Files
samiandClaude Sonnet 5 60363a7142 proto: land B4 and B2a — buffer bounds, budget knobs, effective readback (1.1.0)
Minor bump on 1.0.0, per core/docs/buffer-sizing.md.

B4 — bufferBytes bounds corrected in all four locations (DownloadSpec,
TaskDetail, download.update's patch, Settings.connection.bufferBytes): was
4 KiB-8 MiB with no stated default, now 64 KiB-16 MiB with a 1 MiB default.
64 KiB because 4 KiB is smaller than one libcurl HTTP/2 write-callback delivery;
16 MiB because throughput from write size is flat past ~1-4 MiB and past 16 MiB
there is stall-cover left to buy but no memory left to spend it on; 1 MiB
default because it is the only candidate for which docs/04's 60 MB RSS target
actually holds once buffers are counted per segment, not per download.

Two new settings keys: connection.maxTotalBufferBytes (128 MiB default) and
connection.maxActiveSegments (32 default). Without them CORE's clamp — reduce
every live segment's buffer to fit the global cap — has no wire configuration
surface, and "20 active downloads" has no meaning distinct from 160 live TLS
connections.

B2a — TaskDetail.effectiveBufferBytes: what a segment is actually using right
now, after the clamp. Placed on TaskDetail next to bufferBytes, following the
requested/effective pattern ADR 0010 already established for segments. The
download.get fixture now demonstrates a real clamp (16 MiB requested, 4 MiB
effective) rather than a case where the cap happens not to bind.

docs/04-engine-design.md §4 and §8 updated in the same change per CORE's
request and CLAUDE.md rule 5: the RSS target is now stated as conditional on
maxActiveSegments = 32, and the old 4 MiB/64 MiB/256 MiB numbers are corrected
to match the schema. ADR 0012 records the reasoning and explicitly keeps the
60 MB target over CORE's offered 120 MB alternative, with the arithmetic that
makes 60 MB achievable with margin.

Numbered 0012 rather than 0011: DAEMON is independently drafting ADR 0011
(admission control / segment budget split) in a peer session at time of
writing, so 0011 was reserved to avoid a collision.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_012fgjnqFCS5h5L7gZTZo3rV
2026-09-09 23:20:58 +04:00
..

contracts/ — the wire contract

This directory is the interface between every lane. Owner: agent PROTO. Nobody else commits here. Everybody else generates from here.

Status: v1.0.0 — FROZEN (2026-09-09)

v1.0.0 froze 38 methods, 9 events, 26 named types. v1.1.0 (current) is a minor bump on top of it: bufferBytes bounds widened to 64 KiB - 16 MiB across all four locations, two new settings keys (connection.maxTotalBufferBytes, connection.maxActiveSegments), and TaskDetail.effectiveBufferBytes — see docs/adr/0012-buffer-and-segment-budget.md. See also docs/adr/0005-... for the versioning rule and docs/adr/0010-... for the failure taxonomy and segment ranges.

Lane requests are answered in writing: contracts/proto-answers-m1.md responds to core/docs/proto-requests-m1.md point by point.

What each lane can rely on, starting now:

You need It is here
C++ types, parsing, dispatch core/generated/velox_proto.{hpp,cpp} (target libveloxproto)
TypeScript types, typed client, runtime validators extension/src/shared/protocol/
The API document to read contracts/openrpc.json
A daemon to build against today tools/mockd — both transports, 4 Hz progress, unhappy-path flags
Proof you have not drifted ./tests/conformance/run.sh

Changing this is a PR to contracts/ alone. Optional field or new method → minor. Rename, remove or retype → major, plus an ADR. File a request; do not add a field locally.

contracts/
├── VERSION                     # protocol semver — frozen at 1.0.0
├── openrpc.json                # human-readable API doc (generated from schema/)
├── schema/
│   ├── envelope.schema.json    # JSON-RPC 2.0 envelope + our error codes
│   ├── types/                  # Task, Segment, Category, Queue, Settings, CaptureRules…
│   ├── methods/                # one file per method: params + result
│   └── events/                 # one file per server→client notification
├── fixtures/                   # golden request/response pairs, replayed by conformance
└── codegen/
    ├── schema_ir.py            # the one loader/IR all generators share
    ├── gen_cpp.py              # → core/generated/  (structs + to_json + parse + dispatch)
    ├── gen_ts.py               # → extension/src/shared/protocol/ (types, client, validators)
    ├── gen_openrpc.py          # → contracts/openrpc.json
    └── gen_cpp_conformance.py  # → tests/conformance/cpp/fixture_dispatcher.hpp

Each subdirectory has its own README: codegen/ documents the supported JSON Schema subset, fixtures/ documents the fixture shape and the placeholder rules.

Rules

  1. Generated code is committed. No lane may be blocked because it can't run Python.
  2. Hand-editing generated files is a merge blocker. Fix the schema and regenerate.
  3. Every method needs at least one fixture — a success case and, where meaningful, an error case. A method with no fixture is not done.
  4. Versioning: adding an optional field or a new method → minor bump. Removing, renaming, retyping, or changing a default → major bump and a written migration note in docs/adr/. session.hello rejects a major mismatch with error -32001 and a message the GUI renders as "Velox needs updating".
  5. Changes arrive as a PR to contracts/ alone, containing: schema edit + fixtures + regenerated code + VERSION bump. Lanes rebase onto it. This is the only synchronization point in the whole project — keep it cheap and frequent rather than big and rare.

