feat: request ids on uploads, fp calls and evaluator calls

CHANGELOG.md

@@ -1,5 +1,26 @@
 # Changelog
 
+## 1.0.10-beta.0 — 2026-09-30
+
+### Features
+
+- Uploads, `fp` calls and evaluator calls carry a request id; the daemon also sends batch and machine ids, and `fp` errors show a `ref` (#872)
+
+### Fixes
+
+- The daemon logs each failed upload attempt at warning level with its request id and batch id, so every attempt of a batch can be found; before, only the last one was visible (#872)
+- Parking a failed batch works when the spool and state directories are on different filesystems (e.g. separate Docker volumes); before, the batch stayed in the spool and was re-sent on every sweep. The move is crash-safe: the original is deleted only after the copy's directory entry is on disk (#872)
+
+### Docs
+
+- Troubleshooting, HTTP API and Cloud CLI pages explain the `ref` / `request_id` to quote to support, and where the daemon logs them (#872)
+
+### Dependencies
+
+- Python SDK dev lockfile: oauthlib 3.3.1 → 4.0.0 and pyjwt 2.13.0 → 2.15.1, clearing three OSV advisories (#872)
+- Pin brace-expansion 5.0.9 → 5.0.12 (package.json override), clearing three OSV advisories (#872)
+- Python SDK and `fp` CLI lockfiles: urllib3 2.7.0 → 2.8.0, clearing three OSV advisories (#872)
+
 ## 1.0.9 — 2026-09-29
 
 Action needed if you use Jev: its log-only mode is now `observe`, a `jev.json` still set to `shadow` is refused (Jev stays off until `failproofai jev setup` is run again), and Jev's checks now come only from `failproofai policies add FailproofAI/jev-policies`. For Hermes, `failproofai update` moves every profile from the old shell hooks (never run for cron jobs) to the native plugin, and every agent config failproofai edits is written crash-safely with a `.failproofai-backup`. Collects 1.0.9-beta.0 to beta.2 below.

crates/failproofaid/src/main.rs

@@ -790,8 +790,10 @@ fn collector_tasks() -> Vec<fpai_collect::TaskSpec> {
         ingest.key.clone(),
         cfg.failed_dir.clone(),
     )
-    .map(|u| u.with_redact(cfg.settings.redact))
-    {
+    .map(|u| {
+        u.with_redact(cfg.settings.redact)
+            .with_machine_id(cfg.settings.machine_id.as_deref())
+    }) {
         Ok(u) => std::sync::Arc::new(u),
         Err(err) => {
             eprintln!("[failproofaid] collector disabled: {err}");

crates/fpai-collect/Cargo.toml

@@ -49,6 +49,11 @@ rusqlite = { version = "0.40", features = ["bundled"] }
 # reason rustls is chosen above: no C toolchain on the four cross-compiled legs.
 ruzstd = "0.9"
 time = { version = "0.3", default-features = false, features = ["std", "formatting", "macros"] }
+# Request ids and batch ids on each upload (see uploader.rs). Both were already
+# in the lockfile — sha2 via failproofaid, getrandom via the TLS stack — so
+# neither adds a crate to the four cross-compiled legs; `uuid` would have.
+getrandom = "0.3"
+sha2 = "0.11"
 
 [dev-dependencies]
 tokio = { version = "1", features = ["rt-multi-thread", "macros", "time", "sync", "test-util", "net"] }

crates/fpai-collect/src/delivery.rs

@@ -37,7 +37,7 @@ use tokio::sync::Semaphore;
 
 use crate::spool::is_batch_file;
 use crate::supervisor::{Shutdown, TaskError};
-use crate::uploader::{ParkedName, Uploader};
+use crate::uploader::{ParkedName, Uploader, batch_id};
 
 /// Concurrent uploads across the watcher and sweeper combined.
 ///
@@ -141,7 +141,14 @@ impl Delivery {
 
         match self.uploader.upload_file(&path).await {
             Ok(()) => tracing::debug!(file = %path.display(), "uploaded"),
-            Err(err) => tracing::warn!(file = %path.display(), %err, "upload failed"),
+            // The batch id ties this to the attempt and park lines, which
+            // carry the request ids.
+            Err(err) => tracing::warn!(
+                file = %path.display(),
+                batch_id = %batch_id(&path),
+                %err,
+                "upload failed"
+            ),
         }
     }
 }
@@ -324,6 +331,12 @@ async fn stale_batches(dir: &Path, min_age: Duration, max: usize) -> Vec<PathBuf
 /// Skips anything poison or carrying a definitive client status — those fail
 /// identically until a human fixes the key or the URL, and retrying them burns
 /// permits that batches which could succeed are waiting for.
+/// Whether the parked-batch sweep may retry the file with this name.
+fn sweepable(name: &str) -> bool {
+    // A dotfile is a park still being copied across filesystems.
+    !name.starts_with('.') && ParkedName::parse(name).is_auto_retryable()
+}
+
 async fn retry_parked(delivery: &Delivery, failed_dir: &Path, sd: &Shutdown) {
     let Ok(mut rd) = tokio::fs::read_dir(failed_dir).await else {
         return;
@@ -336,7 +349,7 @@ async fn retry_parked(delivery: &Delivery, failed_dir: &Path, sd: &Shutdown) {
         let Some(name) = path.file_name().and_then(|n| n.to_str()) else {
             continue;
         };
-        if !ParkedName::parse(name).is_auto_retryable() {
+        if !sweepable(name) {
             continue;
         }
         let Ok(meta) = entry.metadata().await else {
@@ -414,6 +427,14 @@ mod tests {
         std::fs::remove_dir_all(&dir).ok();
     }
 
+    #[test]
+    fn the_parked_sweep_skips_rejected_batches_and_half_copied_parks() {
+        assert!(sweepable("hooks-a-1-0.a1.jsonl"));
+        assert!(!sweepable("hooks-a-1-0.a1.c401.jsonl"));
+        assert!(!sweepable("hooks-a-1-0.a3.jsonl.poison"));
+        assert!(!sweepable(".hooks-a-1-0.a1.jsonl.partial"));
+    }
+
     #[tokio::test]
     async fn the_sweeper_only_claims_batch_files() {
         let dir = tmpdir("filter");