pin v2 payload: engine 2610041105001, cuda 2610040656001, cuda12 2610040730001, sbsa 2610040704001, vulkan 2610041121001, macos 2610041000001, media 2610040945001

.gitattributes

@@ -226,3 +226,7 @@ components/media/2610040945001/oc-espeak-linux-x86_64.so filter=lfs diff=lfs mer
 components/media/2610040945001/oc-espeak-linux-aarch64.so filter=lfs diff=lfs merge=lfs -text
 components/media/2610040945001/oc-espeak-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
 components/media/2610040945001/oc-espeak-macos-aarch64.dylib filter=lfs diff=lfs merge=lfs -text
+components/engine/2610041105001/opencoti-0.10.5-c7-2610041105001 filter=lfs diff=lfs merge=lfs -text
+components/engine/2610041105001/opencoti-0.10.5-c7-2610041105001.exe filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610041121001/ggml-vulkan-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610041121001/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text

components/engine/2610041105001/BUILD_INFO.engine.md

@@ -0,0 +1,53 @@
+# engine 2610041105001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `opencoti-0.10.5-c7-2610041105001` | any | bin | `683f182805dfaf90a8743be95db24beab65eb538fa3399cf1b9d1bdc5d95517d` | 134675275 |
+| `opencoti-0.10.5-c7-2610041105001.exe` | win-x86_64 | bin | `683f182805dfaf90a8743be95db24beab65eb538fa3399cf1b9d1bdc5d95517d` | 134675275 |
+
+ABI `ggml` = `41ba479c24d7e3a624c7b13aa3cbe357e41fac04a504c54ec2ee170de0bff9a1` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h af6dc423ee7c6c3f67a6a187d6dfbfeb94e52ddc1db27c45ee7440289bc44282
+llama.cpp/ggml/include/ggml-vulkan.h 7eae5dad2cc7bb4d3eca828539f441816d5fd59fdc7b224d49aa229fc7b7248c
+llama.cpp/ggml/include/ggml.h f2bb8b127e3e601a78fe32242ffab7235e6a4db3bab5370b72fb6bacba2b2433
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h d17c504fbab50ff9376105a1f2c79d99fcde4b431886b17b90718a2a359278de
+```
+
+ABI `media` = `b099409b3029a6e380268e31d1caefa6706569961d32e72b3543fbc73903d95c` (exported) — sha256 of this list:
+
+```
+llamafile/oc-audiocpp/audiocpp.h 6ad0b55d5f28adc9ad60a45bf9a7caa428c4f5441389a6932d6670a8e2a90d73
+llamafile/oc-codec/oc_codec_sidecar.h f2c9001d1414aa2a55533b867b79c52a52c682558e0234c5bebd30bf97b91005
+```
+
+ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc` (exported) — sha256 of this list:
+
+```
+.cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
+```
+
+The APE: one file for Linux x86_64 / aarch64, Windows and macOS arm64; the `.exe` row is the same bytes under the name Windows needs.
+
+Chain through 0555 graceful-shutdown (bug-3919; docs/features/server_shutdown.md), on top of 0554 abi-export:
+- `POST /shutdown` — loopback only (403 otherwise), 503 while the model is still loading, behind the API key when one is set. The same path as SIGTERM. Feature `graceful_shutdown_v1`.
+- The stop path now frees the media engines, the models and the contexts (`shutdown: models and contexts freed`), then for every Vulkan device waits for it to be idle and destroys it (`ggml_vulkan: shutdown: Vulkan0 idle and destroyed`), then the instance (`ggml_vulkan: shutdown: N device(s) destroyed, 0 kept, instance destroyed`), exit 0. Before 0555 the server left with the model allocated and every device alive, on a clean stop as on a kill.
+- Deadline `OPENCOTI_SHUTDOWN_DEADLINE_MS` (default 10000, 0 = none) from the moment the task loop stops: `the teardown did not finish within the deadline, exiting now`, exit 0.
+- Parent watch: `OPENCOTI_PARENT_PID=<pid>` (on Windows a Windows process id) — `parent watch: armed on pid N`; when that process is gone, `shutdown: parent pid N is gone` and the same stop path. `NOT armed` when the pid cannot be watched. Feature `parent_watch_v1`.
+- `--list-devices` destroys the Vulkan instance before it exits; `OPENCOTI_LIST_DEVICE_IDS=1` appends ` id=<PCI address | uuid:<32 hex>>` to each line (the stock line is unchanged without it). Feature `list_device_ids_v1`.
+- The Vulkan teardown needs the `vulkan` component of this index (2610041121001); with an older Vulkan library the engine says `shutdown: the library predates the device teardown (0555): devices NOT destroyed`.
+No interface header changed: the `ggml`, `media` and `loader` digests are those of engine 2610040950001, so every other component of the previous index fits unchanged.
+
+Gated on this exact file: Linux x86_64 (bs2), `.opencoti/graceful-shutdown-gate.sh` 12/12 — POST /shutdown and SIGTERM on Vulkan (exit 0 in 1.6-3.5 s, device and instance destroyed), non-loopback 403, stop during a stream (the stream ends with a 503 `server is shutting down` event), deadline, parent watch armed / not armed, CUDA POST /shutdown, list-devices teardown and ids. NOT run on this file: Windows and macOS (the previous engine, 2610040950001, passed macOS 21/21 and Windows 30/30; this one differs by 0555 only), Linux aarch64, the bs2 2M-context swarm.
+
+Image / video features are listed for the platforms they were gated on.