Per-method annotations

Every method schema carries these, and both generators emit them as data the code can act on rather than as prose a reader has to honour:

Key Meaning
x-privileged refused over the WebSocket transport with -32003
x-transports which listeners serve it (uds, ws)
x-deadlineMs how long a client waits before giving up
x-errors the error codes this method is documented to return
x-wsRestrictions extra limits when the call arrives from the extension

19 of the 38 methods are privileged: everything that reconfigures the daemon, destroys user data, or names an arbitrary destination path. The extension may request a download; it may not choose where the bytes land.

Transport framing

Client Transport Framing
GUI, CLI $XDG_RUNTIME_DIR/velox/velox.sock newline-delimited JSON (NDJSON)
nmhost ← Firefox stdio 4-byte little-endian length prefix (Firefox's format)
nmhost → daemon same Unix socket NDJSON
Extension (fallback) ws://127.0.0.1:520xx one JSON message per WS text frame

All four carry the same JSON-RPC 2.0 payloads. The framing differences stop at the transport layer; no method behaves differently depending on how it arrived — except that methods marked "privileged": true in the schema are refused over the WebSocket transport.

Method surface (v1.0.0 target — expand only via PR)

Session

Method Params → Result
session.hello {clientType, clientName, protocolVersion, token?}{daemonVersion, protocolVersion, capabilities[], sessionId}
session.pair {clientName, extensionId}{token, expiresAt} (WS only; triggers user prompt)
session.subscribe {events[]}{ok}

Downloads

Method Params → Result
download.probe {url, headers?, cookies?, referrer?, userAgent?}{filename, sizeBytes?, mime, resumable, effectiveUrl, suggestedCategoryId}
download.add {url, headers?, cookies?, referrer?, userAgent?, filename?, saveDir?, categoryId?, segments?, bufferBytes?, startMode:"now"|"later"|"queue", queueId?, description?, checksum?}{taskId, state}
download.addBatch {items[], defaults}{taskIds[]}
download.list {filter?, sort?, offset?, limit?}{total, items: TaskSummary[]}
download.get {taskId}TaskDetail (includes segments[])
download.start | .pause | .resume | .cancel {taskIds[]}{updated[]}
download.remove {taskIds[], deleteFile:bool}{removed[]}
download.update {taskId, patch:{filename?, saveDir?, categoryId?, queueId?, description?, segments?, bufferBytes?}}TaskSummary
download.refreshUrl {taskId, url, headers?}{ok} (IDM's "Refresh Download Address")

Organisation

category.list · category.upsert · category.remove · queue.list · queue.upsert · queue.start · queue.stop · queue.reorder · rules.list · rules.upsert · schedule.get · schedule.set

Settings & limits

settings.get {keys?} · settings.set {values} · limiter.get · limiter.set {globalBps?, enabled}

Browser integration

Method Notes
capture.offer {url, method, headers, cookies, contentType?, contentLength?, contentDisposition?, tabUrl, filename?}{action:"take"|"ignore", taskId?, reason?}must answer within 750 ms; the extension gives up and lets Firefox handle it otherwise
capture.getRules Extension mirrors the daemon's monitored types so the two never disagree
media.listVariants {manifestUrl, headers}{variants:[{id,resolution,bitrate,codec,sizeEstimate}]}
media.addVariant {manifestUrl, variantId, ...addParams}{taskId}

Grabber

grabber.start {startUrl, depth, includePatterns[], excludePatterns[], fileTypes[]}{jobId}; grabber.status {jobId}; grabber.harvest {jobId, select[]}{taskIds[]}

Events (server → client notifications)

Event Payload
event.task.added / .removed {taskId, summary?}
event.task.state {taskId, state, error?}
event.task.progress Batched array, emitted at ≤4 Hz: [{taskId, downloaded, speedBps, etaSec, segments:[{i,completed,speedBps}]}]
event.speed.global {downBps, activeCount}
event.auth.required {taskId, host, realm, scheme}
event.notify {level, title, body, taskId?}
event.settings.changed {keys[]}
event.grabber.progress {jobId, found, crawled, done}

Two error spaces, and why they are not the same

This trips people up, so it is stated once, loudly:

ErrorCode TaskErrorCode
Says why a call failed why a download failed
Space JSON-RPC integers (-32xxx) strings ("server_file_changed")
Lives in the JSON-RPC envelope's error TaskError.code, on a task
Example -32602 — your params were malformed checksum_mismatch — the bytes arrived and were wrong

A download fails while every RPC involved succeeds. That is the normal case. Never put a -32xxx into a TaskError, and never invent a JSON-RPC code for a transfer failure.

TaskErrorCode's 27 values mirror vdm::Error in core/include/vdm/util/error.hpp by name, so DAEMON's projection from the engine taxonomy is lossless and a new engine failure that has no wire spelling is a visible hole rather than a silent collapse to internal.

Segment ranges are inclusive

Segment.startByte and Segment.endByte describe a closed range [startByte, endByte]: endByte is the last byte, not one past it, and the segment covers endByte - startByte + 1 bytes. The two fields are copied verbatim into Range: bytes=<startByte>-<endByte>, which RFC 9110 defines as inclusive, so there is no arithmetic between the wire and the socket and nowhere for an off-by-one to hide. Conformance enforces contiguity and full coverage; a fixture written half-open fails.

Requested is not effective

DownloadSpec.segments is what a client asked for. TaskSummary.segments is what is in use right now, after the per-host cap and after the demotion to 1 for a non-resumable source. They are routinely different and the GUI must render the effective one.

Error codes

Code Meaning
-32600/-32601/-32602/-32603 Standard JSON-RPC
-32001 Protocol major version mismatch
-32002 Not paired / invalid token
-32003 Method not permitted on this transport
-32010 Task not found
-32011 Invalid destination path (outside allowed roots, or not writable)
-32012 Disk full
-32013 Probe failed (with data.httpStatus)
-32014 Rate limited (pairing brute-force lockout)