BigMoeOnEdge/tests
Raffaele e677c1c5e1
feat(models): a model's name is enough, the Hugging Face field searches (#282)
Typing owner/name was too rigid: a person knows a model's name, not who
published it. The one field now takes either. Text that is not a
repository id or URL is refused by /api/hf/models with its own code,
not_a_repo, and the UI then asks /api/hf/search: the site's repositories
with gguf files whose name matches, the most downloaded first, twenty at
most, a gated one marked. The repositories found are offered to pick from,
and picking one lists it as before. Web UI and app alike.

The query is percent-encoded by the server (hf_search_url) and goes to the
one endpoint the server was started with; the answer is read by
parse_hf_search, which drops an entry without an id or with one this
server could not list. Both tested.
2026-10-18 14:47:00 +02:00
..
sanitizers refactor: no hardcoding, no regex, the duplication removed and the server and app hardened (#262) 2026-10-18 10:49:39 +02:00
chat_messages_test.cpp build: llama.cpp to upstream master of 2026-10-18 (#278) 2026-10-18 13:28:31 +02:00
chat_parse_test.cpp build: llama.cpp to upstream master of 2026-10-18 (#278) 2026-10-18 13:28:31 +02:00
CMakeLists.txt feat(build): compute backends as modules beside the engine, on every OS (#263) 2026-10-18 12:15:28 +02:00
config_test.cpp feat(prefill): the NPU prefill reads only routed experts (+ --decide-probe); 0.28.0 (#208) 2026-09-29 16:08:23 +02:00
csv_summary_test.cpp refactor: no hardcoding, no regex, the duplication removed and the server and app hardened (#262) 2026-10-18 10:49:39 +02:00
decide_test.cpp feat: --decide, choose from a list from one prefill, with no decode (#201) 2026-09-28 12:01:25 +02:00
engine_host_test.cpp refactor: no hardcoding, no regex, the duplication removed and the server and app hardened (#262) 2026-10-18 10:49:39 +02:00
expert_layout_test.cpp refactor(moe): which tensors are the routed experts is read from the file, not from a table of architectures (#266) 2026-10-18 11:10:38 +02:00
mapping_release_test.cpp feat(io): release the model file's mapping after load (--release-mmap) (#185) 2026-09-07 20:43:02 +02:00
moe_gates.cpp build: llama.cpp to upstream master of 2026-10-18 (#278) 2026-10-18 13:28:31 +02:00
ngram_test.cpp feat(engine): self-speculative decoding — the model's own MTP head, or n-gram lookup (#134) 2026-08-02 00:09:39 +02:00
params_test.cpp refactor(moe): which tensors are the routed experts is read from the file, not from a table of architectures (#266) 2026-10-18 11:10:38 +02:00
planner_test.cpp fix(plan): the host's share of the experts counts the blocks that carry them (#280) 2026-10-18 13:47:24 +02:00
server_test.cpp feat(models): a model's name is enough, the Hugging Face field searches (#282) 2026-10-18 14:47:00 +02:00
stall_union_test.cpp fix(telemetry): attribute critical-path stall explicitly (#169) 2026-08-27 17:45:51 +02:00
tensor_names_test.cpp refactor(moe): which tensors are the routed experts is read from the file, not from a table of architectures (#266) 2026-10-18 11:10:38 +02:00
think_control_test.cpp build: llama.cpp to upstream master of 2026-10-18 (#278) 2026-10-18 13:28:31 +02:00