feat(srt): SRT relay addresses with a sticky id (srt+relay://) by srperens · Pull Request #629 · Eyevinn/open-live

srperens · 2026-11-11T16:57:00Z

Needs Strom 0.8.2 (Eyevinn/strom#1260 capabilities, #1261, #1262 named SRT relays, all merged). Merge together with the studio, Eyevinn/open-live-studio#259. Also covers #628.

What

SRT sources and outputs can be reached through an SRT relay (srt-splice) on one public UDP port, instead of a listener port per source or output. Strom dials the relay from inside, and the outside party dials it with a streamid. Strom may have several relays, each with its own name.

Address scheme srt+relay://<id>[@<relay>][?query] in SourceDoc.address (srt/efp) and OutputDoc.url (mpegtssrt/efpsrt). It is passed to Strom's srt_uri unchanged, and Strom resolves it.
- @<relay> names one of the relays Strom reports. Names are [a-z0-9-], 1–32 characters, and are validated on save. Without @ the address uses the relay default, as in Strom, and Open Live writes the default relay without @.
Sticky id:
- srt+relay://auto[@<relay>] means "generate an id for me", like srt://:0 for ports. Open Live generates a 128-bit id once and stores it.
- Activation, regeneration and a Strom redeploy never change the id. A PATCH with auto keeps the stored id. Open Live owns the id because it creates a new Strom flow on every activation.
- auto without @ also keeps the stored relay, so renewing the passphrase (srt+relay://auto?passphrase=…) never moves the stream to another relay. auto@<relay> moves the stored id to that relay, and auto@default moves it back to the default relay.
Passphrase on an auto PATCH of a document that already holds a relay address:
- no passphrase key: the stored passphrase (still encrypted, not re-encrypted) and the stored pbkeylen are kept. A relay switch therefore keeps the passphrase: it protects the stream end to end and the relay never sees it, so there is no reason to make the operator type it again.
- passphrase= (empty): the passphrase is removed. An empty passphrase= is never stored or sent to Strom.
- passphrase=<value>: sets it, as before.
Rules for ids:
- An explicit id needs 16–64 characters. Shorter ids are refused because they can be guessed.
- Ids are unique across sources and outputs, and across relays. A duplicate gets 409. This is checked before the write and again after it, on POST and PATCH.
No port reservation for relay addresses. The legacy port broker is unchanged.
Strom capabilities come from Strom GET /api/capabilities. It replaces GET /api/srt-relay, which Strom removed before it was ever released. Open Live caches the answer and refreshes it in the background, so a request never waits on Strom (30 s fresh, 5 s after a failure, warmed at startup).
- state is ok, unsupported, unreachable or unknown.
- unsupported means a 404 from an older Strom. Every mode is offered as before, and there is no relay.
- unreachable keeps the last known values.
- Open Live keeps no relay addresses or env vars of its own, because Strom knows where its relays are.

API: GET /api/v1/server-info now has srt. It replaces the unreleased srtRelay.

"srt": {
  "state": "ok",
  "modes": { "listener": false, "caller": true, "rendezvous": false },
  "relays": [ { "name": "default", "public": "srt.example.com:9000" }, { "name": "eu-north", "public": "srt-eu.example.com:9000" } ],
  "inboundReachable": false,
  "ports": [ { "first": 47100, "last": 47199 } ]
}

The studio offers only the directions whose mode is true (Only offer the SRT directions the Strom instance supports #628), and offers "Via relay" only when relays is non-empty, with a relay picker when there is more than one.
The relays' internal addresses are never exposed.

API: connect (the dial-in address) is on SRT sources and outputs.
- For a relay address it is srt://<that relay's public>?streamid=<id>. It never includes the passphrase, and it has relay: "<name>".
- reason (only when uri is null) can be srt-relay-not-configured, srt-relay-unknown (Strom does not report that name), srt-relay-unreachable or srt-relay-public-unknown.
- New warning field: srt-listener-unsupported or srt-rendezvous-unsupported, when Strom reports that the mode the address needs cannot work.
SRT modes are reported, never refused. The operator may know better. A listener address on a Strom with listener: false is still stored, and it gets the warning.
- Strom's own host for listener addresses is resolved as before: authored host, then SRT_PUBLIC_HOST, then the STROM_URL host. Strom deliberately reports no public host.
Relay refusals: a relay address gets 422 when Strom definitely does not have the relay it names. The body is { error, statusCode, code, relays }.
- code: "srt-relay-not-configured" when Strom has no relays at all, or is an older Strom.
- code: "srt-relay-unknown" when Strom has no relay of that name. relays lists the names Strom has.
- While Strom cannot be asked, relay addresses are accepted, and connect.reason says why the dial-in address is unknown.
One resolver (resolveSrtAddress, usedSrtBindings, writeWithSrtAddress) handles listener ports and relay ids in all four handlers.
SRT addresses are checked at save the way GStreamer reads them (lib/srt-uri-check.ts, a port of Strom's validate_srt_uri), on create and on a PATCH that sends address/url. These get a 400 that names the right spelling:
- a key or mode in the wrong case
- a key given twice
- a key without = (new: ?latency crashes GStreamer 1.28.6, and Strom now refuses it)
- a malformed escape or %00
- a raw #
- an unknown mode
- a pbkeylen other than 0/16/24/32
A stored address is not checked again, so renaming an older document keeps working.
Passphrase limit is 10–80 bytes (libsrt), measured decoded. ! * ' ( ) ~ are accepted in the query.
Separate fix (own commit, cbf8152): SRT output passphrases were sent to Strom still encrypted (encv1:...), because outputs, unlike sources, were not decrypted before flow generation.
docs/openapi.yaml documents all of the above. ADR-014 and the Only offer the SRT directions the Strom instance supports #628 spec have a dated note on how Strom's API landed.

Strom configuration (for reference)

STROM_SRT_RELAY is now STROM_SRT_RELAY_PUBLIC (plus STROM_SRT_RELAY_INTERNAL, and STROM_SRT_RELAYS=name=public[|internal],... for named relays). STROM_INBOUND_REACHABLE=false turns off listener and rendezvous in modes. See Strom docs/SRT_RELAY.md.

Verified

2026-11-13, after the passphrase-keep change: tsc --noEmit passes, vitest passes: 184 files, 2233 tests (keep on switch, remove with empty value, set new, same-relay no-op, route tests with real encryption).
Earlier the same day, after the move to capabilities: pnpm build passes.
- The tests cover capabilities parsing (404, empty relays, several relays, malformed entries), the cache states, @relay parse and write, dial-in per relay, sticky renewal keeping the relay, the 422 codes, mode warnings and the server-info.srt shape.
- The repo has no lint script, so lint was not run.
Against the merged Strom (#1262): the local stack (Open Live, studio, native Strom with two relays) answers server-info.srt with both relays. Media through Strom relay blocks was verified end to end on both relays (ffmpeg → relay → Strom input → encoder → Strom output → relay → ffprobe, encrypted H.264/AAC), driven through the Strom API, not through Open Live.
Checked by hand through the studio against this branch and the merged Strom.

