Pipeline complet Radiacode 103 - identification automatique d'isotopes

- VegaModel CNN-FCNN 34.5M params, 82 isotopes, val acc 99.89%
- Generation 50k spectres synthetiques 1D (12-24h durees)
- Entrainement 100 epochs sur RTX 5060 Ti (CUDA 12.8, Blackwell)
- Detection continue avec soustraction du background
- Capture background 24h avec gestion deconnexion
- Docker Compose : conteneur train (GPU) + detect (CPU/USB)
- Modele entraite inclus (vega_best.pt, 395 Mo)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

train/vega_ml/analyzer/README.md

@@ -0,0 +1,37 @@
+# Analyzer
+
+This folder contains small utilities for inspecting the **last inference request/response** processed by your middleware API.
+
+## Fetch last inference (PowerShell)
+
+Runs against the server’s log endpoints:
+- `GET /logs?limit=...`
+- `GET /logs/{id}`
+
+### Default (uses `http://99.122.58.29:443`)
+
+```powershell
+powershell -ExecutionPolicy Bypass -File analyzer/fetch_last_inference.ps1
+```
+
+### Override the base URL
+
+```powershell
+powershell -ExecutionPolicy Bypass -File analyzer/fetch_last_inference.ps1 -BaseUrl "http://99.122.58.29:443"
+```
+
+### Output
+
+Artifacts are written to:
+- `analyzer/out/last_inference_summary_*.json`
+- `analyzer/out/last_inference_detail_*.json`
+
+If the server includes `request` and/or `response` fields in the detail payload, those are also saved as:
+- `analyzer/out/last_inference_request_*.json`
+- `analyzer/out/last_inference_response_*.json`
+
+## Notes
+
+- This script assumes the log list items have at least: `id`, `method`, `path`.
+- It selects the most recent matching `POST` to an `identify` endpoint.
+- If your server uses HTTPS with a real cert, switch the URL to `https://...`.