fix: precompute skips if cache exists + integrated into compile-and-train

- precompute_features early-exits if {cache_key}.fxcache already exists
  (cache key = SHA256 of filenames + sizes, only changes when data changes)
- compile-and-train copies binaries to PVC /data/bin/ after compile
- precompute step added to compile-and-train DAG (runs after compile,
  before training, parallel with GPU warmup)
- precompute template reads binary from PVC (no GitLab download)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

crates/ml/examples/precompute_features.rs

@@ -210,6 +210,22 @@ async fn main() -> Result<()> {
 
     let t0 = Instant::now();
 
+    // ── Early exit if cache already exists for this data ─────────────────────
+    let hex_key_early = ml::feature_cache::calculate_dbn_cache_key_full(
+        &data_dir,
+        mbp10_dir.as_deref(),
+        trades_dir.as_deref(),
+    ).context("Failed to compute cache key")?;
+    let early_check_path = output_dir.join(format!("{hex_key_early}.fxcache"));
+    if early_check_path.exists() {
+        let size_mb = std::fs::metadata(&early_check_path)
+            .map(|m| m.len() as f64 / 1_048_576.0)
+            .unwrap_or(0.0);
+        println!("Cache already exists: {} ({:.1} MB) — skipping extraction",
+            early_check_path.display(), size_mb);
+        return Ok(());
+    }
+
     // ── Step 1: Load OHLCV bars from DBN files ──────────────────────────────
     info!("Loading OHLCV bars from {} (symbol={})...", data_dir.display(), opts.symbol);
     let mut dbn_files = collect_dbn_files_filtered(&data_dir, Some(&opts.symbol));

infra/k8s/argo/compile-and-train-template.yaml

@@ -115,9 +115,12 @@ spec:
               parameters:
                 - name: tag
                   value: "{{tasks.compile-training.outputs.parameters.tag}}"
+          - name: precompute
+            template: precompute
+            dependencies: [compile-training]
           - name: hyperopt
             template: hyperopt
-            dependencies: [fetch-binary, gpu-warmup]
+            dependencies: [fetch-binary, gpu-warmup, precompute]
             when: "{{workflow.parameters.hyperopt-trials}} != 0"
           - name: train-best
             template: train-best
@@ -173,6 +176,8 @@ spec:
             readOnly: true
           - name: cargo-target-cuda
             mountPath: /cargo-target
+          - name: training-data
+            mountPath: /data
         args:
           - |
             set -e
@@ -275,9 +280,52 @@ spec:
                 "${GITLAB}/api/v4/projects/1/packages/generic/foxhunt-training/${TAG}/${BIN_NAME}"
             done
 
+            # Copy binaries to PVC so precompute and other workflows can use them
+            mkdir -p /data/bin
+            cp "$BUILD/bin/training/"* /data/bin/
+            echo "=== Binaries deployed to PVC /data/bin/ ==="
+            ls -lh /data/bin/
+
             echo "=== Compile + upload done (${TAG}) ==="
             echo "$TAG" > /tmp/tag
 
+    # ── precompute: generate fxcache if data changed (skips if cache exists) ──
+    - name: precompute
+      nodeSelector:
+        k8s.scaleway.com/pool-name: ci-compile-cpu
+      tolerations:
+        - key: node.cilium.io/agent-not-ready
+          operator: Exists
+          effect: NoSchedule
+      container:
+        image: ubuntu:24.04
+        command: ["/bin/bash", "-c"]
+        env:
+          - name: RUST_LOG
+            value: info
+        args:
+          - |
+            set -e
+            BINARY=/data/bin/precompute_features
+            chmod +x "$BINARY"
+            $BINARY \
+              --data-dir /data/futures-baseline \
+              --mbp10-data-dir /data/futures-baseline-mbp10 \
+              --trades-data-dir /data/futures-baseline-trades \
+              --output-dir /data/feature-cache \
+              --symbol ES.FUT \
+              --yes
+        volumeMounts:
+          - name: training-data
+            mountPath: /data
+        resources:
+          requests:
+            cpu: "4"
+            memory: 16Gi
+          limits:
+            cpu: "28"
+            memory: 56Gi
+
     # ── gpu-warmup: trigger GPU autoscale during compile ──
     - name: gpu-warmup
       nodeSelector:

infra/k8s/argo/precompute-features-template.yaml

@@ -83,13 +83,12 @@ spec:
         args:
           - |
             set -e
-            # Fetch latest binary from GitLab Package Registry
+            # Binary deployed to PVC by compile-and-train workflow
             BINARY=/data/bin/precompute_features
-            GITLAB="http://gitlab.foxhunt.svc.cluster.local:8181"
-            echo "Downloading precompute_features from GitLab packages..."
-            curl -f -o "$BINARY" \
-              "${GITLAB}/api/v4/projects/1/packages/generic/foxhunt-training/latest/precompute_features" \
-              || echo "WARN: GitLab download failed, using existing binary on PVC"
+            if [ ! -f "$BINARY" ]; then
+              echo "ERROR: $BINARY not found on PVC. Run compile-and-train first."
+              exit 1
+            fi
             chmod +x "$BINARY"
 
             SYMBOL="{{inputs.parameters.symbol}}"