Known limitations

Relay ids are unique only within one Open Live instance. The check reads only this instance's database. Two Open Live instances that share one Strom and relay could store the same explicit id, and the relay would then pair the wrong callers. Generated (auto) ids are 128-bit random, so this needs two explicit ids chosen alike. It is unlikely in practice, and it is not handled.
If a stored relay is later removed from Strom, the stored address shows srt-relay-unknown. A passphrase renewal with auto on that document is then refused with 422 until another relay is chosen (auto@<name>).

Strom 0.8.2: feat(api): GET /api/capabilities and inbound reachability strom#1260 (GET /api/capabilities), #1261 (caller output queue), #1262 (named SRT relays; supersedes #1258's GET /api/srt-relay)
Studio: the "Via relay" choice, relay picker and mode-aware direction picker (feat(studio): SRT "Via relay" and dial-in addresses from the backend open-live-studio#259)
Only offer the SRT directions the Strom instance supports #628: offer only the SRT directions Strom supports (covered here on the backend: modes are reported in server-info.srt.modes and connect.warning)

🤖 Generated with Claude Code

eyevinn-developer · 2026-11-11T17:27:24Z

This PR looks like part of a multi-PR dependency stack together with, in dependency order: Eyevinn/strom#1258 (Strom resolves srt+relay://), #629 (backend), Eyevinn/open-live-studio#259 (studio). Each PR body says it must not merge until the others are ready together, and the open-live and studio PRs are drafts. Holding off on automated review/merge here — this needs a human (or the relevant dev agent) to assess the whole stack together, not a per-PR pass. Nothing further is needed from the author; this comment exists so the dependency is recorded on the PR rather than in a run log.

Related: #628 (spec merged in #630) builds on this stack's relay-status cache, and its implementation waits for #629 to merge.

… Strom SRT output URLs are stored with the passphrase encrypted at rest (#260), like source addresses, but only source addresses were decrypted in activateStromFlow. Output URLs went into srt_uri raw, so Strom got the encv1: ciphertext as the passphrase and the receiver could never decrypt the stream. Decrypt a copy of each mpegtssrt/efpsrt output URL before the output blocks are built. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Sources (srt/efp) and SRT outputs (mpegtssrt/efpsrt) accept srt+relay://<id>[?params]: a session on the SRT relay (srt-splice) that Strom fronts and resolves itself. Open Live never learns where the relay is; it owns the id and the passphrase. - srt+relay://auto asks the server for an id, like srt://:0 asks for a port: a random 128-bit base64url id is generated once and stored. A patch with auto keeps the id the doc already has, so the address the outside party dials survives activation, restarts, flow regeneration and Strom redeploys. An explicit id is stored as given. - srtUrl() validates the relay form: id [A-Za-z0-9_-]{1,64}, the same query charset and 10-79 passphrase rule, and rejects mode, streamid, localaddress and localport (Strom sets them when it calls the relay). The passphrase is encrypted at rest as for any SRT address. - Relay addresses skip listener-port reservation and clash checks, and are never declared as flow ports. Flow generation passes the (decrypted) relay URI to srt_uri unchanged. - StromClient.system.srtRelay() reads GET /api/srt-relay; the result is cached ~30 s (404 from an older Strom = not configured; other failures serve the last good answer). - connect: outputs' dial-in address resolves relay addresses to srt://<relay public>?streamid=<id> (source srt-relay, mode caller, streamid), or uri null with reason srt-relay-not-configured / srt-relay-public-unknown. port is now nullable. Sources get the same connect field. Strom is only asked when a relay doc is in the reply. - GET /api/v1/server-info adds srtRelay: { configured, public }. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…icit ids A relay id is the streamid both callers use on the relay, which pairs exactly two callers per id, so an id may be held by one document only. Sources and outputs share one namespace. - Create and patch refuse an id another source or output holds with 409 and a message naming that document. Like listener ports, create checks again after the write: a clash undoes the write, then a generated (auto) id is regenerated and retried, an explicit one is refused. A doc's own id never counts against it. - Explicit ids must be [A-Za-z0-9_-]{16,64}; a shorter one is a 400 that says why (anyone who guesses the id can connect to the public relay port). auto is unchanged and generates 22 characters. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…y status Code-review follow-ups on the SRT relay branch: - One resolver, resolveSrtAddress() (next to resolveListenerAddress), and one scan, usedSrtBindings(), for listener ports and relay ids alike; the resolve/write/re-check/undo loop is shared by sources and outputs, create and update (services/srt-address-write.ts). relay-ids.ts is folded in. - PATCH resolves and checks the EFFECTIVE address whenever the address or the stream type changes, so a whip -> srt patch can no longer store a literal `auto` relay id or one another document holds. - PATCH re-checks after the write like POST, and undoes the write (restoring the previous document) on a clash. - The relay status is stale-while-revalidate: requests are served from cache and refreshed in the background, the cache is warmed at startup, and only a lookup with nothing cached waits (at most 3 s). - Strom unreachable is no longer "not configured": the status keeps the last known values with state unreachable/unknown; server-info carries `state`, and connect.reason says srt-relay-unreachable. - Create/patch to a relay address is refused with 422 when Strom positively reports no relay; accepted while Strom cannot be asked. - srtUrl() matches the srt+relay:// scheme case-insensitively, like the parser and Strom. - srt-connect uses one host:port parser for authored URIs and the relay's public address. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

GStreamer's SRT elements in Strom match query keys and mode values exactly and drop most problems silently: `PassPhrase=` or `mode=Caller` is ignored (an address that looks encrypted runs unencrypted), a bad escape or %00 sets nothing, and a raw '#' cuts the URI. Strom now refuses such a URI when it builds the flow, so Open Live accepted it on save and the production failed only at activation. - lib/srt-uri-check.ts: checkSrtUriAsGStreamerReads, a port of Strom's validate_srt_uri (types/src/srt_uri.rs): scheme exactly srt:// (or srt+relay://), no raw '#', no malformed escape or %00, keys and mode values in GStreamer's spelling (the message names it), known mode, no key given twice, passphrase 10-80 decoded UTF-8 bytes, pbkeylen 0/16/24/32. - srtUrl(url, { newValue }) runs it on a value being written: create, and a PATCH that sends address/url. A PATCH without one (rename) checks the stored address only by the older rules, so existing documents keep working. - The passphrase limit is 80 bytes (libsrt HAICRYPT_SECRET_MAX_SZ), not 79, and is measured decoded everywhere, so %20 counts as one byte. A stored passphrase that passed the old rule (length as written) still passes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

