.config_json['data']['estimator'] = {
            "type": model_obj['data']['selection'],
            "settings": model_obj['data']['settings']
        }

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

renderer/components/mainPages/settings.jsx

  useEffect(() => {
    ipcRenderer.invoke("get-settings").then((receivedSettings) => {
      console.log("received settings", receivedSettings)
      setSettings(receivedSettings)
      if (pythonEmbedded.pythonEmbedded) {
        setCondaPath(pythonEmbedded.pythonEmbedded)
      } else if (receivedSettings?.condaPath) {
        setCondaPath(receivedSettings?.condaPath)
      }
      if (receivedSettings?.seed) {
        setSeed(receivedSettings?.seed)
      }
    })
    // ipcRenderer.invoke("server-is-running").then((status) => {
    //   setServerIsRunning(status)
    //   console.log("server is running", status)
    // })
    checkMongoIsRunning()
    checkServer()
    getJupyterStatus()
-  }, [])
+  }, [pythonEmbedded])


renderer/components/learning/results/node/modelsResults.jsx

+      // FIX: if overall_metrics exists, use these new metrics (optimal threshold)
+      const overallMetrics = selectedResults?.data?.overall_metrics
+      if (overallMetrics) {
+        Object.keys(overallMetrics).forEach((key) => {
+          modifiedRow[key] = overallMetrics[key].mean
+        })
+      }
+
+      modifiedRow["Parameters"] = model.params
+      modifiedRow["OverallMetrics"] = overallMetrics
+      modifiedRow = Object.assign({ Name: model.name }, modifiedRow)
+      allModelsData.push(modifiedRow)


pythonCode/med_libs/MEDml/nodes/ModelHandler.py

            self.optimize_threshold = self.config_json['data']['internal'].get('optimizeThreshold', False)
            if self.optimize_threshold:
                self.threshold_optimization_metric = self.config_json['data']['internal'].get('threshOptimizationMetric', 'Accuracy')
+                            # Normalisation du nom de métrique pour PyCaret
+                METRIC_NAME_MAP = {
+                    'Youden': 'Youden Index',
+                    'BAC': 'Balanced Accuracy',
+                    'Specificity': 'Specificity',
+                    'NPV': 'NPV',
+                }
+                self.threshold_optimization_metric = METRIC_NAME_MAP.get(
+                    self.threshold_optimization_metric, 
+                    self.threshold_optimization_metric
+                )


…path defintion

kalm7073 and others added 13 commits February 2, 2026 23:25

Updated default inverse tooltip

4eb1c0f

Merge branch 'learningFixes' of github.com:MEDomics-UdeS/MEDomicsLab …

e56a108

…into learningFixes

Merge branch 'learningFixes' of github.com:MEDomics-UdeS/MEDomicsLab …

88d952b

…into learningFixes

Merge branch 'learningFixes' of github.com:MEDomics-UdeS/MEDomicsLab …

ef2a5f6

…into learningFixes

Merge branch 'learningFixes' of github.com:MEDomics-UdeS/MEDomicsLab …

3c0f2a4

…into learningFixes

Merge branch 'learningFixes' of github.com:MEDomics-UdeS/MEDomicsLab …

d94f6fd

…into learningFixes

Fixed optimize_threshold problems (front and back)

b2540ac

added base files to resources

40a19fc

fix in conda env path in settings

6866fae

renamed prediction score to confidence score

c08bf3a

script kill bug fix

1343887

error handling bug fix and added stack trace

341f242

Improved error handling, errors outside start() were not correctly ha…

b48eaea

…ndled

MahdiAll99 requested a review from Copilot April 9, 2026 15:54

Copilot started reviewing on behalf of MahdiAll99 April 9, 2026 15:55 View session

Copilot AI reviewed Apr 9, 2026

View reviewed changes

MahdiAll99 added 3 commits April 9, 2026 14:14

Progress bar message update

60ed7b2

Fix: avoid duplicate imports

04db9df

Huge fix on code generation feature and final model training (use act…

b2ebdb2

…ual pycaret exp and not fold exp)

MahdiAll99 requested a review from Copilot April 10, 2026 16:13

Copilot started reviewing on behalf of MahdiAll99 April 10, 2026 16:14 View session

Copilot AI reviewed Apr 10, 2026

View reviewed changes

MahdiAll99 added 3 commits April 10, 2026 12:26

Unused function: set_model

2c9a050

Seperated useEffects to better handle changes in python or conda env …

e035f19

…path defintion

minor comment fix

57118f1

MahdiAll99 merged commit db9ac0a into develop Apr 10, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Learning fixes for v1.8.1#550

Learning fixes for v1.8.1#550
MahdiAll99 merged 19 commits intodevelopfrom
learningFixes

MahdiAll99 commented Apr 9, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

MahdiAll99 commented Apr 9, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants