# FileServer A minimal Julia service that receives files over HTTP and hands them off to a pool of worker threads for processing. The HTTP endpoint does no real work: it spools each uploaded file to disk, pushes a lightweight reference onto a work queue, and responds immediately, staying free to accept the next upload. The per-file processing is deliberately thin. Each file runs through a small neural-network classifier that labels it **known** (a file type resembling the training set) or **unknown**, and logs the result. See "File classifier" below. ## Architecture The pipeline is four stages, each with its own bounded queue and its own worker pool (tuned independently, since classification is CPU-bound, known-file enrichment is process-/IO-bound, content triage is cheap IO, and language enrichment mixes CPU with a subprocess): ``` POST /upload (multipart) │ ▼ ┌─────────────────┐ stream bytes to disk (never buffered) │ HTTP handler │────────────────────────► data/spool/- │ (streaming) │ the file's home for its └────────┬─────────┘ enqueue reference whole time in flight │ (non-blocking) ▼ │ 202 + job IDs ▼ (503 if full) ┌────────────────────┐ │ stage-1 queue │ classification └─────────┬──────────┘ │ dequeue ┌───────────────────┼───────────────────┐ ▼ ▼ ▼ classify wkr 1 classify wkr 2 … classify wkr N │ ┌────────────┴────────────┐ :unknown :known the file itself never moves; │ │ routing is the enqueue alone │ enqueue (blocking) │ enqueue (blocking backpressure) ▼ ▼ ┌────────────────────┐ ┌────────────────────┐ │ unknown queue │ │ known queue │ enrichment └─────────┬──────────┘ └─────────┬──────────┘ │ dequeue │ dequeue ┌────────┼────────┐ ┌───────────┼───────────┐ ▼ ▼ ▼ ▼ ▼ ▼ unk 1 unk 2 … unk K known wkr 1 known wkr 2 … known wkr M │ binary-vs-text sniff │ exiftool → normalized sidecar ├─► data/binary/- success ──┴──► data/done/- │ (terminal — moved) data/done/-.meta.json │ (sidecar-first commit) │ :text enqueue (blocking failure ───────► data/failed/- ▼ backpressure) ┌────────────────────┐ │ text queue │ language enrichment └─────────┬──────────┘ │ dequeue ┌────────┼────────┐ ▼ ▼ ▼ txt 1 txt 2 … txt P │ Languages.jl (natural language) + github-linguist (programming language) └─► data/text_done/- + data/text_done/-.meta.json (sidecar-first commit) ``` A file is written once, into `data/spool/`, and stays there for its entire time in the pipeline. Stages hand it on by enqueueing its small `Job` reference, never by moving bytes: the queue holding the reference *is* the record of which stage the file has reached. The only move is the last one, into a terminal sink (`data/done/`, `data/text_done/`, `data/binary/`) or into `data/failed/` if a worker throws. Stage 1 used to rename each file into `data/known/` or `data/unknown/` first, and stage 3 into `data/text/`. Those three directories are gone, and with them 11.6 µs per file, the equal of the classifier itself. Stage 1 now runs at ~85k files/s on one worker instead of ~35k. Stages 2 (known-file enrichment) and 3 (content triage) run in parallel: stage 1 feeds both the known and unknown queues. Stage 3 in turn feeds stage 4 (language enrichment) for every file it sorts as text. Key properties: - **Fast intake:** the queue only ever carries small references; file bytes live on disk, so memory stays flat regardless of file size. This holds end to end: intake streams each upload from the socket to the spool file a chunk at a time (`FS_UPLOAD_CHUNK_BYTES`, default 64 KiB) rather than buffering the body, and every worker reads only a bounded prefix. Measured: uploads of 256 MiB, 1 GiB and 2 GiB each grow resident memory by ~20 MiB, a flat line in file size. See "Streaming intake" and "Benchmarking" below. - **Backpressure:** each queue is bounded (default 1000). When the *intake* queue is full, uploads get `503 Service Unavailable`. When the *known* queue is full, the stage-1 worker blocks and retries (a classified file is never dropped). - **Crash-resilient:** files survive on disk. On startup everything left in `data/spool/` is re-enqueued (`recovered` in the log) and **replays from stage 1**. That is safe rather than merely tolerable: classification and the binary/text sniff are pure functions of the file's bytes, and the terminal commits rename with `force=true`, so a replayed file lands where it would have landed and overwrites its own sidecar. Recovery *blocks* on a full queue rather than dropping the excess, and runs with the worker pools already live, so a backlog larger than one queue's capacity takes longer to re-drive but none of it is abandoned. The cost of replay is redoing stages a file had already cleared. That is a property of the *queue*, not of the directory layout: the default queues are in-process (`src/queue.jl`), so a crash destroys the only record of how far each file got. Per-stage directories used to stand in for that record, at the price of a rename per file per stage on the hot path: a permanent cost on every file, to buy a cheaper restart. The seam in `src/queue.jl` is where that is actually fixed: run with `FS_QUEUE_BACKEND=rabbitmq` and a job stays unacked until its handler commits, so a restart resumes each file at the stage it had reached instead of re-driving `spool/` from stage 1. See "Queue backends". - **Graceful shutdown:** SIGINT (Ctrl-C) and SIGTERM (systemd/Docker/k8s `stop`) both stop accepting uploads, then drain the stages *in order*: close the stage-1 queue and wait out the classify workers (the only producer of the known *and* unknown queues), then close those queues and wait out the enrich and content-triage workers (content triage being the only producer of the text queue), then close the text queue and wait out the language-enrichment workers. (See "Shutdown" below for one cosmetic caveat on SIGTERM.) - **Safe filenames:** client-supplied names are sanitized and prefixed with a server-minted UUID before touching the filesystem (no path traversal). ### Streaming intake The upload endpoint never holds a file in memory. Bytes go socket → spool file in `FS_UPLOAD_CHUNK_BYTES` chunks, so resident memory per in-flight upload is set by the chunk size, not the file size. A 2 GiB upload costs about what a 2 KiB one does. Two pieces make that work, and both are deliberate: - **`src/multipart.jl`, an incremental multipart parser.** HTTP.jl's `parse_multipart_form` takes the *complete* body as a byte vector, so using it means every file in the request is in memory at once (and copied again per part). The reader here pulls fixed-size chunks and hands each part's bytes straight to its spool file. Its interface is two calls in a loop (`next_part!` then `write_part_body!`, or `skip_part_body!`), so the handler keeps ordinary control flow instead of inverting into callbacks. The subtle part is that a boundary delimiter can straddle two chunks, so the buffer always retains the last `length(delimiter)-1` bytes; the test suite parses the same body at chunk sizes from 1 byte upward to put that split at every offset. - **`/upload` bypasses Oxygen's router.** Oxygen's root handler wraps `HTTP.streamhandler`, which does `request.body = read(stream)` *before* dispatching, even for an Oxygen `@stream` route, so no route can stream an upload. `run` therefore passes its own `handler` to `serve` (`root_stream_handler`), which intercepts `POST /upload` at the stream level and delegates everything else to Oxygen unchanged. The trade-off: `/upload` is absent from Oxygen's built-in metrics and docs. Streaming also changes what the endpoint can promise. A buffered handler knows up front how many files a request holds; this one discovers them as they arrive. So when the intake queue fills mid-request it does not abandon the connection: it stops spooling (discarding the remaining parts rather than writing files it can't queue), drains the body, and answers `503` with the `accepted` list of whatever got in first. Files already queued stay queued, and the client can retry the rest. A client that hangs up mid-upload is treated as routine: the partial spool file is removed (so restart recovery can never pick up a truncated upload as if it were complete) and the event is logged `upload aborted by client`. One cosmetic caveat, like the SIGTERM one below: when a request body is cut short, HTTP.jl's own `closeread` logs an `EOFError` after the handler returns, because the connection promised more bytes via `Content-Length` than arrived. It's harmless noise from inside HTTP.jl: the partial file is already cleaned up and the connection closed. ### Metadata enrichment (stage 2) Files the classifier labels **known** are handed to a second pool that extracts metadata with [`exiftool`](https://exiftool.org/) (`exiftool -json -G -n`), chosen because no native Julia library comes close to its multi-format coverage. The output is normalized into a small, stable, documented schema and written as a JSON **sidecar** next to the file in `data/done/`, e.g. `data/done/-.meta.json`. The original bytes are never modified. > **Prerequisite:** `exiftool` must be on `PATH` (e.g. `apt install > libimage-exiftool-perl`). The server fails fast at startup if it's missing. Sidecar top-level fields (all nullable, present only when available), plus the complete raw `exiftool` object under `raw`: | Field | Meaning | |---|---| | `id`, `original_name` | job id and client-supplied name | | `file_type`, `mime_type` | e.g. `PDF` / `application/pdf` | | `file_size` | bytes (authoritative, from intake, not exiftool) | | `created_date`, `modified_date` | content timestamps | | `author` | person (`Author`/`Artist`/`By-line`) | | `created_by` | authoring app/tool (`Producer`/`CreatorTool`/`Creator`/`Software`/…) | | `dimensions` | `{width, height}` for media | | `duration` | seconds, for audio/video | | `page_count` | for documents | | `error` | set on a *degraded* sidecar (see below) | | `raw` | full `exiftool` output | Each normalized field is a coalesce over a priority list of exiftool tags (`src/metadata.jl`); extend a field by appending tag names. If extraction fails or `exiftool` times out (`FS_EXIFTOOL_TIMEOUT`, default 30s), the file still completes to `data/done/` with a **degraded sidecar** (`file_size`/`file_type` plus an `error` note) rather than being quarantined, because it's still a wanted known file. Only genuine I/O errors (can't write the sidecar or move the file) send it to `data/failed/`. The sidecar is committed *before* the file is moved into `data/done/`, so a file's presence there always implies its sidecar is already present; a crash in between leaves only a harmless orphan sidecar, and recovery re-enriches idempotently. ### Content triage (stage 3) Files the classifier labels **unknown** are handed to a third pool that sorts them into two coarse buckets so downstream tooling can treat them differently: - **binary:** the file looks like binary data. This is the end of the live path, so the file is committed to `data/binary/`, which doubles as the corpus for the offline stage-5 discovery sweep below. - **text:** the file looks like text. Stage 4 is still to come, so nothing moves; the file stays in `data/spool/` and its reference goes onto the stage-4 queue, ending up in `data/text_done/` once enriched. The test is a **UTF-8 sniff**: read the first 8000 bytes and call the file text when that window is valid UTF-8 and holds no control bytes outside the text-safe set (tab, newline, CR, and friends, plus ESC for ANSI-colored logs); otherwise binary. It's cheap (no full read) and Unicode-aware. Where the older NUL-byte and printable-ASCII heuristics misfiled non-ASCII text, this keeps accents, CJK and emoji in the text bucket, while binary formats, which rarely form valid UTF-8 near their start, still read as binary. A NUL byte is valid UTF-8 but not a text control byte, so it too reads as binary. A multi-byte character split by the 8000-byte boundary is trimmed before the check so it isn't mistaken for malformed bytes. An empty file is treated as text (`src/content.jl`). ### Language enrichment (stage 4) Files that stage 3 sorts as **text** are handed to a fourth pool that identifies their language and writes a `.meta.json` sidecar, mirroring the stage-2 known-file enrichment. Two detectors run per file: - **natural language:** [`Languages.jl`](https://github.com/JuliaText/Languages.jl)'s `LanguageDetector` (a Julia port of the `whatlang` n-gram model) reads a bounded prefix (up to `LANG_SAMPLE_BYTES`, 64 KiB) and reports the language's English name, ISO 639-3 code, and a confidence in `[0,1]`. Pure Julia, no subprocess. The detector is built once at startup and shared read-only across the pool. - **programming language:** the [`github-linguist`](https://github.com/github-linguist/linguist) CLI recognizes source and markup by extension + content heuristics (e.g. `Python`, `Markdown`). Plain prose reports as `Text` and unrecognized content as `null`; both collapse to *no programming language*. The sidecar schema: | field | meaning | |---|---| | `id`, `original_name` | from intake | | `file_size` | bytes (authoritative, from intake) | | `content_type` | always `"text"` | | `language` | natural-language English name (e.g. `English`), or `null` | | `language_code` | ISO 639-3 code (e.g. `eng`), or `null` | | `language_confidence` | detector confidence in `[0,1]`, or `null` | | `programming_language` | e.g. `Python`, `Markdown`, or `null` | | `error` | set if natural-language detection produced nothing | > **`github-linguist` and the git-repo quirk:** run against a path *inside* a git > repository, linguist reads the file's committed git blob, not the on-disk > bytes, and an untracked file (which everything under `data/` is) has no blob, > so it crashes. Stage 4 sidesteps this by copying each file to a fresh temp dir under > `/tmp` (outside any repo, preserving the name so extension heuristics still > fire) and pointing linguist there. > > Programming-language detection is best-effort: if `github-linguist` is > missing (a startup warning, not a fatal error, unlike `exiftool`), fails, or > times out (`FS_LINGUIST_TIMEOUT`, default 30s), `programming_language` is simply > `null` and the file still completes. Natural-language detection failing produces > a **degraded sidecar** (with an `error` note) rather than a quarantine, because > the file is still wanted text. Like stage 2, the sidecar is committed *before* the file is moved into `data/text_done/`, so the file's presence there always implies its sidecar is present; recovery re-enriches idempotently (`src/language.jl`). ### Unknown-format discovery (stage 5, offline) The `binary/` sink from stage 3 is the pile of genuinely *unrecognized* files. Stage 5 mines it for recurring new file formats by clustering files on their header bytes into a growing catalog of discovered formats, each with a magic-byte signature that can eventually be promoted into the classifier's fast path. Unlike stages 1–4 it is not on the request hot path: it is a single-owner *batch* process (the catalog is mutable shared state, the opposite of the stateless classifier), and because promotion is human-gated nothing here is latency-sensitive. The full rationale, and the assumptions we deliberately rejected, live in [`model/DESIGN_clustering.md`](model/DESIGN_clustering.md). The model (`src/cluster.jl`, base-Julia, no extra deps) is a Dirichlet-process mixture of **per-position categoricals** over the first 32 header bytes, on a 257-symbol alphabet (byte `0–255` plus a `past-EOF` symbol so short fixed-length formats are modeled honestly). Bytes are treated as categorical rather than numeric (`0x89` and `0x88` are not "close"), so this deliberately does *not* reuse the classifier's `[0,1]` byte scaling. A fixed uniform *background* component absorbs structureless (compressed/encrypted) blobs so they don't mint spurious clusters. A cluster's spiked positions become a libmagic-style signature; clusters with enough members and enough fixed positions self-**nominate** for promotion (a human does the one irreversible step, redefining "known"). **Status:** both phases are implemented and calibrated. Phase A (offline Gibbs) is the science; phase B (`src/catalog.jl`) is the live catalog: a durable single-owner state that sweeps `binary/`, folds each new file into a cluster with the deterministic CRP-predictive rule, and writes promotion nominations. The catalog process is run periodically (cron), single-threaded. It is the only writer of the catalog, so it needs no locking: ```bash julia --project=. bin/cluster_sweep.jl # incremental live sweep of new binary/ files julia --project=. bin/cluster_sweep.jl --compact # offline Gibbs re-cluster (seed / recompact) ``` The catalog is a single durable file (`FS_CLUSTER_CATALOG`, default `data/catalog.json`) committed with the same sidecar-first temp→fsync→rename→fsync-dir discipline as the stage-2 sidecars, so a crash can neither corrupt it nor lose a write. On the first run (empty catalog) the sweep auto-promotes to a `--compact` pass to seed clusters; later runs assign incrementally, touching only files they have not seen. A cluster that clears the member/magic thresholds writes a nomination (a hex magic template, member count, and example filenames) into `FS_NOMINATED_DIR` (default `data/nominated/`) for a human to glance at and promote. Under the calibrated `bg_mass > α`, the live sweep never mints single-file clusters; genuinely new formats surface from the periodic `--compact` re-clustering of the background residue, not the live path. Calibration is its own offline script, like training, and never in the request path. It is scored against magic-collapsed ground truth (so `docx`≡`zip` and the whole ELF family count as one format each, which is the *correct* answer, not an error): ```bash julia --project=. bin/cluster_calibrate.jl [training_set_dir] # defaults to ../training_set ``` It grid-tunes the hyperparameters to maximize Adjusted Rand Index against known formats and cross-checks against a model-free NCD (gzip) baseline. On the 700-file training corpus the calibrated defaults (`n=32`, `α=1.0`, `β=0.1`) recover the known formats at **ARI 0.77** (0.885 excluding tar), with `gzip`, `pkzip` (`docx`+`zip` merged), and `jpeg` forming clean, promotable clusters; the NCD baseline agrees. See `DESIGN_clustering.md` §11 for the full results, including the one known limitation: ELF and these tarballs share a long run of header zero-padding and merge. The v2 fix is inverse-entropy position weighting. ## Queue backends The HTTP handler and the workers only ever call `enqueue!`, `dequeue!`, `ack!`, `nack!` and `close!` on a `JobQueue` (`src/queue.jl`). Two implementations sit behind those five methods, chosen at startup by `FS_QUEUE_BACKEND`: | | `channel` (default) | `rabbitmq` | |---|---|---| | Where jobs live | in-process, bounded buffer | durable broker queues, persistent messages | | Crash recovery | everything in `spool/` replays from stage 1 | each file resumes at the stage it had reached | | External dependency | none | a RabbitMQ broker | | Capacity | hard limit, enforced per enqueue | advisory, checked against a polled depth | | Delivery | exactly once (nothing to redeliver) | at least once | `ack!` is the whole difference. A job is settled only after its handler commits, so a crash mid-enrichment leaves that job on the enrich queue and the restart picks it up there — not at stage 1, and not lost. `ChannelQueue` implements `ack!`/`nack!` as no-ops, which is honest rather than lazy: an in-process queue has no delivery to settle, and its recovery story is `recover_dir!`. ### Running it ```bash # broker + server, both in compose docker compose -f docker-compose.yml -f docker-compose.rabbitmq.yml up # just the broker, with the server on the host (5672 and the management UI on # 15672 are published for exactly this) docker compose -f docker-compose.yml -f docker-compose.rabbitmq.yml up -d rabbitmq FS_QUEUE_BACKEND=rabbitmq \ FS_AMQP_URL=amqp://fileserver:fileserver@localhost:5672/ \ julia --project=. -t auto bin/server.jl ``` Queues are named `.