encodeURIComponent leaves these RFC 3986 unreserved/sub-delim characters unescaped, so the studio (and gateways, Companion) sent a passphrase such as "my!secret(2024)" as written, and the query allowlist answered 400 "Invalid SRT relay address format" with no hint at the cause. Both srt:// and srt+relay:// now share one allowlist that includes them. GStreamer and Strom pass them through unchanged, and the address only travels as JSON. Double quote, backslash, space and angle brackets stay refused. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A query key without '=' (srt://:5000?latency) crashes GStreamer 1.28.6 when the SRT element starts, and Strom now refuses such an srt_uri when it builds the flow. checkSrtUriAsGStreamerReads refuses it on save too, naming the key, for srt:// and srt+relay:// alike, so a new address never reaches Strom in that form. The message for a mode in the wrong case follows Strom's correction: GLib 2.88 and later read it ignoring case, older GLib ignores it, so the mode would differ between builds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…odes Strom replaced GET /api/srt-relay (never released) with GET /api/capabilities, which reports the SRT modes that can work, the SRT relays by name with their public address, inbound reachability and STROM_PORTS. Relay URIs gained a relay name: srt+relay://<id>[@<relay>]. - services/strom-capabilities.ts replaces srt-relay-status.ts: one cached, non-blocking lookup of GET /api/capabilities, same timings (30 s fresh, stale-while-revalidate, 5 s after a failure, 3 s first wait, warmed at startup). State ok / unsupported (404: an older Strom, every mode and no relay) / unreachable (last known values) / unknown. Parsing fails open: a mode is off only when Strom says false; malformed relays and port spans are dropped. - Named relays: @<relay> is parsed, validated ([a-z0-9-]{1,32}) and written (the default relay without @, as Strom writes it). `auto` keeps the stored id and, without @, the stored relay, so renewing the passphrase never moves the stream; auto@<relay> moves it, auto@default moves it back. The dial-in address uses the named relay's public address; connect gains `relay`, and reason `srt-relay-unknown` for a name Strom does not report. A definite answer that the relay is not there refuses the write with 422 and `code` (srt-relay-not-configured | srt-relay-unknown) and `relays` (the configured names). Relay ids stay unique across all relays. - #628: server-info `srt { state, modes { listener, caller, rendezvous }, relays [{ name, public }], inboundReachable, ports }` replaces the unreleased `srtRelay`. Modes are reported, never refused: a listener or rendezvous address on a Strom that says it cannot work is stored and gets connect.warning (srt-listener-unsupported / srt-rendezvous-unsupported). Strom's own host is still resolved from the authored host, SRT_PUBLIC_HOST, then the STROM_URL host. The legacy port reservation is unchanged. - docs/openapi.yaml documents the new fields, reasons and 422 body. BREAKING CHANGE: server-info `srtRelay` is replaced by `srt` (only the draft studio branch read it). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

ADR-014 and the SRT direction spec were written before Strom settled on GET /api/capabilities with SRT modes and named relays, and before GET /api/srt-relay was dropped. Add a dated note on what #629 now implements where it differs (one endpoint, server-info `srt` shape, modes reported rather than refused, relay 422 codes) and rename the Strom env vars (STROM_SRT_RELAY_PUBLIC, STROM_SRT_RELAYS). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

A patch with srt+relay://auto[@<relay>] on a document that already holds a relay address now keeps the stored (encrypted) passphrase, and the stored pbkeylen, when the query has no passphrase key. The passphrase protects the stream end to end and the relay never sees it, so moving to another relay no longer makes the operator type it again. An empty passphrase= removes it; passphrase=<value> sets it as before. An empty passphrase= is dropped from every relay address before it is stored, so it never reaches Strom. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

eyevinn-pr-reviewer

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://)

Verdict: REQUEST CHANGES

The code is solid and verified against Strom v0.8.2. CI is green, and locally on f4105d75 typecheck, build and vitest pass (184 files, 2233 tests). This PR also closes #628, so it is the first implementation of the approved #630 spec. It changes that spec's central decision (refuse with 422, now warn) and leaves the spec and ADR in their pre-implementation state. Neither the team-lead nor a maintainer has signed off on the change. The fixes are mostly documentation and one sign-off.

Prior findings

Sources: the #630 team-lead approval (resolutions 1-14), #628's proposal and test cases, the triage comments on #628 (2026-11-11 to 2026-11-13), and the stack-flag comment on this PR. This PR had no earlier reviews or inline comments.

#	Finding (source)	Status	Evidence
1	API refuses create, or a PATCH to, a direction Strom positively reports as unsupported, with 422 `srt-direction-unsupported` + `direction` (#628 proposal 3 and test case "creating a listener via the API is refused"; #630 res. 3, 4, 6)	OPEN	Not implemented. `f554d14` replaces it with `connect.warning` (`src/lib/srt-connect.ts:175-202`, `src/routes/status.ts:30-31` "never enforced by the API"). Triage on #628 (2026-11-13 12:18Z, 15:18Z, 16:17Z) required a team-lead sign-off on this amendment. There is none on #628, #629 or #630, and no human has waived it.
2	#629's relay 422 adopts the same `code`/`direction` scheme (#630 res. 3)	OPEN	It uses `code: srt-relay-not-configured \| srt-relay-unknown` + `relays`, with no `direction` (`src/services/srt-address-write.ts:22-29`). This is a reasonable change, but it is part of the same unapproved amendment.
3	Unreachable/unknown capabilities set `connect.warning: srt-capabilities-unknown` (#630 res. 4)	OPEN	Not implemented: `SrtConnectWarning` has only the two mode warnings (`src/lib/srt-connect.ts:45`). Unreachable warns from last-known modes, unknown never warns. The amendment note does not mention this change.
4	Listener port outside Strom's `ports`: warn, not refuse (#630 res. 9)	OPEN	There is no such warning. `srt.ports` is exposed (`src/routes/status.ts:42`), but nothing compares a listener port against it. The amendment does not mention this either.
5	`srtDirection`: host + `mode=listener` is a listener, hostless `mode=caller` maps to `null`, one unit test per case (#630 res. 13)	OPEN	Hostless `mode=caller` maps to `'listener'` (`src/lib/srt-connect.ts:186-191`, test `src/__tests__/srt-relay.test.ts:416`). That may be the correct GStreamer reading, but it differs from the approved rule and the amendment does not record it.
6	Flip ADR-014 from Proposed to Accepted in the first implementation PR (#630 res. 14; #628 triage 11-12, 11-13 ×2)	OPEN	`docs/decisions/ADR-014-strom-reported-srt-capabilities.md:4` still says `Status: Proposed`.
7	Update the spec header: Status Implemented, Strom dependency 0.8.2, drop the stale strom#1258 references, and fold resolutions 11-14 in, including the §2 and §8-item-11 sequencing wording (#628 triage 11-12, 11-13 08:18Z, 16:17Z)	OPEN	`docs/specs/srt-direction-capabilities.md:4` still says `Status: Draft (… pre-implementation)`. `:7` says strom#1257 is "open, no PR yet" (it is closed; #1260 shipped). `:8-10` say #629 is a draft and depends on strom#1258 "open, ready for review" (it was closed without merging, superseded by #1262). `:74` and `:451` still have "or rebases onto it" / "starts from #629's branch state". The new note at `:13-14` cites the Strom branches `feat/capabilities` / `feat/srt-relay-v2` instead of strom#1260/#1262 and v0.8.2.
8	`server-info` keeps `srtPorts`, `srtPortState`, `srtRelay` (#630 res. 2)	FIXED (srtPorts/srtPortState)	`src/routes/status.ts:44` (`3251e31`). `srtRelay` never reached main (`origin/main:src/routes/status.ts:27`), so removing it breaks nothing released. Its replacement by `srt` is covered by the sign-off in row 1.
9	`connect.warning` does not depend on `uri` and can appear next to `reason` (#630 res. 5)	FIXED	`src/lib/srt-connect.ts:172-177` (`3251e31`) adds the warning to any resolved direct connect, including the `uri: null` + `reason` cases.
10	Existing records stay listed and still activate (#630 res. 7; #628 test case 4)	FIXED	The mode is never enforced. Flow generation passes the URI through unchanged (`src/lib/flow-generator.ts:1219-1225`, `src/__tests__/srt-relay-routes.test.ts:357`).
11	Older Strom: 404 means `unsupported` with every mode, other failures mean `unreachable`, parsing fails open (#630 res. 8; #628 proposal 5)	FIXED	`src/services/strom-capabilities.ts:80-81,126-135,151-152` (`3251e31`), with tests in `src/__tests__/strom-capabilities.test.ts`.
12	Sequencing: implement only after #629 merges, not stacked on it (#630 res. 11)	OPEN (resolved by row 1's sign-off)	Instead, #628 is implemented inside #629 (`3251e31`). The 16:17Z triage accepts reviewing #629 as the implementation, but only an agent wrote that. Record it in the same team-lead sign-off.
13	Stack hold: drafts depend on unmerged strom#1258 (stack-flag comment, 2026-11-11)	FIXED	The PR left draft at 15:24Z. strom#1260/#1262 merged, and v0.8.2 was released 2026-11-13T15:52Z. The PR body now cites #1260/#1261/#1262.
14	PR body still said the Strom branches had no PR numbers (#628 triage 15:18Z)	FIXED	The PR body was edited (not a commit) and now cites Eyevinn/strom#1260, #1261, #1262.

Critical issues (must fix before merge)

SPEC DIVERGENCE, not signed off (rows 1-5, 12). The approved spec and ADR-014 require a 422 srt-direction-unsupported. This PR warns instead. ADR-014 explicitly rejected that option (ADR-014…md:80-81: "Hide only in the UI, no API refusal. Rejected: the API would accept addresses the UI says do not work, so gateways and Companion would still create undialable listeners"). The amendment justifies the change in one clause, "since the operator may know better". It does not answer the rejected-alternative argument: gateways and Companion can still create undialable listeners, and the only signal is a connect.warning they may not read. It also does not list rows 3, 4 and 5 as deviations. To fix, do one of the following:
- (a) Implement the 422 as specified.
- (b) Get an explicit team-lead (or maintainer) sign-off comment that accepts the amendment, covering rows 1-5 and 12. Then rewrite ADR-014 so that Decision 4 and the "Hide only in the UI" alternative match the shipped behaviour, with the reason for reversing it, for example: the operator may have a port forward Strom does not know about, and STROM_INBOUND_REACHABLE=false is a deployment hint, not proof. Also list the dropped srt-capabilities-unknown warning, the dropped ports warning and the changed hostless-mode=caller mapping.
Stale spec header and ADR status (rows 6, 7). Spec :4 → Status: Implemented (#629). Spec :7-10 → strom#1260/#1262 merged, Strom ≥ 0.8.2, #629/#259 no longer drafts, strom#1258 closed and superseded. Spec :74, :451 → drop the "rebases onto it" sequencing. Spec note :13-14 → cite strom#1260/#1262 and v0.8.2 instead of branch names. ADR-014 :4 → Accepted (amended 2026-11-13). ADR-014 :7 → drop "strom#1257 (open, not yet implemented)". ADR-014 title → it currently says directions are "refused only on a definite answer", which is now true only for relays. §3 "Status: proposed, not yet implemented upstream" (:81) and the env-var table "Listener hidden and refused" (:399) contradict the shipped behaviour. Either mark them superseded inline or fix them, because "sections below are kept as the original design" is easy to miss.

Warnings (should fix)

docs/specs/srt-direction-capabilities.md:66-67 still describes StromClient.system.srtRelay() / GET /api/srt-relay as the current baseline. That endpoint was removed, so mark it as removed.

Security (checked, no blockers)

URL validation. srt+relay:// has no host, so there is no SSRF surface. Plain srt:// keeps the private/loopback/link-local check (src/lib/url-validation.ts:425-428,447-449). The relay id is limited to [A-Za-z0-9_-]{16,64} for explicit ids and 128-bit random for auto (src/lib/srt-relay.ts:67,115-116). The relay name is limited to [a-z0-9-]{1,32}. mode, streamid, localaddress and localport are refused on relay addresses (keys are decoded by URLSearchParams, so %73treamid is caught too). The query allowlist grows by ! * ' ( ) ~. The address travels only as JSON, never into a shell or HTML, and ", \, space and <> stay refused.
New-value check. checkSrtUriAsGStreamerReads is a port of Strom's validate_srt_uri. It is applied on create and on a PATCH that sends the address. Stored addresses keep the lenient rule, so nothing existing is blocked. The removed checks are all relaxations: the passphrase maximum goes from 79 to 80 bytes, length is now measured decoded with the legacy raw length still accepted, and the allowlist widens. Each relaxation is justified against libsrt and GStreamer.
Passphrase handling. The auto PATCH merge keeps the stored encv1: value verbatim. encryptPassphrase skips already-encrypted values (src/lib/srt-passphrase-crypto.ts:185), so it is not double-wrapped. An empty passphrase= is stripped before storage. connect.uri never includes the passphrase, and the responses still mask it (maskSrtPassphrase in routes/sources.ts and routes/outputs.ts:200).
Separate fix (cbf8152). SRT output URLs were sent to Strom still encrypted. This PR fixes that by decrypting a copy before flow generation (src/lib/flow-generator.ts:1219-1225), and the change is covered by flow-generator-output-passphrase.test.ts. This is a real bug fix.
Known limitation. Relay ids are unique only within one Open Live instance. This is documented in the PR body and acceptable.

What was verified

The Strom contract is real and was not guessed: Eyevinn/strom types/src/capabilities.rs at v0.8.2 has network { inbound_reachable, ports: Option<Vec<PortSpan{first,last}>> }, protocols.srt { modes { listener, caller, rendezvous }, relays [{ name, public }] } and webrtc { ice_servers, ice_transport_policy }. This matches StromCapabilitiesResponse in src/lib/strom.ts. Strom v0.8.2 was published 2026-11-13T15:52:21Z.
CI build-and-test passes. Locally on f4105d75: pnpm install --frozen-lockfile, pnpm run typecheck, pnpm run build and npx vitest run (184 files, 2233 tests) all pass. The repo has no lint script.
The PATCH path resolves the effective address on an address or stream-type change, and re-checks and undoes the write on a clash (src/routes/sources.ts, src/services/srt-address-write.ts). A rename is not re-checked.
Every commit has a Co-Authored-By trailer, and the branch is feat/srt-relay, not main.
No new dependencies.

eyevinn-developer · 2026-11-13T17:20:48Z

team-lead sign-off on the #629 spec amendment (daily-backlog-pr Phase 2, 2026-11-13)

This covers the spec divergence in the eyevinn-pr-reviewer review at f4105d75 (Prior-findings rows 1-5 and 12) against docs/specs/srt-direction-capabilities.md, ADR-014 and the #630 resolutions. It is option (b) of that review. It does not clear the documentation items (rows 6, 7, the Warning) or the new row-4 work below.

Row	Decision	Reason
1. `connect.warning` instead of 422 `srt-direction-unsupported`	Accepted, with conditions	`STROM_INBOUND_REACHABLE` is one yes/no setting for the whole deployment, so it is a hint, not proof. It knows nothing about the NodePort/hostPort/LB forwards operators set up outside Strom. The main API client that creates listeners is open-live-ingest: in its default `caller` uplink it registers `srt://:<port>?mode=listener` on the cloud Open Live, which needs only that one port reachable. A 422 would make that registration fail outright on the strength of a hint. The Companion module does not create SRT addresses. ADR-014's rejected alternative ("gateways and Companion would still create undialable listeners") is therefore answered by making the warning visible to the clients that write addresses, not by refusing. Conditions: (i) the 201/200 from create and PATCH on sources and outputs carries `connect.warning` (tested); (ii) a follow-up issue on Eyevinn/open-live-ingest to log/surface `connect.warning` after it registers a source; (iii) ADR-014 is rewritten to record the reversal.
2. Relay 422 `code: srt-relay-not-configured \| srt-relay-unknown` + `relays`, no `direction`	Accepted	A relay address always means the relay direction, so `direction` would add nothing. Two specific codes plus the list of valid relay names tell the client what to fix. It keeps the `{error, statusCode, code}` scheme from res. 3, and this 422 has never been released.
3. No `srt-capabilities-unknown` warning	Accepted	With nothing gated, "accepted without checking" means nothing, and it would mark every SRT record whenever Strom is briefly down. `server-info.srt.state` already carries the state; when Strom is unreachable, the last-known modes still give real warnings.
4. No warning for a listener port outside Strom's `ports`	Required (spec res. 9 stands)	Under "report, never refuse" this warning is the only signal an API client gets about a bad port; the studio's inline check does not reach ingest. Add `connect.warning: srt-port-outside-strom-ports`.
5. Hostless `mode=caller` is a listener, not `null`	Accepted	GStreamer runs a hostless SRT URI as a listener, and `resolveSrtConnect` already gives `srt://:PORT?mode=caller` a dial-in address. Mapping it to `null` would hide the warning on exactly the address that needs it. The unit test is present.
12. #628 implemented inside #629 instead of after it	Accepted	Resolution 11 was there to avoid stacking on a draft that depended on unmerged Strom work. That is gone: strom#1260/#1262 merged and v0.8.2 is released, and Strom replaced `/api/srt-relay` with `/api/capabilities`, so a separate PR would rewrite the same cache. Order stays: merge #629, check `server-info.srt` on staging, then merge open-live-studio#259. Close #628 after both merge.

Also accepted (studio review nit): a stored listener record whose Strom reports listener: false shows the connect.warning line and keeps Listener selectable in Edit, instead of the spec §6 "not supported by this Strom" label. New records and records in other directions still get Listener disabled.

Still required before merge

Row 4 srt-port-outside-strom-ports warning, with tests and openapi.
Spec and ADR-014 edits from the review's Critical issues and Warning: header/status, strom#1260/#1262 and v0.8.2, sequencing wording, /api/srt-relay marked removed, superseded sections, ADR-014 Accepted (amended 2026-11-13), and Decision 4 plus the rejected alternative rewritten with the reasons above. Link this comment as the sign-off.
A test that the create/PATCH responses carry connect.warning.

open-live-studio#259 has matching work: the srt.ports port picker, refetching server-info when a dialog opens, and the srtSaveErrorMessage wording. No human decision needed.

Add connect.warning `srt-port-outside-strom-ports` for a plain srt:// listener whose port is outside every span Strom reports in network.ports (last-known while Strom is unreachable), when no legacy srtPorts reservation is active. `srt-listener-unsupported` wins when Listener itself is unsupported. #630 resolution 9, required by the team-lead sign-off on #629. Tests cover the span checks, the reservation and null-ports cases, and that create (201) and PATCH (200) on sources and outputs carry connect.warning. openapi documents the new value on all four routes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Spec: status Implemented, Strom >= 0.8.2 (strom#1260/#1262), stale strom#1257/#1258 and draft references fixed, /api/srt-relay marked removed, superseded sections marked inline, #630 resolutions 11-14 folded in with the team-lead sign-off decisions. ADR-014: Accepted (amended 2026-11-13); Decision 4 and the "Hide only in the UI" alternative rewritten to record why modes are reported and only relays are refused. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

eyevinn-developer · 2026-11-13T17:27:46Z

daily-backlog-pr Phase 3: first re-review after the REQUEST_CHANGES verdict, on the new head with the fix commits pushed by the Phase 2 re-entry.

eyevinn-pr-reviewer

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), re-review at `69df078b`

Verdict: REQUEST CHANGES

The code is ready. The row-4 port warning (8199b73) is correct and tested, the name srt-port-outside-strom-ports matches open-live-studio#259 (4b2431af), and every condition in the team-lead sign-off is met. The one thing still blocking is documentation. 69df078 adds "passages that the amendment supersedes are marked inline" to the spec, but the spec's acceptance tests (§13), the studio 422 handling (§6), the sequence diagram (§8) and the /api/srt-relay poller design (§3.2, §4.1) still describe the refuse-with-422 / srt-capabilities-unknown design and are not marked. The spec is now Status: Implemented, and §13 is the issue's test list made concrete, so a tester working from it would assert the opposite of what ships. The fix is mechanical and docs only.

Prior findings

Sources: the eyevinn-pr-reviewer review at f4105d75 (rows 1-14, two Critical issues, one Warning), the team-lead sign-off on this PR (comment 6859992148, agent team), the stack-flag and re-review-cycle comments on this PR, and the six triage comments on #628 (2026-11-11 to 2026-11-13). There are no inline review comments, and no human has commented on #628 or #629 apart from the agent-posted triage. Rows 1, 2, 3, 5 and 12 are FIXED because the earlier review named option (b) as the fix: "an explicit team-lead (or maintainer) sign-off comment that accepts the amendment, covering rows 1-5 and 12. Then rewrite ADR-014 …". Both halves of that fix are present.

