unom/punktfunk
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases 8 Wiki Activity

feat(host,web): multi-GPU selection — GPU inventory + preference API, web-console GPU card
apple / swift (push) Successful in 1m9s
Details
ci / rust (push) Successful in 1m50s
Details
ci / web (push) Successful in 56s
Details
ci / docs-site (push) Successful in 57s
Details
decky / build-publish (push) Successful in 11s
Details
android / android (push) Successful in 3m13s
Details
apple / screenshots (push) Successful in 5m32s
Details
deb / build-publish (push) Successful in 3m15s
Details
docker / build-push (--build-arg FEDORA_VERSION=44, ci, ci/fedora-rpm.Dockerfile, punktfunk-fedora44-rpm) (push) Successful in 5s
Details
windows-host / package (push) Successful in 7m35s
Details
docker / build-push (., web/Dockerfile, punktfunk-web) (push) Successful in 31s
Details
docker / build-push (ci, ci/fedora-rpm.Dockerfile, punktfunk-fedora-rpm) (push) Successful in 4s
Details
docker / build-push (ci, ci/rust-ci.Dockerfile, punktfunk-rust-ci) (push) Successful in 4s
Details
docker / build-push (docs-site, docs-site/Dockerfile, punktfunk-docs) (push) Successful in 4s
Details
ci / bench (push) Successful in 4m53s
Details
rpm / build-publish (bazzite, punktfunk-fedora-rpm) (push) Successful in 9m58s
Details
rpm / build-publish (fedora-44, punktfunk-fedora44-rpm) (push) Successful in 9m54s
Details
docker / deploy-docs (push) Successful in 18s
Details

Browse Source 
- new crate::gpu (compiled on all platforms so the OpenAPI doc stays
  platform-independent): DXGI / sysfs GPU inventory with reboot-stable ids
  (PCI vendor:device + occurrence — LUIDs are per-boot), persisted auto/manual
  preference (<config>/gpu-settings.json, atomic temp+rename with in-memory
  rollback), one selection with precedence console preference >
  PUNKTFUNK_RENDER_ADAPTER > max VRAM and graceful fallback when the preferred
  GPU is absent, plus a live "in use" record (RAII session guard wrapped around
  every encoder open_video returns)
- fix: windows_gpu_vendor derived the encoder backend from DXGI adapter 0
  instead of the selected render adapter — on a hybrid box (e.g. Intel iGPU at
  index 0 + NVIDIA dGPU) the backend could disagree with the GPU the capture
  ring / IddCx render pin sit on. The NVENC 4:4:4 probe now also runs on the
  selected adapter (was: OS default), the codec/4:4:4 probe caches are keyed
  per selected GPU (were process-lifetime OnceLocks), and an explicit
  PUNKTFUNK_ENCODER conflicting with the selected GPU's vendor warns up front
- mgmt API: GET /api/v1/gpus (inventory + mode + preferred + next-session
  selection with reason + in-use GPU/backend/session-count) and
  PUT /api/v1/gpus/preference (validates mode/gpu_id before writing);
  openapi.json regenerated; the vdisplay render pin now also engages for a
  console preference (not just the env pin)
- web console: GPU card on the Host page — list with vendor + VRAM,
  Automatic / Prefer controls, Preferred / Next session / "In use · backend"
  badges, missing-preferred-GPU warning and env-pin note; en + de messages
- Linux: a matched manual preference picks the VAAPI render node and the
  NVENC-vs-VAAPI auto choice; auto mode is exactly the previous behavior