` — `fileserver.classify`, `.enrich`, `.triage`, `.language` — so the broker's queue list reads like the pipeline, and `/stats` reports each one's depth from a once-a-second poll. ### What it guarantees, and what it deliberately doesn't - **At least once, not exactly once.** Stages 1 and 3 publish downstream *then* ack upstream, so a crash in that window redelivers a job that was already routed. Safe for the same reason replay is: every stage is a pure function of the file's bytes and every commit is idempotent. The one new case is that a duplicate can run *concurrently* with the original and find the file already committed; `worker_loop` recognises a vanished `job.path` and settles it quietly, rather than quarantining a file that in fact succeeded. - **Publisher confirms on intake only.** A `202` means the broker has the file, because a client may delete its copy on the strength of it. That costs a round trip per file and serializes intake publishes; `FS_AMQP_CONFIRMS=false` turns it off. Inter-stage publishes are fire-and-forget by design — a lost one leaves its source job unacked, and redelivery repairs it for free, on a handoff that otherwise costs 0.12 µs. - **Advisory capacity.** `enqueue!` compares `FS_QUEUE_CAPACITY` against a depth polled once a second (and adjusted locally in between), so the existing `503` and `blocked_ns` behaviour still works, but a burst can overshoot by up to a poll interval. Making the limit real would mean `x-max-length` with `overflow: reject-publish`, which needs a confirm per message to detect. - **No reconnect.** A dropped connection invalidates every in-flight delivery tag, so reconnecting would silently reprocess whatever the workers were holding. Instead the server drains and exits non-zero, and the restart policy brings it back to redelivered messages and an intact `spool/`. - **No dead-letter queue.** A failed job is quarantined to `failed/` and then acked: the file's story and the message's story end in the same place. A DLQ would hold messages pointing at files that had already moved. - **One consumer process.** `Job` is a claim check carrying a *local* path, so a second server on another host would be handed jobs whose files it cannot see. The broker buys durable resume across restarts, not horizontal scale; scaling out would additionally need `spool/` on shared storage and an aggregated `/stats`. - **Restart recovery skips `spool/`.** The broker is the record of what is in flight, so re-driving `spool/` would duplicate the whole backlog on every restart. `FS_RECOVER_SPOOL=true` forces it, for the one case the broker cannot cover: it was purged or recreated and the spooled files are all that is left. ### Tests The pure parts (URL parsing, the wire format, backend selection, the duplicate branch) run in the normal suite. The round trip against a real broker is gated: ```bash docker compose -f docker-compose.yml -f docker-compose.rabbitmq.yml up -d rabbitmq FS_TEST_AMQP_URL=amqp://fileserver:fileserver@localhost:5672/ \ julia --project=. -e 'using Pkg; Pkg.test()' ``` Without `FS_TEST_AMQP_URL` those tests are skipped with a notice, so the suite still passes on a machine with no Docker. ## Running ```bash # install deps (first time) julia --project=. -e 'using Pkg; Pkg.instantiate()' # external tools: exiftool (stage 2, required) and github-linguist (stage 4, # optional, for programming-language detection). e.g. on Debian/Ubuntu: # apt install libimage-exiftool-perl # gem install github-linguist # start the server; -t sets the number of OS threads available to workers julia --project=. -t auto bin/server.jl ``` By default the pipeline's queues are in-process, and nothing external is needed. For durable queues that survive a crash, run against RabbitMQ instead — see "Queue backends" above. ## Shutdown Both SIGINT and SIGTERM trigger the same idempotent graceful drain (stop serving → close queue → wait for workers → exit): - **SIGINT** is caught as an `InterruptException` (we call `Base.exit_on_sigint(false)`), so shutdown is clean and quiet. One caveat specific to the RabbitMQ backend: Julia delivers SIGINT to whichever task happens to be running on thread 1, which is usually the main loop but is not guaranteed to be — AMQPClient runs receiver tasks of its own, and an interrupt that lands in one of those kills the broker connection instead of reaching the main loop. That is not a stuck server: the depth poller notices the closed connection within a second and asks for the same drain, so the process still stops (within ~4s, exiting non-zero, and logging it as a lost connection rather than an interrupt). **Under a process manager, prefer SIGTERM for the broker backend** — it goes through `atexit`, which has no such lottery. Shutdown on either signal also prints a few `Consumer ... task exiting` warnings from AMQPClient, which are cosmetic: they are its consumer tasks noticing that we cancelled them. - **SIGTERM** can't be intercepted directly: Julia blocks it on worker threads and handles it in its own runtime, so a user `signal()` handler never fires. Instead we hook the drain into an `atexit` handler, which Julia's SIGTERM path does run. Caveat: Julia prints its own `signal 15: Terminated` backtrace *before* `atexit` runs. It's harmless noise, and the drain still completes right after it, but for a fully quiet stop under a process manager, configure it to send SIGINT instead (systemd: `KillSignal=SIGINT`; Docker: `STOPSIGNAL SIGINT`). Give the stop timeout enough headroom to drain in-flight work (systemd: `TimeoutStopSec`). ## File classifier Each file is scored by a fixed-structure neural network (Lux.jl) that answers a single binary question: is this file **known** (like the types in the training set) or **unknown**? It's novelty detection, not exact file-typing: it won't tell you "PDF", just "this looks like something I was trained on, or not". - **Features:** the first 16 bytes + last 16 bytes of the file, each scaled 0–255 → `[0,1]`, giving a 32-dim input. Files under 32 bytes can't form that window and are classified `unknown` without touching the model. - **Architecture:** `Dense(32→64,relu) → Dense(64→16,relu) → Dense(16→2)`, raw logits; decision is `argmax` (class 1 = known, class 2 = unknown). - **Artifact:** trained weights live in `model/classifier.jld2` (committed), so the server just loads them at startup. Missing/unreadable ⇒ the server fails fast rather than run without classification. - **Effect today:** *active routing*. The class is logged (`classification=known|unknown`) and drives the pipeline split: `:known` files go onto the stage-2 queue for metadata enrichment, `:unknown` files onto the stage-3 queue for content triage. The class chooses the downstream stage (the file itself stays in `data/spool/` either way); what's still unproven is the model's *accuracy*, not whether the routing path runs. The architecture and byte→feature mapping are defined once in `src/model.jl` and shared by the trainer and the server, so they can't drift apart. ### Training Training is a separate, offline script. It never runs in the request path: ```bash julia --project=. bin/train.jl [negatives_dir] ``` - **positives_dir:** every file in it (≥32 bytes) is a "known" example. - **negatives_dir** *(optional)*: a grab-bag of *other* real file types used as "unknown" examples. Negatives are generated ~1:1 with positives, split 50/50 between uniform-random byte vectors and grab-bag files. With no grab-bag dir, negatives are all random (weaker: the net may just learn "high entropy = unknown" rather than your actual types, so a grab-bag of real off-distribution files is recommended). The script uses an 80/20 seeded split, reports validation accuracy, and writes `model/classifier.jld2` (path overridable via `FS_MODEL_PATH`). A fixed seed (`FS_TRAIN_SEED`, default 42) drives negative generation, the split, and weight init, so the artifact is exactly regenerable from the same inputs. ## Configuration (environment variables) | Variable | Default | Meaning | |---------------------|----------------|------------------------------------------| | `FS_HOST` | `127.0.0.1` | Bind address | | `FS_PORT` | `8080` | Port | | `FS_WORKERS` | `nthreads()` | Stage-1 (classification) worker tasks | | `FS_QUEUE_CAPACITY` | `1000` | Max pending intake jobs before `503` | | `FS_KNOWN_WORKERS` | `nthreads()` | Stage-2 (enrichment) worker tasks | | `FS_KNOWN_QUEUE_CAPACITY` | `1000` | Max pending enrichment jobs (then backpressure) | | `FS_UNKNOWN_WORKERS` | `nthreads()` | Stage-3 (content triage) worker tasks | | `FS_UNKNOWN_QUEUE_CAPACITY` | `1000` | Max pending triage jobs (then backpressure) | | `FS_TEXT_WORKERS` | `nthreads()` | Stage-4 (language enrichment) worker tasks | | `FS_TEXT_QUEUE_CAPACITY` | `1000` | Max pending language jobs (then backpressure) | | `FS_SPOOL_DIR` | `data/spool` | Every in-flight file, at every stage | | `FS_BINARY_DIR` | `data/binary` | Stage-3 sink: unknown files that look binary | | `FS_DONE_DIR` | `data/done` | Enriched known files (+ `.meta.json`) | | `FS_TEXT_DONE_DIR` | `data/text_done` | Enriched text files (+ `.meta.json`) | | `FS_FAILED_DIR` | `data/failed` | Files whose processing threw | | `FS_MODEL_PATH` | `model/classifier.jld2` | Classifier artifact loaded at startup | | `FS_UPLOAD_CHUNK_BYTES` | `65536` | Socket read size at intake; bounds intake memory per in-flight upload | | `FS_EXIFTOOL_TIMEOUT` | `30` | Seconds before a stuck exiftool is killed | | `FS_LINGUIST_TIMEOUT` | `30` | Seconds before a stuck github-linguist is killed | | `FS_CLUSTER_DIR` | `data/binary` | Stage-5 input: the unknown/binary pile to sweep | | `FS_CLUSTER_N` | `32` | Header bytes modeled per file | | `FS_CLUSTER_ALPHA` | `1.0` | CRP concentration (propensity to spawn new formats) | | `FS_CLUSTER_PSEUDOCOUNT` | `0.1` | Dirichlet pseudocount β (calibrated) | | `FS_CLUSTER_BG_MASS` | `5.0` | Mass of the uniform background component | | `FS_PROMOTE_MIN_MEMBERS` | `20` | Cluster size threshold for promotion nomination | | `FS_PROMOTE_MIN_MAGIC` | `3` | Required fixed signature positions to nominate | | `FS_CLUSTER_CATALOG` | `data/catalog.json` | Durable stage-5 catalog file (single-owner) | | `FS_NOMINATED_DIR` | `data/nominated` | One JSON per self-nominated cluster (human promote gate) | | `FS_QUEUE_BACKEND` | `channel` | `channel` (in-process) or `rabbitmq` (durable); see "Queue backends" | | `FS_AMQP_URL` | `amqp://guest:guest@localhost:5672/` | Broker connection, credentials included | | `FS_AMQP_PREFIX` | `fileserver` | Queue names are `.` | | `FS_AMQP_PREFETCH` | worker count | Unacked messages the broker hands one stage at a time | | `FS_AMQP_CONFIRMS` | `true` | Wait for a publisher confirm before a `202` (intake only) | | `FS_RECOVER_SPOOL` | `false` | Re-drive `spool/` at startup even on the broker backend (use after a purged broker) | > To get real parallelism, start Julia with enough threads (`-t N`) to cover all > pools. If `FS_WORKERS + FS_KNOWN_WORKERS + FS_UNKNOWN_WORKERS + FS_TEXT_WORKERS` > exceeds available threads you'll get a warning (non-fatal) and workers will > share threads. ## Usage ```bash # health check curl http://127.0.0.1:8080/health # {"status":"ok"} # upload one or more files (multipart/form-data) curl -F "a=@report.pdf" -F "b=@data.csv" http://127.0.0.1:8080/upload # 202 {"accepted":[{"id":"","name":"report.pdf"}, ...]} # per-stage counters curl http://127.0.0.1:8080/stats ``` Each file in a request becomes its own job. Responses: - `202 Accepted` — all files spooled and queued (with per-file job IDs) - `400 Bad Request` — not multipart, or no files present - `503 Service Unavailable` — queue full, retry later - `500 Internal Server Error` — failed to write a file to disk ### `GET /stats` — per-stage counters The pipeline counts its own work (`src/stats.jl`), because nothing outside it can. Every in-flight file sits in `data/spool/` no matter which stage it has reached. The stage is a property of the queue holding its reference, and only the pipeline can see that; there is no directory to poll. ```jsonc { "now": 1785725300.5, "since": 1785725291.9, "uptime_seconds": 8.5, "intake": { "requests": 400, "files": 400, "bytes": 6553600, "rejected": 0 }, "stages": [ { "stage": 4, "name": "language", "workers": 16, "queue_depth": 184, "queue_capacity": 1000, "completed": 200, "failed": 0, "bytes": 3276800, "busy_seconds": 170.5, // summed handler time across the pool "blocked_seconds": 0.0, // of that, time parked on a full downstream queue "in_flight": 3 } ] } ``` Counters are monotonic since startup, Prometheus-style. Rates are the reader's job, so a scrape is stateless and two readers can't disturb each other. Take two scrapes Δt apart and subtract: ``` throughput = Δcompleted / Δt utilization = (Δbusy_seconds − Δblocked_seconds) / (Δt × workers) ``` **Utilization is the number that names the bottleneck.