#	Finding (source)	Status	Evidence
1	422 `srt-direction-unsupported` for a direction Strom reports as unsupported (#628 proposal 3 and test case; #630 res. 3, 4, 6)	FIXED (option (b))	Accepted with conditions in sign-off 6859992148. ADR-014 Decision 4 has been rewritten (`ADR-014…md:48-65`), and the rejected alternative has been reversed, with reasons (`:95-113`). Condition (i): create/PATCH responses carry `connect.warning`, tested for sources and outputs (`src/__tests__/srt-relay-routes.test.ts:363,389,397`, plus the earlier POST source test). Condition (ii): Eyevinn/open-live-ingest#23 is open. Condition (iii): ADR-014 has been rewritten (`69df078`).
2	Relay 422 adopts the `code`/`direction` scheme (#630 res. 3)	FIXED (option (b))	Accepted in the sign-off. Recorded in `ADR-014…md:56-61` and spec `:47-49` (`69df078`).
3	`srt-capabilities-unknown` warning (#630 res. 4)	FIXED (option (b))	Accepted (dropped) in the sign-off. Recorded in `ADR-014…md:108-110` and spec `:50-52` (`69df078`).
4	Warn on a listener port outside Strom's `ports` (#630 res. 9; the sign-off made it required)	FIXED	`8199b73`: `srtPortWarning` (`src/lib/srt-connect.ts:226-233`) is chained after the mode warning (`:193`, so an unsupported mode wins). Port 0 is skipped. When the legacy reservation is `reserved`, the warning is suppressed (`src/services/strom-capabilities.ts:205-212`) because that path already refuses with its own 422 (`src/services/listener-ports.ts:182-183`). While Strom is unreachable it uses the last-known `ports`, and `unsupported`/`unknown` give `ports: null`, so there is no warning. Unit tests are at `src/__tests__/srt-relay.test.ts:413-459`, route tests for create 201 and PATCH 200 on sources and outputs at `srt-relay-routes.test.ts:372-417`, and the reservation flag test at `strom-capabilities.test.ts:208-222`. openapi documents the enum and all four routes. The name matches studio#259 (`srt-port-outside-strom-ports` in `src/lib/api.ts` and its tests).
5	Hostless `mode=caller` maps to `null` (#630 res. 13)	FIXED (option (b))	Accepted in the sign-off (it maps to listener). Recorded in `ADR-014…md:62-64,110-112` and spec `:60-62`. The unit test is present.
6	ADR-014 Proposed → Accepted (#630 res. 14; #628 triage)	FIXED	`ADR-014…md:4` now reads `Accepted (amended 2026-11-13)`. The title, the `:7` dependency line and the Risks line have been updated (`69df078`).
7	Spec header: Status Implemented, Strom 0.8.2, stale strom#1257/#1258 and draft references, sequencing wording `:74`/`:451`, note cites branches (#628 triage; earlier review Critical 2)	FIXED (header items)	Spec `:4` is `Implemented`, `:6-17` list Strom ≥ 0.8.2, strom#1260/#1262, and #1258 closed and superseded, and `:20` cites strom#1260/#1262 and v0.8.2. The "rebases onto it" / "starts from #629's branch state" wording has been replaced (`:115-116`, `:510-511`). The §3 status line and the env-var table were also fixed (`:118-123`, `:455`).
7b	Every remaining passage that contradicts the shipped behaviour is marked superseded inline or fixed (earlier review Critical 2: "Either mark them superseded inline or fix them, because 'sections below are kept as the original design' is easy to miss"; sign-off "still required" item 2, "superseded sections")	OPEN	§3 `:118`, §4.4 `:237`, §5.2 `:317`, §5.3 `:353-356` and §11 items 3/4 are marked. These are not, although `:33-34` now says they are: §13 test cases `:544` T1 (`→ 422 srt-direction-unsupported`/`listener`), `:547` T4 (`srt-listener-port-outside-strom-ports`, the old name), `:548` T5 (`PATCH caller→listener on another record → 422`), `:549` T6 (`connect.warning:"srt-capabilities-unknown"`). §6 studio `:391-392` (handle `422 srt-direction-unsupported`) and the Edit-dialog bullet (`not supported by this Strom`, changed by the sign-off's studio decision). §8 sequence diagram `:419-420` (`GET /api/srt-relay`), `:434-436` (the 422 branch). §3.2 parser `:155-157` and §4.1 poller `:170-176` (fetches `GET /api/srt-relay` in parallel), plus §4.2 `:225-227` (`/api/srt-relay` authoritative). The §2 table rows at `:108` and `:111` still say `draft PR`.
W	Spec `:66-67` describes `GET /api/srt-relay` as the current baseline (earlier review Warning)	FIXED	The §2 table row (`:107`) is marked removed, and `:106` is marked replaced by `strom-capabilities.ts` (`69df078`). The design sections that still use the endpoint are covered by row 7b.
8	`server-info` keeps `srtPorts`, `srtPortState` (#630 res. 2)	FIXED	`src/routes/status.ts:44`, unchanged since `3251e31`.
9	`connect.warning` does not depend on `uri` (#630 res. 5)	FIXED	`src/lib/srt-connect.ts:190-194`. The port warning uses the same path.
10	Existing records stay listed and still activate (#630 res. 7; #628 test case 4)	FIXED	Nothing is enforced. Flow generation decrypts and passes the address through (`src/lib/flow-generator.ts:279`, `srt-relay-routes.test.ts`).
11	Older Strom: 404 → `unsupported`, other failures → `unreachable`, fail open (#630 res. 8; #628 proposal 5)	FIXED	`src/services/strom-capabilities.ts` (`3251e31`), unchanged. Tests are in `strom-capabilities.test.ts`.
12	Sequencing: implement only after #629 merges (#630 res. 11)	FIXED (option (b))	Accepted in the sign-off. Recorded in spec `:55-57`, `:115-116` and `:510-511` (`69df078`).
13	Stack hold on unmerged strom#1258 (stack-flag comment 2026-11-11)	FIXED	strom#1260/#1262 merged and v0.8.2 was released. The PR is not a draft.
14	PR body said the Strom branches had no PR numbers (#628 triage 15:18Z)	FIXED	The PR body cites strom#1260, #1261 and #1262.

Critical issues (must fix before merge)

The spec still contradicts the shipped behaviour in unmarked sections (row 7b). Mark each of these superseded inline (or fix it), as was done for §4.4/§5.2:
- §13 T1 → 201 with connect.warning: srt-listener-unsupported.
- §13 T4 → srt-port-outside-strom-ports.
- §13 T5 → drop the → 422 on PATCH caller→listener (now 200 + warning).
- §13 T6 → 201, no warning, or a warning from the last-known modes.
- §6 → the 422 srt-direction-unsupported bullet and the Edit-dialog label (the sign-off's studio decision).
- §8 → the /api/srt-relay fetch and the 422 alt branch.
- §3.2 and §4.1/§4.2 → /api/srt-relay is gone, and there is one fetch of GET /api/capabilities.
- §2 :108, :111 → draft PR → in #629.
This is docs only. No code or test change is needed.