Validated live on the hybrid laptop (RTX 3500 Ada + Intel Arc Pro, which
enumerates twice — the occurrence ids disambiguate): enumerate, prefer,
bad-id 400, restart persistence, auto-restore keeping the stored pick.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Enrico Bühler
2026-07-02 13:57:18 +02:00
parent 40fefd73ca
commit 019f2677a7
17 changed files with 1881 additions and 200 deletions
Download Patch File Download Diff File Expand all files Collapse all files
api/openapi.json
+281
View File
@@ -138,6 +138,100 @@
}
}
},
"/api/v1/gpus": {
"get": {
"tags": [
"gpu"
],
"summary": "GPU inventory and selection",
"description": "Lists the host's hardware GPUs, the persisted auto/manual preference, the GPU the next session\nwill use (and why), and the GPU live sessions encode on right now.",
"operationId": "listGpus",
"responses": {
"200": {
"description": "GPU inventory + selection state",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/GpuState"
}
}
}
},
"401": {
"description": "Missing or invalid bearer token",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/ApiError"
}
}
}
}
}
}
},
"/api/v1/gpus/preference": {
"put": {
"tags": [
"gpu"
],
"summary": "Set the GPU preference",
"description": "`auto` restores automatic selection (`PUNKTFUNK_RENDER_ADAPTER` pin, else max dedicated VRAM);\n`manual` pins capture + encode to the given GPU. Persisted across restarts; applies to the\n**next** session (a running session keeps its GPU). If the preferred GPU is absent at session\nstart the host falls back to automatic selection rather than failing.",
"operationId": "setGpuPreference",
"requestBody": {
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/SetGpuPreference"
}
}
},
"required": true
},
"responses": {
"200": {
"description": "Preference stored; the new selection state",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/GpuState"
}
}
}
},
"400": {
"description": "Unknown mode, or `gpu_id` missing / not a listed GPU",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/ApiError"
}
}
}
},
"401": {
"description": "Missing or invalid bearer token",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/ApiError"
}
}
}
},
"500": {
"description": "Preference could not be persisted",
"content": {
"application/json": {
"schema": {
"$ref": "#/components/schemas/ApiError"
}
}
}
}
}
}
},
"/api/v1/health": {
"get": {
"tags": [
@@ -1373,6 +1467,40 @@
},
"components": {
"schemas": {
"ApiActiveGpu": {
"type": "object",
"description": "The GPU live sessions are encoding on right now.",
"required": [
"id",
"name",
"vendor",
"backend",
"sessions"
],
"properties": {
"backend": {
"type": "string",
"description": "The encode backend in use (`nvenc` | `amf` | `qsv` | `vaapi` | `software`)."
},
"id": {
"type": "string",
"description": "Stable id matching an entry of `gpus` (empty for the CPU/software encoder)."
},
"name": {
"type": "string"
},
"sessions": {
"type": "integer",
"format": "int32",
"description": "Number of live encode sessions on it.",
"minimum": 0
},
"vendor": {
"type": "string",
"description": "`nvidia` | `amd` | `intel` | `other`."
}
}
},
"ApiCodec": {
"type": "string",
"description": "Video codec identifier.",
@@ -1394,6 +1522,64 @@
}
}
},
"ApiGpu": {
"type": "object",
"description": "One hardware GPU on the host (software/WARP adapters are never listed).",
"required": [
"id",
"name",
"vendor",
"vram_mb"
],
"properties": {
"id": {
"type": "string",
"description": "Stable identifier (`vendorid-deviceid-occurrence`, hex PCI ids) — pass to `setGpuPreference`.\nStable across reboots and driver updates, unlike an adapter index or LUID.",
"example": "10de-2c05-0"
},
"name": {
"type": "string",
"description": "Adapter/marketing name.",
"example": "NVIDIA GeForce RTX 5070 Ti"
},
"vendor": {
"type": "string",
"description": "`nvidia` | `amd` | `intel` | `other`."
},
"vram_mb": {
"type": "integer",
"format": "int64",
"description": "Dedicated VRAM in MiB (0 where the platform doesn't expose it).",
"minimum": 0
}
}
},
"ApiSelectedGpu": {
"type": "object",
"description": "The GPU the **next** session's pipeline will be created on, and why. (A preference change\napplies to the next session; a running session keeps the GPU it opened on.)",
"required": [
"id",
"name",
"vendor",
"source"
],
"properties": {
"id": {
"type": "string"
},
"name": {
"type": "string"
},
"source": {
"type": "string",
"description": "Why this GPU was selected: `preference` (the manual choice), `env`\n(`PUNKTFUNK_RENDER_ADAPTER`), `auto` (max dedicated VRAM / platform default), or\n`preference_missing` (a manual choice is set but that GPU is absent — auto-selected\ninstead so the host keeps streaming)."
},
"vendor": {
"type": "string",
"description": "`nvidia` | `amd` | `intel` | `other`."
}
}
},
"ApprovePending": {
"type": "object",
"description": "Approve-pending-device request body. Send `{}` to keep the device's own name.",
@@ -1671,6 +1857,75 @@
}
}
},
"GpuState": {
"type": "object",
"description": "Full GPU-selection state for the console: inventory, the persisted preference, what the next\nsession will use, and what is in use right now.",
"required": [
"gpus",
"mode",
"preferred_available"
],
"properties": {
"active": {
"oneOf": [
{
"type": "null"
},
{
"$ref": "#/components/schemas/ApiActiveGpu",
"description": "The GPU live sessions use right now (absent while nothing is streaming)."
}
]
},
"env_override": {
"type": [
"string",
"null"
],
"description": "`PUNKTFUNK_RENDER_ADAPTER` (the host.env pin), when set — it applies while `mode` is\n`auto`; a manual preference overrides it."
},
"gpus": {
"type": "array",
"items": {
"$ref": "#/components/schemas/ApiGpu"
},
"description": "The host's hardware GPUs."
},
"mode": {
"type": "string",
"description": "`auto` or `manual`."
},
"preferred_available": {
"type": "boolean",
"description": "Whether the preferred GPU is currently present."
},
"preferred_id": {
"type": [
"string",
"null"
],
"description": "The manually preferred GPU's stable id, when one is stored (kept while `mode` is `auto` so\na console can offer returning to it). May reference a GPU that is currently absent."
},
"preferred_name": {
"type": [
"string",
"null"
],
"description": "The stored name of the preferred GPU (a usable label even when it is absent)."
},
"selected": {
"oneOf": [
{
"type": "null"
},
{
"$ref": "#/components/schemas/ApiSelectedGpu",
"description": "The GPU the next session will use."
}
]
}
}
},
"Health": {
"type": "object",
"description": "Liveness + version probe.",
@@ -2047,6 +2302,28 @@
}
}
},
"SetGpuPreference": {
"type": "object",
"description": "Request body for `setGpuPreference`.",
"required": [
"mode"
],
"properties": {
"gpu_id": {
"type": [
"string",
"null"
],
"description": "Required when `mode` is `manual`: the stable `id` of a currently listed GPU\n(see `listGpus`).",
"example": "10de-2c05-0"
},
"mode": {
"type": "string",
"description": "`auto` (env pin, else max dedicated VRAM — the default) or `manual`.",
"example": "manual"
}
}
},
"StageTiming": {
"type": "object",
"description": "One pipeline stage's latency in an aggregation window (microseconds).",
@@ -2267,6 +2544,10 @@
"name": "host",
"description": "Host identity, capabilities, and liveness"
},
{
"name": "gpu",
"description": "GPU inventory and selection: list the host's GPUs, choose automatic or a preferred GPU, see the one in use"
},
{
"name": "clients",
"description": "Paired Moonlight client management"