** In a pipeline every stage completes the same files, so at steady state they all report near-identical files/s no matter which one is the constraint; what separates them is how hard each pool worked to keep up. The bottleneck sits near 1.0 with its queue backing up while its neighbours idle. `blocked_seconds` is what keeps that true. Stages 1 and 3 apply *blocking* backpressure (a full downstream queue means the handler parks and retries rather than dropping the file), and that wait happens inside the handler. Counting it as busy would pin stage 1 at 1.0 whenever stage 2 is the real jam, making every stage upstream of a jam look like the jam. Subtracted out, utilization means "doing its own work", and a high blocked share becomes its own signal: a stage blocked 90% of the time is naming its successor. The counters are a handful of atomic adds per file, recorded in `worker_loop`: the one place every stage's work passes through, so a new stage is instrumented the moment it is wired up, and never on the read path. ## Benchmarking (throughput + memory) There are five harnesses. Only the first needs a running server: | script | measures | server? | |---|---|---| | `bin/bench.jl` (below) | intake, end-to-end and per-stage throughput; server RSS | **yes** | | [`bin/bench_stage1.jl`](#stage-1-component-benchmark-binbench_stage1jl) | stage 1 taken apart: classify vs. rename vs. enqueue vs. logging | no | | [`bin/bench_stage2.jl`](#stage-2-component-benchmark-binbench_stage2jl) | stage 2 taken apart: exiftool spawn vs. extraction vs. commit | no | | [`bin/bench_model.jl`](#model-microbenchmark-binbench_modeljl) | the classifier alone: inference, feature reads, thread scaling | no | | [`bin/cluster_calibrate.jl`](#unknown-format-discovery-stage-5-offline) | stage-5 clustering quality vs. an NCD baseline | no | Running all of them from a clean checkout: ```bash julia --project=. -e 'using Pkg; Pkg.instantiate()' # once # 1. the model, on its own; no server involved julia --project=. -t auto bin/bench_model.jl # 1b. stage 1 taken apart; also no server julia --project=. -t auto bin/bench_stage1.jl # 1c. stage 2 taken apart; needs a directory of real files, not generated ones julia --project=. -t auto bin/bench_stage2.jl # 2. the pipeline. Start the server in one terminal… julia --project=. -t auto bin/server.jl # …and drive it from another. Restart the server between memory runs: Julia's # GC returns memory to the OS lazily, so a second run starts inflated. julia --project=. -t auto bin/bench.jl --files 2000 --size 8k --concurrency 32 julia --project=. -t auto bin/bench.jl --files 2 --size 1g --concurrency 1 # memory julia --project=. -t auto bin/bench.jl --corpus ../training_set --concurrency 16 # 3. stage-5 clustering quality (offline, needs a labelled corpus) julia --project=. bin/cluster_calibrate.jl ../training_set ``` The test suite is `julia --project=. -t auto -e 'using Pkg; Pkg.test()'`. `bin/bench.jl` must run on the same machine as the server (it reads the sink dirs and `/proc`); otherwise it only speaks HTTP: `/upload`, `/health`, and `/stats` for the per-stage numbers. Three properties of this design dictate how it measures: - **HTTP latency is not throughput.** `/upload` returns `202` once the bytes are spooled and a reference is enqueued, and all four stages run *after* the response, so `ab`/`hey`/`wrk` would only ever measure intake. The bench instead uploads a corpus and polls the terminal sinks (`done/`, `text_done/`, `binary/`, `failed/`) until the count stops moving, and reports both numbers separately: intake rate *and* end-to-end completion rate. It also samples `spool/`, whose peak depth is the high-water mark of files in flight. But *which* stage they are waiting on comes from `/stats`, not from disk. - **End-to-end throughput doesn't name the slow stage.** The four stages run concurrently behind their own queues, so the pipeline's rate *is* the slowest stage's rate and the others are invisible in it. The bench scrapes [`/stats`](#get-stats--per-stage-counters) before and after the run and subtracts, giving each stage its own throughput, mean service time and utilization: ``` PER-STAGE (server counters, delta over the end-to-end window) stage files/s MiB/s svc ms util blocked peak queue failed 1 classify 32.52 0.51 38.7 0.08 0.0% 209/1000 0 2 enrich 0.24 0.0 541.0 0.01 0.0% 0/1000 0 3 triage 32.27 0.5 33.9 0.07 0.0% 83/1000 0 4 language 16.26 0.25 852.6 0.87 0.0% 184/1000 0 bottleneck stage 4 (language) at 87.0% utilization of 16 worker(s) ``` (400 mixed files, 16 KiB each, concurrency 16. Stage 4 is the constraint, `github-linguist` being a process spawn per file at ~850 ms, and the 200 text files queue up behind it while stages 1 and 3 idle at under 10%.) Read *down* the util column, not the files/s column: each stage sees a different subset of the corpus, so a low rate can just mean little work was routed there. - **Memory should be flat in file size, and the sweep is what proves it.** Both halves of the pipeline are bounded: the workers read bounded prefixes (16+16 bytes to classify, 8 KB to sniff, 64 KB to language-detect), and intake streams each upload to disk a chunk at a time. So peak RSS should track *concurrency*, not size. Measured on this machine (16 threads, 64 KiB chunk): | upload size | concurrency | RSS growth | |---|---|---| | 256 MiB × 4 | 1 | 14.0 / 20.4 MiB (two runs) | | 1 GiB × 2 | 1 | 22.3 MiB | | 2 GiB × 1 | 1 | 31.3 MiB | | 256 MiB × 4 | 4 | none measurable (peak stayed under the baseline) | **Read the shape, not the digits.** Growth is flat across an 8× range of file sizes: 14–31 MiB whether the upload is 256 MiB or 2 GiB. That is the claim that matters, because nothing scales with file size, so nothing is buffering. The absolute figures are *not* precise to the megabyte. A freshly started server settles anywhere in an ~860–985 MiB band, so run-to-run variance in the baseline is comparable to the growth being measured; the residual is GC churn from the chunk reads, not retained buffers. Two consequences worth knowing before quoting these numbers: - **Let the server settle ~30s after startup** before a memory run, or the baseline is sampled mid-fall and the run reports less growth than it caused (or none at all, as the concurrency-4 row above did). - **Linear-in-concurrency is not currently demonstrated.