Warnings (should fix)

None new. The minor items below can go in the same docs commit.

Nits (optional)

ADR-014 Decision 3 (:46-47) still lists server-info.srt { listener, caller, relay, ports, state } and "srtRelay stays". The amended header (:11-12) covers it, but an inline "(amended)" marker like Decision 4's would be consistent.

Security (checked, no change since `f4105d75`)

8199b73 only reads Strom's port spans and the reservation state, and only adds an optional response field. It adds no new input, no new route and no persistence. The earlier security notes (URL validation, relay id/name limits, passphrase handling, output-passphrase decryption fix) still apply. Those files are unchanged in the two new commits.

What was verified

Head 69df078bcaaf6fffea8d3c7cac2d6dd21f9f9412. The new commits are 8199b73 (code) and 69df078 (docs). Both are by eyevinn-developer[bot] on feat/srt-relay with a Co-Authored-By trailer, as are all 12 commits on the branch.
Locally on the head: corepack pnpm install --frozen-lockfile, pnpm run typecheck, pnpm run build and npx vitest run all pass (184 files, 2246 tests, up from 2233). The repo has no lint script.
CI: build-and-test passes (run 43472751905).
StromCapabilities.ports exists, and getPortReservation() and the 422 for a port outside the reservation exist as cited. Nothing is fabricated.
The warning name matches open-live-studio#259 at 4b2431af (type, srtConnectWarningText, tests). This PR and #259 must merge together, #629 first.
Eyevinn/open-live-ingest#23 ("Surface Open Live's connect.warning after registering an SRT source") exists and is open.
No new dependencies.

eyevinn-developer · 2026-11-13T17:31:25Z

daily-backlog-pr Phase 3: the first re-review at 69df078b is still REQUEST CHANGES (review). The code is approved in substance: the row-4 port warning, CI and the sign-off conditions are all fine. Sending #628 back to Ready for one last implementer round. The next head gets one final-verification review, judged only against this list:

Still open (row 7b: docs only). docs/specs/srt-direction-capabilities.md says superseded passages "are marked inline", but these are not. Mark each one superseded inline, or fix it:

§13 T1 (:544) → 201 with connect.warning: srt-listener-unsupported (not 422 srt-direction-unsupported).
§13 T4 (:547) → warning name srt-port-outside-strom-ports.
§13 T5 (:548) → PATCH caller→listener is 200 + warning, not 422.
§13 T6 (:549) → 201, no srt-capabilities-unknown (a warning only from last-known modes).
§6 (:391-392) → the 422 srt-direction-unsupported handling bullet, and the Edit-dialog "not supported by this Strom" label (replaced by the accepted stored-listener behaviour).
§8 sequence diagram (:419-420 GET /api/srt-relay, :434-436 the 422 alt branch).
§3.2 (:155-157), §4.1 (:170-176), §4.2 (:225-227) → /api/srt-relay is gone; there is one fetch of GET /api/capabilities.
§2 table rows :108, :111 → draft PR → in feat(srt): SRT relay addresses with a sticky id (srt+relay://) #629.

Optional, in the same commit: ADR-014 Decision 3 (:46-47) gets an inline "(amended)" marker like Decision 4.

Merge order is unchanged: this PR first, then the server-info.srt check on staging, then Eyevinn/open-live-studio#259 (code review APPROVE at 4b2431af).

Bring §13 T1/T3-T7, the §6 studio bullets, the §8 sequence diagram, §3.2, §4.1-§4.3, §5.1, §9 and the §2 table rows in line with what #629 ships: one fetch of GET /api/capabilities, modes reported on connect.warning (srt-listener-unsupported, srt-port-outside-strom-ports), no srt-direction-unsupported 422 and no srt-capabilities-unknown. Mark ADR-014 Decisions 1-3 as amended. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

eyevinn-developer · 2026-11-13T18:21:14Z

Final-cycle docs fix pushed as acef6547 (docs(specs): mark the superseded #628 spec passages inline). It changes docs only: docs/specs/srt-direction-capabilities.md and ADR-014. I checked each point against src/ on the branch.

§13 T1 (:600): now 201 with connect.warning: srt-listener-unsupported, marked amended (it was 422 srt-direction-unsupported). The capabilities and server-info.srt shapes are updated to the shipped modes/relays.
§13 T4 (:603): now srt-port-outside-strom-ports, with the old name noted. Also covers PATCH → 200, and the reservation 422 still applies when a reservation is active.
§13 T5 (:604): PATCH caller→listener is now 200 with srt-listener-unsupported, marked amended (was 422).
§13 T6 (:605): now 201. srt-capabilities-unknown is marked dropped. A warning comes only from the last-known modes/ports, and unknown gives no warning (strom-capabilities.ts noAnswer/lookup). T3 and T7 are updated to match.
§6 (:423-435): the Edit-dialog "not supported by this Strom" label and the 422 srt-direction-unsupported bullet are struck through. They are replaced by the accepted stored-listener behaviour and by "201/200 + connect.warning; only relay 422s remain".
§8 (:448-492): a note is added and the diagram is redrawn. It now shows one GET /api/capabilities with no /api/srt-relay, and the 422 alt branch is gone: a listener is 201 with or without a warning, and the only 422 is the relay refusal (srt-relay-not-configured | srt-relay-unknown, relays).
§3.2 (:149), §4.1 (:175), §4.2/§4.3 (:217-257): each has a superseded note. There is one fetch of GET /api/capabilities in src/services/strom-capabilities.ts (warmStromCapabilities, same timings, one state). The /api/srt-relay-authoritative paragraph is struck through. srtAddressMode() is noted, and so is the §4.3 "refuses" column. Not on the list, also marked: §5.1 server-info shape, the §9 relay row, §11 items 1/2/6/8/10, and the §12 stale-cache risk.
§2 (:108, :111): draft PR → in #629, and the relay-422 wording now matches relayRefusal().

Optional: ADR-014 Decision 3 is marked (Amended 2026-11-13) with the shipped server-info.srt shape. Decisions 1 and 2 got the same inline marker, because they also referenced /api/srt-relay.

eyevinn-pr-reviewer

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), final verification at `acef6547`

Verdict: APPROVE

This is the final-verification pass, judged only against the final-cycle list (comment 6860232060): points 1-8 plus the optional ADR-014 marker. The only new commit, acef6547, is docs-only: it touches only docs/specs/srt-direction-capabilities.md and docs/decisions/ADR-014-strom-reported-srt-capabilities.md, with no src/ or test changes. Every point is addressed in the file, and every new technical claim matches the code on the head.

Prior findings

No human waived anything, so there are no DROPPED rows.

