Adds a `-tensorrt-jp7` Frigate image for JetPack 7.2 / L4T R39.2 Jetson hosts,
built on nvcr.io/nvidia/tensorrt:26.02-py3-igpu (TensorRT 10.11, CUDA 13, py3.12),
keeping the existing JP6 path unchanged. ONNX Runtime GPU is built from source
(no public aarch64 onnxruntime-gpu wheel), TensorRT-Python branch is selected by
the base image, and TensorRT runtime library checks are major-version aware.
VALIDATED on a real AGX Orin (L4T R39.2 / nv_tegra_release R39 rev 2.0):
- image frigate:test-tensorrt-jp7 builds (rc=0, 16.3GB,
sha256:c9e4d382f1603ee130ee4a7315b4f71f9461405e3785707251505e2d6d088df3)
- ONNX Runtime 1.25.1 exposes CUDAExecutionProvider, and a real Add-model
inference RAN on the iGPU CUDA EP (functional, not just listed)
- frigate.util.model.get_ort_providers(False,"AUTO") = [CUDA, CPU] (CUDA first,
CPU last, TensorRT EP excluded) -> the ONNX detector GPU-accelerates on JP7
- /etc/TENSORRT_VER = 10.11.0
Native `type: tensorrt` (.trt gen) + the ORT TensorRT EP stay DRAFT-GATED: the
L4T R39 host ships NO libnvdla_compiler.so (absent from host AND base image), so
`import tensorrt` and libonnxruntime_providers_tensorrt.so fail to load. ONNX
detector GPU acceleration is the supported JP7 path; native trt is deferred.
Build fixes the new noble/CUDA-13 base surfaced (beyond the plan):
- build_nginx.sh: enable deb-src for the deb822 ubuntu.sources (Ubuntu 24.04)
- tensorrt_libyolo.sh: strip -lnvToolsExt (removed in CUDA 13) + -lnvparsers
(dropped in TensorRT 10) when those libs are absent
- docker/main: noble/py3.12 build adjustments (Dockerfile, build_sqlite_vec.sh)
Reproducible: `make -C docker/tensorrt ... local-trt-jp7` (JETPACK7_ARGS) on any
arm64 builder; built on-device only because the GPU smoke test needs the iGPU.
* swap sqlite_vec for chroma in requirements
* load sqlite_vec in embeddings manager
* remove chroma and revamp Embeddings class for sqlite_vec
* manual minilm onnx inference
* remove chroma in clip model
* migrate api from chroma to sqlite_vec
* migrate event cleanup from chroma to sqlite_vec
* migrate embedding maintainer from chroma to sqlite_vec
* genai description for sqlite_vec
* load sqlite_vec in main thread db
* extend the SqliteQueueDatabase class and use peewee db.execute_sql
* search with Event type for similarity
* fix similarity search
* install and add comment about transformers
* fix normalization
* add id filter
* clean up
* clean up
* fully remove chroma and add transformers env var
* readd uvicorn for fastapi
* readd tokenizer parallelism env var
* remove chroma from docs
* remove chroma from UI
* try removing custom pysqlite3 build
* hard code limit
* optimize queries
* revert explore query
* fix query
* keep building pysqlite3
* single pass fetch and process
* remove unnecessary re-embed
* update deps
* move SqliteVecQueueDatabase to db directory
* make search thumbnail take up full size of results box
* improve typing
* improve model downloading and add status screen
* daemon downloading thread
* catch case when semantic search is disabled
* fix typing
* build sqlite_vec from source
* resolve conflict
* file permissions
* try build deps
* remove sources
* sources
* fix thread start
* include git in build
* reorder embeddings after detectors are started
* build with sqlite amalgamation
* non-platform specific
* use wget instead of curl
* remove unzip -d
* remove sqlite_vec from requirements and load the compiled version
* fix build
* avoid race in db connection
* add scale_factor and bias to description zscore normalization
Jorge Ramirez