** An earlier measurement put 4 concurrent 256 MiB uploads at 86.9 MiB (21.7 per in-flight upload), but that run predates a fix to how the baseline was sampled: it was read *before* the peak counter was reset, against a different origin. The re-measurement above cannot reproduce it, and the growth sits under the noise floor. Expect concurrency to cost memory; don't trust a specific coefficient without a quieter machine or many more runs. Before intake was streamed, the same 256 MiB upload grew RSS by ~700 MiB and 4 concurrent ones pushed a 950 MiB baseline past 2 GiB. So a `--size` sweep that slopes upward is the regression signal: something has started buffering bodies again. Flags: `--files`, `--size` (`8k`/`64m`/`1g`), `--concurrency`, `--kind` (`binary`/`text`/`mixed`, which chooses the stages that get loaded), `--corpus DIR` (real files, the only way to exercise stage 2's exiftool path), `--pid`, `--no-mem`, `--no-stats`, `--sample-ms`, `--timeout`, `--json PATH`. Full list in the script header. The `--json` output carries the per-stage numbers too, so a sweep can be compared run to run. Two caveats the script reports rather than hides: it counts sink *deltas*, so it warns if the pipeline isn't idle at the start (in-flight leftovers would be counted as its own throughput); and because Julia's GC returns memory to the OS lazily, a second run in the same process starts from an inflated baseline. It resets the kernel's peak-RSS counter (`/proc//clear_refs`) per run and flags a drifted baseline, but for a clean growth figure restart the server between memory runs. ### Stage-1 component benchmark (`bin/bench_stage1.jl`) `bin/bench.jl` reports stage 1 as one number and `bin/bench_model.jl` takes the *classifier* apart, but stage 1 is more than the model. Per file it also renames the file into its stage directory, pushes a reference onto the downstream queue, and logs. `bin/bench_stage1.jl` times each of those in isolation, then times the real `handle_classify_job` end to end so the parts can be checked against the whole: ```bash julia --project=. -t auto bin/bench_stage1.jl ``` Measured on this machine (Ryzen 7 2700X, 8 cores/16 threads, Julia 1.12; 2000 × 64 KiB files, minimum of 5 trials): | component | per file | share of the handler | |---|---|---| | `classify()` | 10.7 µs | 38% | | ↳ `read_features` | 7.9 µs | 28% | | ↳ `Lux.apply` | 2.3 µs | 8% | | `move_to` (rename) | 11.7 µs | 41% | | `enqueue_blocking!` | 0.12 µs | 0.4% | | per-file logging (disabled `@debug`) | 0.29 µs | 1% | | **`handle_classify_job`** | **28.3 µs** | 100% | **This benchmark is why stage 1's per-file log lines are `@debug` rather than `@info`.** As `@info` they cost ~71 µs of the handler's ~118 µs, about 6× the classifier and 6× the rename, and nearly all of it was `ConsoleLogger` *formatting* (~64 µs), not the `FlushLogger`'s per-message flush (~8 µs on top). Demoting them took stage 1 from 8.5k files/s to 35.3k files/s on a single worker, a 4.2× speedup for no algorithmic change. The script still prices a formatted line, so the cost of turning them back on is visible: running the handler under `JULIA_DEBUG=FileServer` measures 133 µs per file, a 4.7× slowdown. That is the trade. Per-file tracing is available when you want it and off by default, with `GET /stats` giving per-file observability that is counted, not formatted. What's left is evenly split between the rename and the classifier, and neither has an easy 2×. Two things worth knowing: - **The rename, not the model, is the single largest component** (11.7 µs), and it's a plain `mv` within one filesystem. Inside `classify`, the same pattern holds: 7.9 µs of the 10.7 µs is `read_features` (the `open`, the two reads and the `seek`) against 2.3 µs of actual inference. Stage 1 is now a filesystem-bound stage with a neural network attached, not the reverse. - **Stage 1 now peaks at ~4 workers.** With the logger removed from the hot path the sweep reads 35.0k/s at 1 worker, 66.6k/s at 2, **73.3k/s at 4**, then *falls back* to 60.1k/s at 8 and 53.3k/s at 16, because every worker renaming into the same two directories contends on the same directory inode. That ceiling coincides with the one `bin/bench_model.jl` finds for inference, so ~4 is the number from both directions: raising `FS_WORKERS` past it costs throughput. Reported times are the minimum over trials. Flags: `--files`, `--reps`, `--trials`, `--size`, `--dir`, `--model`, `--threads`, `--no-threads`, `--json PATH`. ### Stage-2 component benchmark (`bin/bench_stage2.jl`) Stage 2 is the one stage whose cost is dominated by something outside Julia entirely: it forks `exiftool`, a Perl program, once per file. `bin/bench.jl` reports the stage as a single throughput number, which can't distinguish "the extraction is slow" from "the *spawn* is slow", and those have opposite fixes. `bin/bench_stage2.jl` times each piece in isolation, then times the real `handle_known_job` end to end: ```bash julia --project=. -t auto bin/bench_stage2.jl ``` Two things make this benchmark different from the stage-1 one: - **The corpus must be real files.** exiftool's cost depends on what it finds; a file of random bytes bails out early and understates the stage by ~10×. The default corpus is `data/done`, files that already went through stage 2 on this machine. `--corpus PATH` points it elsewhere. - **It prices the alternatives to one-fork-per-file**, because if the fork dominates then the only fixes are to stop paying it per file. `exiftool (batched Nx)` runs the whole corpus through one process; `exiftool (-stay_open)` keeps one process alive and feeds it one file at a time over a pipe, which is the shape a streaming pipeline could actually adopt. Both are measured, not assumed. Measured on this machine (Ryzen 7 2700X, 8 cores/16 threads, Julia 1.12, exiftool 12.40; 150 real files / 102 MiB, minimum of 2 trials): | component | per file | share of the handler | |---|---|---| | `run_exiftool()` | 135.9 ms | 98% | | ↳ bare fork + Perl boot (`exiftool -ver`) | 76.7 ms | 55% | | ↳ `JSON3.read` + tag map | 10 µs | 0.0% | | `normalize_metadata` | 1.7 µs | 0.0% | | `commit_enriched!` (sidecar + fsyncs + rename) | 2.0 ms | 1.5% | | per-file logging (`@info`, flush→file) | 47 µs | 0.0% | | **`handle_known_job`** | **138.3 ms** | 100% | | *alt:* `exiftool -stay_open` | 40.1 ms | 29% | | *alt:* `exiftool` batched 150× | 37.3 ms | 27% | **Stage 2 is exiftool and nothing else.** Everything the Julia code does (parsing, normalizing, the durable sidecar-first commit, the log line) sums to about 1.5% of the stage. There is no point optimizing any of it. **More than half the stage is interpreter startup, not metadata extraction.** The bare `exiftool -ver` (fork, Perl boot, module loads, read no file) costs 76.7 ms against a 135.9 ms full call. Both fork-free alternatives agree on what's left: ~37–40 ms of actual work per file. So a persistent exiftool would cut the stage by ~70%, and `-stay_open` gets there without giving up the one-file-in, one-result-out shape the pipeline needs. That remains the single biggest available win in this stage; it is measured here but not yet implemented. **This benchmark is also why `run_with_timeout` no longer polls.