CCTV Street View fallback: the browser's pose at the catalog position

The fallback keeps the catalog position but takes heading, field of view
and pitch from the request again, as before, clamped as before. A catalog
camera with no pose (every world-cameras row on main) otherwise asked
Google for heading 0, pitch 0, FOV 80 while the browser draws it at the
id-hash heading, pitch -17 and FOV 74, and a calibrated camera lost its
calibration. Cost and the per-IP budget are unchanged.

- The Street View budget is built per proxy, so tests no longer depend on
  GEV_RATELIMIT_GOOGLE_PER_MIN in the shell; the fixture pins it.
- New tests: a pose-less catalog camera, and the per-IP budget (no call
  without a key, the cap, a separate budget per address).
- The client no longer sends lat/lon on /frame.
- CURRENT-STATE, the changelog fragment and SECURITY.md say the fallback
  has its own per-client limit, and that unlisted (seed) cameras get the
  placeholder.
This commit is contained in:
German Khokhlov committed 2026-11-19 19:06:53 +04:00
1 parent f579b23f52
commit 11e1c9d26d
8 files changed
+235 -36

No files matched your search

+2 -2
View File
@@ -68,7 +68,7 @@ browser-key fallback for existing single-key setups.
The data proxies under `server/providers/` are written so the browser cannot turn the server into an open relay:
- **No arbitrary-URL fetching.** The CCTV frame proxy fetches only server-registered camera/frame URLs — clients cannot pass an upstream URL to fetch (SSRF mitigation). Other proxies target fixed upstream hosts.
- **No arbitrary-URL fetching.** The CCTV frame proxy fetches only server-registered camera/frame URLs — clients cannot pass an upstream URL to fetch (SSRF mitigation). The key-bearing Street View fallback likewise answers only for a registered camera, always at its catalog position: a request sets only the view direction, field of view (20–120°) and pitch (−40–20°), so it cannot buy Street View of a place of its choosing. Other proxies target fixed upstream hosts.
- **Live HLS is bounded and origin-pinned.** Registered HTTP(S) HLS sources use a Node puller that rejects redirects and off-origin playlist references, bounds playlists and segment bodies while streaming, and aborts timed-out or released sessions. At most two sessions retain 12 segments/24 MiB each, plus bounded download buffers. Stores are memory-only and idle sessions expire after 15 seconds. RTMP/ffmpeg execution is not enabled. Locally configured source URLs remain an operator trust boundary.
- **Radio is not an audio relay.** `/api/radio/stations` contacts only allowlisted Radio Browser HTTPS hosts and paths, rejects redirects, rejects any hostname with a loopback/private/link-local/metadata/non-public A or AAAA result, and pins each TLS connection to a validated address. It returns normalized public HTTPS stream URLs; `/api/radio/click/:uuid` applies the same destination policy and accepts only station IDs from the current bounded catalog. The browser then connects directly to the broadcaster after an explicit playback action, so the broadcaster sees the listener's IP address. GEV never proxies, caches, records, or redistributes audio.
- **No verbatim client headers upstream.** The CCTV media route is the one route that relays a request header (`Range`, for video seeking). It is parsed and canonicalized before it is forwarded: one `bytes=` range only, with every accepted form — explicit span, open-ended and suffix — bounded to the same 64 MiB ceiling the relay applies to a declared response body. Multi-range, malformed, inverted and non-`bytes` values are dropped and the request proceeds without a `Range`, as RFC 7233 §3.1 prescribes. Bounding the request bounds what is asked for: a response that declares no length — live streamed media, or a chunked body from an upstream that ignores the `Range` — has no ceiling. An upstream request the browser has stopped waiting for is cancelled rather than left running, whether the viewer leaves before the headers arrive or during the body.
@@ -95,7 +95,7 @@ The dev server is a **key broker**: every server-side key above is spendable by
- **Local-only by default.** `./scripts/dev-fresh.sh` (and the Vite config itself) bind to `localhost`, so only your machine can reach the server — and only local names are accepted (`allowedHosts` stays restricted, which also blunts DNS-rebinding tricks). Binding to all interfaces keeps that restriction: IP addresses and `localhost` work, and LAN hostnames must be listed in `GEV_ALLOWED_HOSTS` (for example `GEV_ALLOWED_HOSTS=globe.lan`); suffix and wildcard entries are ignored.
- **LAN exposure is an explicit opt-in**: `HOST=0.0.0.0 ./scripts/dev-fresh.sh`. The launcher prints a prominent warning plus your LAN URL. Understand what opting in means: **every device on that network can drive the proxies and spend your OpenAI / Google / OpenSky / AISStream / TomTom / FIRMS quota** for as long as the server runs. Do this only on networks you trust.
- **App-level throttles (on by default):** `GEV_RATELIMIT_OPENAI_PER_MIN` (default `30`) and `GEV_RATELIMIT_GOOGLE_PER_MIN` (default `120`) — the values the Pinokio build already ships — cap the cost-bearing endpoints per client IP per minute (over-limit requests receive a sanitized `429`). Typed chat keeps its own per-IP window, set by `GEV_RATELIMIT_OPENAI_CHAT_PER_MIN` and `GEV_RATELIMIT_GEMINI_CHAT_PER_MIN`; left unset, each takes its provider's resolved cap (`GEV_RATELIMIT_OPENAI_PER_MIN`, or `GEV_RATELIMIT_GEMINI_PER_MIN`, default `30`), so lowering a provider throttle lowers chat too, while an explicit chat value, including `0`, wins. Both defaults sit well above what the app itself generates; set either to exactly `0` to run unthrottled. They are **per-IP, process-local, in-memory guards** — they reset on restart, a caller with several addresses gets a bucket per address, and they are **not billing caps**.
- **App-level throttles (on by default):** `GEV_RATELIMIT_OPENAI_PER_MIN` (default `30`) and `GEV_RATELIMIT_GOOGLE_PER_MIN` (default `120`) — the values the Pinokio build already ships — cap the cost-bearing endpoints per client IP per minute (over-limit requests receive a sanitized `429`). The CCTV Street View fallback has its own per-IP window on `GEV_RATELIMIT_GOOGLE_PER_MIN`, separate from the Places and Geocoding one; over it, a frame gets the placeholder instead of a `429`. Typed chat keeps its own per-IP window, set by `GEV_RATELIMIT_OPENAI_CHAT_PER_MIN` and `GEV_RATELIMIT_GEMINI_CHAT_PER_MIN`; left unset, each takes its provider's resolved cap (`GEV_RATELIMIT_OPENAI_PER_MIN`, or `GEV_RATELIMIT_GEMINI_PER_MIN`, default `30`), so lowering a provider throttle lowers chat too, while an explicit chat value, including `0`, wins. Both defaults sit well above what the app itself generates; set either to exactly `0` to run unthrottled. They are **per-IP, process-local, in-memory guards** — they reset on restart, a caller with several addresses gets a bucket per address, and they are **not billing caps**.
- **Docker trusts one address, and only because the port is on loopback.**
Docker delivers the host's own browser to the container from the container
network's gateway, never from `127.0.0.1`, so a loopback-only Provider