components/engine/2610041105001/opencoti-0.10.5-c7-2610041105001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:683f182805dfaf90a8743be95db24beab65eb538fa3399cf1b9d1bdc5d95517d
+size 134675275

components/engine/2610041105001/opencoti-0.10.5-c7-2610041105001.exe

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:683f182805dfaf90a8743be95db24beab65eb538fa3399cf1b9d1bdc5d95517d
+size 134675275

components/macos/2610041000001/BUILD_INFO.macos.md

@@ -32,6 +32,8 @@ ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc
 .cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
 ```
 
+Apple notarisation (Accepted), submission id(s): `eea7c66d-e8a4-4860-b9da-71e9416a2805 7b83b4fc-da94-4dec-a829-43c8cc89e494`.
+
 The prebuilt APE loader (unchanged bytes) and the prebuilt Metal library, built on the Mac by engine 2610040950001 from its embedded sources: it states `oc-abi ggml`. **Start the engine as `ape-macos-aarch64 <engine> <args…>`.** Developer ID signed, hardened runtime; both accepted by Apple's notary service.
 
 Gated: Mac mini (macOS 27), release gate 21/21, fresh HOME and TMPDIR, no compiler on PATH; `metal: abi ggml 41ba479c24d7 matches this engine`.

components/media/2610040945001/BUILD_INFO.media.md

@@ -30,6 +30,8 @@ llamafile/oc-audiocpp/audiocpp.h 6ad0b55d5f28adc9ad60a45bf9a7caa428c4f5441389a69
 llamafile/oc-codec/oc_codec_sidecar.h f2c9001d1414aa2a55533b867b79c52a52c682558e0234c5bebd30bf97b91005
 ```
 
+Apple notarisation (Accepted), submission id(s): `aa53f9d9-14ba-443a-a059-64f82416f86a`.
+
 Twelve sidecars, three per platform, ALL rebuilt: each states the `media` interface in its bytes (`oc_codec_abi_media()`, `audiocpp_oc_abi_media()`, and for `oc-espeak-*` — a file the engine reads, it does not load it — a 96-byte block `oc-abi media <digest>` right before its data block). An engine from 2610040950001 on checks them at load; older engines ignore the statement and use the files as before (not gated here: `engine-min` is 2610040950001).
 
 Take the three of a platform as a set: `oc-espeak-*` (eSpeak-ng 1.52.0, GPL-3.0-or-later, a separate file) works only with the `oc-audiocpp-*` of this component.

components/vulkan/2610041121001/BUILD_INFO.vulkan.md

@@ -0,0 +1,31 @@
+# vulkan 2610041121001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610041105001; gated floor (engine-min) 2610041105001; gated with 2610041105001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-vulkan-x86_64.so` | x86_64 | vulkan | `a176bdc8b76222183ef57d73cb2d0641628322ccb3a5799e991ec0d61c6b801e` | 56290696 |
+| `ggml-vulkan-win-x86_64.dll` | win-x86_64 | vulkan | `5ea74840db7930452e08ae6dde4937c35ca8e9125b19aaf3e661db004a1ee60f` | 57255285 |
+
+ABI `ggml` = `41ba479c24d7e3a624c7b13aa3cbe357e41fac04a504c54ec2ee170de0bff9a1` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h af6dc423ee7c6c3f67a6a187d6dfbfeb94e52ddc1db27c45ee7440289bc44282
+llama.cpp/ggml/include/ggml-vulkan.h 7eae5dad2cc7bb4d3eca828539f441816d5fd59fdc7b224d49aa229fc7b7248c
+llama.cpp/ggml/include/ggml.h f2bb8b127e3e601a78fe32242ffab7235e6a4db3bab5370b72fb6bacba2b2433
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h d17c504fbab50ff9376105a1f2c79d99fcde4b431886b17b90718a2a359278de
+```
+
+Built from the 0555 tree: states the `ggml` interface in its bytes (the engine prints `abi ggml … matches this engine`), and exports `ggml_backend_vk_oc_shutdown` — the teardown the engine calls on every normal exit (wait for each device to be idle, destroy it, destroy the instance). A Vulkan device's id is its PCI address, or `uuid:<deviceUUID>` where the driver has no `VK_EXT_pci_bus_info` (Windows). Carries 0551 vk-adopt-stack as before.
+
+Gated: Linux x86_64 on bs2 (NVIDIA, Vulkan) with engine 2610041105001 — the graceful-shutdown gate 12/12. The Windows DLL (mingw cross lane) has NOT been run: its bytes carry the teardown symbol and the interface statement, nothing more is claimed. The previous Windows DLL (component 2610040653001) was gated on an RTX 3090, an RX 9070 XT and a Radeon integrated GPU.

components/vulkan/2610041121001/ggml-vulkan-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ea74840db7930452e08ae6dde4937c35ca8e9125b19aaf3e661db004a1ee60f
+size 57255285

components/vulkan/2610041121001/ggml-vulkan-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a176bdc8b76222183ef57d73cb2d0641628322ccb3a5799e991ec0d61c6b801e
+size 56290696