** The original watchdog polled with `sleep(0.1)` and then joined the polling task, so every call paid the remainder of an in-flight sleep *after* the child had already exited: ~25 ms per file here, and a measured 101 ms on a process that exits instantly. Replacing it with a one-shot `Timer` took the stage from 164.6 ms to 138.3 ms per file (6→8 files/s on one worker) and cost nothing in behavior. Stage 4 shares the wrapper and got the same fix for free. Writing the missing tests for that wrapper turned up a second, worse problem: **the timeout was never enforceable.** `wait(proc)` returns only once the captured stdout pipe closes, and any grandchild inherits that pipe, so signalling the child alone left the worker blocked until the whole process tree finished on its own (a `sh -c "trap '' TERM; sleep 30"` child ran the full 30 s against a 1 s timeout, under both the old and new watchdog). The child now runs in its own process group and the timeout signals the group. The trade is that a hard crash of the server orphans an in-flight child rather than taking it down; these children are short-lived and timeout-bounded, which is the cheaper side of it. **Stage 2 scales to ~8 workers, then flattens**: 8 files/s at 1 worker, 14 at 2, 28 at 4, **49 at 8**, and 49 at 16. The machine runs out of cores to run Perl on, which is exactly what you'd expect of a stage that is ~100% subprocess. Note that the sweep pulls from a shared counter rather than splitting the corpus into contiguous slices: per-file exiftool time spans two orders of magnitude on a real corpus (one 2.1 s archive among 48 files), and a static split reports a scaling ceiling that is really just load imbalance. One caveat the numbers raise but don't answer: **`fsync_dir` measures 1.75 µs**, which is far too fast to be a real disk flush. The durability that `commit_enriched!` is written for may not survive power loss on this filesystem, even though the code is correct. That's a correctness question, not a speed one, and it is not yet resolved. Reported times are the minimum over trials. Flags: `--files`, `--reps`, `--trials`, `--corpus`, `--dir`, `--timeout`, `--threads`, `--no-threads`, `--no-stay-open`, `--json PATH`. ### Model microbenchmark (`bin/bench_model.jl`) `bin/bench.jl` reports stage 1 as a single number: the wall time of `handle_classify_job`, which is a feature read, an inference, a rename, a (disabled) debug line, and whatever contention the other three pools create. That's the right number for capacity planning and the wrong one for "is the model slow?". `bin/bench_model.jl` answers that separately, with no server, queue, or HTTP involved: ```bash julia --project=. -t auto bin/bench_model.jl ``` Measured on this machine (Ryzen 7 2700X, 8 cores/16 threads, Julia 1.12): | what | per file | notes | |---|---|---| | `Lux.apply`, batch 1 | **2.3 µs** | 768 B allocated per call | | `read_features` | **4.2–6.0 µs** | flat across a 262,144× size range (1 KiB → 256 MiB) | | `classify()` | **8.4 µs** | 73% feature read, 27% inference | So the model is not the pipeline's problem, by three orders of magnitude: the same run measured stage 1 at 38.7 ms per file, ~4,500× the 8.4 µs `classify()` costs. Whatever stage 1 spends its time on, it isn't the network. (That 38.7 ms predates the `@debug` demotion above and is a whole-pipeline figure. It includes time the stage-1 worker spends *blocked* on a full downstream queue, which is why it is three orders of magnitude above the 28.3 µs the handler costs in isolation. For the uncontended split, see [the stage-1 decomposition](#stage-1-component-benchmark-binbench_stage1jl).) Two findings worth acting on if stage 1 ever *does* become the constraint: - **Batching would buy ~13×.** A 32×1 matmul wastes most of a BLAS call: batch 64 costs 280 ns/file and batch 512 costs 179 ns/file, against 2.34 µs one at a time. The pipeline classifies strictly one file per job today, so it pays the worst row in that table. - **Inference does not scale past ~4 threads.** Concurrent `Lux.apply` on the shared read-only `Classifier` peaks around 1.2M files/s at 4 tasks and then *falls back* to single-thread throughput at 16. The script runs a pure-compute control kernel through the same sweep to place the blame: the control reaches 14.3× at 16 tasks (90% efficiency) on the same box, so the machine parallelizes and `Lux.apply` doesn't. GC is only ~1% of it, so allocation pressure isn't the explanation either. The cause is inside Lux/BLAS and is not diagnosed here. The practical consequence: raising `FS_WORKERS` past ~4 adds no classification throughput. Reported times are the minimum over trials, because for a microbenchmark the floor is the signal and everything above it is scheduler and GC noise. Every timed loop stores its result in a sink so a pure call can't be hoisted out. Flags: `--model`, `--reps`, `--trials`, `--batches`, `--sizes`, `--no-threads`, `--json PATH`. ## Layout ``` src/ FileServer.jl module + run() (startup, recovery, workers, serve, shutdown) config.jl Config struct + env parsing job.jl Job (the queue reference) queue.jl JobQueue seam + in-process ChannelQueue rabbit.jl RabbitMQ-backed JobQueue: durable queues, acks, confirms stats.jl per-stage counters behind GET /stats (throughput, utilization) multipart.jl streaming multipart/form-data reader (intake never buffers a file) spool.jl filename sanitizing, streaming spool/move, startup recovery model.jl NN architecture + byte→feature mapping (shared with trainer) classify.jl load artifact + classify a file at inference time metadata.jl exiftool extraction + normalized sidecar (stage 2) content.jl binary-vs-text sniff for unknown files (stage 3) language.jl natural + programming language enrichment for text (stage 4) cluster.jl header-byte clustering model + Gibbs + scoring core (stage 5, science) catalog.jl durable single-owner format catalog + sweep + nominations (stage 5, phase B) worker.jl parametrized worker loop + classify/enrich/triage/language handlers server.jl HTTP routes + the streaming /upload handler docker-compose.yml the server, in-process queues docker-compose.rabbitmq.yml overlay: adds the broker and switches the backend bin/ server.jl entry point bench.jl throughput + memory harness against a running server bench_model.jl classifier microbenchmark (inference, feature reads, scaling) bench_stage1.jl stage-1 decomposition (classify vs. rename vs. enqueue vs. logging) train.jl offline training script → model/classifier.jld2 cluster_calibrate.jl offline stage-5 hyperparameter calibration + NCD baseline cluster_sweep.jl stage-5 phase-B runner: sweep binary/, update catalog, write nominations model/ classifier.jld2 committed trained weights (loaded at startup) DESIGN_clustering.md stage-5 design rationale + calibration results ```