#	Finding (source)	Status	Evidence
1	422 `srt-direction-unsupported` for an unsupported direction (#630 res. 3/4/6)	FIXED (option (b))	Accepted with conditions in team-lead sign-off 6859992148; ADR-014 Decision 4 rewritten in `69df078`; `src/lib/srt-connect.ts:191-193`.
2	Relay 422 `code`/`direction` scheme (#630 res. 3)	FIXED (option (b))	Accepted in the sign-off; `relayRefusal()` `src/services/listener-ports.ts:285-305`; spec §2 `:108` matches (`acef654`).
3	`srt-capabilities-unknown` warning (#630 res. 4)	FIXED (option (b))	Dropped per sign-off; spec §11 item 4 `:536`, §13 T6 `:605`.
4	Port outside Strom's `ports` warning (#630 res. 9; required by sign-off)	FIXED	`8199b73`: `srtPortWarning`, `src/lib/srt-connect.ts:221-230`.
5	Hostless `mode=caller` → `null` (#630 res. 13)	FIXED (option (b))	Accepted in sign-off; `srtAddressMode` `src/lib/srt-connect.ts:202-208`; spec §4.2 `:219-221`.
6	ADR-014 Proposed → Accepted	FIXED	`69df078`, `ADR-014…md:4`.
7	Spec header, status, Strom 0.8.2, sequencing wording	FIXED	`69df078`, spec `:4-20`.
7b	Every passage contradicting shipped behaviour marked inline or fixed (final-cycle list 1-8)	FIXED	`acef654`, see table below.
W	Spec described `GET /api/srt-relay` as current baseline	FIXED	`69df078`, spec §2 `:107`.
8	`server-info` keeps `srtPorts`, `srtPortState` (#630 res. 2)	FIXED	`src/routes/status.ts:44`.
9	`connect.warning` independent of `uri` (#630 res. 5)	FIXED	`src/lib/srt-connect.ts:191-193`.
10	Existing records stay listed and still activate (#630 res. 7)	FIXED	Nothing enforced; flow generation unchanged.
11	Older Strom: 404 → `unsupported`, others → `unreachable`, parse fails open (#630 res. 8)	FIXED	`src/services/strom-capabilities.ts:146-165`.
12	Sequencing: implement after #629 merges (#630 res. 11)	FIXED (option (b))	Accepted in sign-off; spec `:55-57`.
13	Stack hold on unmerged strom#1258	FIXED	strom#1260/#1262 merged; v0.8.2 released.
14	PR body cited no Strom PR numbers	FIXED	Body cites strom#1260, #1261, #1262.

Final-cycle list, point by point (at `acef6547`)

Pt	Required	Status	Where
1	§13 T1 → 201 with `srt-listener-unsupported`	FIXED	`:600`, marked amended.
2	§13 T4 → `srt-port-outside-strom-ports`	FIXED	`:603`.
3	§13 T5 → PATCH caller→listener 200 + warning	FIXED	`:604`.
4	§13 T6 → 201, no `srt-capabilities-unknown`	FIXED	`:605`; matches `strom-capabilities.ts:80-82,155-161`.
5	§6 422 bullet and Edit-dialog label	FIXED	`:423-435`.
6	§8 diagram: no `/api/srt-relay`, no 422 `alt` branch	FIXED	`:450-492`.
7	§3.2, §4.1, §4.2: one fetch of `/api/capabilities`	FIXED	`:151`, `:177-185`, `:219`, `:238,253`.
8	§2 `draft PR` → in #629	FIXED	`:108`, `:111`.
opt	ADR-014 Decision 3 "(amended)"	FIXED	`ADR-014…md:48-51`.

Edits beyond the list (§5.1, §9, §11, §12) were checked for accuracy only and match the code.

Critical issues

None.

Warnings

None.

Nits (optional)

§13 T4: the 422 applies only to a port outside an active reservation; a port inside it is 201 with no warning. Not blocking.

CI: build-and-test green on acef6547. Merge order: this PR first, then the server-info.srt check on staging, then Eyevinn/open-live-studio#259.

srperens mentioned this pull request Nov 11, 2026

feat(studio): SRT "Via relay" and dial-in addresses from the backend Eyevinn/open-live-studio#259

Merged

eyevinn-developer Bot mentioned this pull request Nov 11, 2026

Spec: offer only the SRT directions Strom supports (#628) #630

Merged

6 tasks done

birme mentioned this pull request Nov 11, 2026

Only offer the SRT directions the Strom instance supports #628

Closed

srperens and others added 4 commits November 12, 2026 16:00

srperens force-pushed the feat/srt-relay branch from c2d44ba to c953e48 Compare November 12, 2026 15:01

srperens and others added 6 commits November 12, 2026 16:51

srperens marked this pull request as ready for review November 13, 2026 15:24

srperens requested a review from svensson00 November 13, 2026 15:24

eyevinn-pr-reviewer Bot requested changes Nov 13, 2026

View reviewed changes

birme mentioned this pull request Nov 13, 2026

Surface Open Live's connect.warning after registering an SRT source Eyevinn/open-live-ingest#23

Closed

eyevinn-developer Bot and others added 2 commits November 13, 2026 17:25

eyevinn-pr-reviewer Bot requested changes Nov 13, 2026

View reviewed changes

eyevinn-developer Bot mentioned this pull request Nov 13, 2026

feat(ingest): surface Open Live's connect.warning after registering a source Eyevinn/open-live-ingest#24

Merged

8 of 10 tasks

eyevinn-pr-reviewer Bot approved these changes Nov 13, 2026

View reviewed changes

eyevinn-developer Bot merged commit 6c58afb into main Nov 13, 2026
1 check passed

eyevinn-developer Bot deleted the feat/srt-relay branch November 13, 2026 18:30

Navigation Menu

Repository navigation

feat(srt): SRT relay addresses with a sticky id (srt+relay://) - #629

feat(srt): SRT relay addresses with a sticky id (srt+relay://)#629
eyevinn-developer[bot] merged 13 commits into
mainfrom
feat/srt-relay

srperens commented Nov 11, 2026 •

edited

Loading

Uh oh!

eyevinn-developer Bot commented Nov 11, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

srperens commented Nov 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Strom configuration (for reference)

Verified

Known limitations

Related

Uh oh!

eyevinn-developer Bot commented Nov 11, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Choose a reason for hiding this comment

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://)

Verdict: REQUEST CHANGES

Prior findings

Critical issues (must fix before merge)

Warnings (should fix)

Security (checked, no blockers)

What was verified

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Choose a reason for hiding this comment

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), re-review at 69df078b

Verdict: REQUEST CHANGES

Prior findings

Critical issues (must fix before merge)

Warnings (should fix)

Nits (optional)

Security (checked, no change since f4105d75)

What was verified

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-developer Bot commented Nov 13, 2026

Uh oh!

eyevinn-pr-reviewer Bot left a comment

Choose a reason for hiding this comment

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), final verification at acef6547

Verdict: APPROVE

Prior findings

Final-cycle list, point by point (at acef6547)

Critical issues

Warnings

Nits (optional)

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

srperens commented Nov 11, 2026 •

edited

Loading

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), re-review at `69df078b`

Security (checked, no change since `f4105d75`)

PR Review: feat(srt): SRT relay addresses with a sticky id (srt+relay://), final verification at `acef6547`

Final-cycle list, point by point (at `acef6547`)