Frank 8d88b3a0a9 frank-small: int8 KV (per-token, per-head) — doubles the pool ...

fp8 KV is hardware-dead on the 3090 (sm_86, no cvt.e4m3x2), so int8 is
the cheapest quant the TRITON_ATTN backend can run: pure ALU dequant,
same backend, no fp8 tensor cores needed. 289,647 -> ~579k-token pool,
two -> four full-context sessions. Quality verified against the bf16
baseline (needle-in-haystack at 213k + short battery) before the swap.

2026-09-18 23:24:37 -07:00

..

.AGENTTRASH

frank-medium: llama.cpp layer split across both R9700s, DFlash2 drafting

2026-09-16 14:59:23 -07:00

patches

frank-default-large: re-enable --flash-attn on via QSA F32-mask patch

2026-08-29 23:43:25 -07:00

bench_tb_fixed.py

docs: rename docs/ -> serverAppConfigs-docs/ per system-wide PROJECTNAME-docs convention

2026-08-28 02:20:06 -07:00

bench-batch-ubatch.sh

scripts: fix the unbound local that killed the sweep at its first cell

2026-09-16 21:43:09 -07:00

bench-decode.sh

feat(llm): rename inference services dropping 'default' from names

2026-09-03 16:15:43 -07:00

bench-llm-soak.py

bench: commit the harnesses the frank-medium numbers came from

2026-09-16 06:59:44 -07:00

bench-spec-ab.py

bench-spec-ab: rewrite (write_env was lost), restore the unit in a finally

2026-09-16 16:38:12 -07:00

bench-threads-batch-http.py

feat(llm): rename inference services dropping 'default' from names

2026-09-03 16:15:43 -07:00

bench-threads-batch.sh

feat(llm): rename inference services dropping 'default' from names

2026-09-03 16:15:43 -07:00

bench-turn-latency.py

frank-medium: the agent-loop shape on one slot, and a harness header that matches the runtime

2026-09-16 15:38:20 -07:00

bench-ubatch.sh

feat(llm): rename inference services dropping 'default' from names

2026-09-03 16:15:43 -07:00

build-llama-rdna4.sh

build: bundle the rocBLAS device libraries at the depth the runtime looks for

2026-09-16 15:15:02 -07:00

check-vault-links.py

docs: three dangling links, and a gate so they cannot accumulate

2026-09-16 07:43:33 -07:00

deploy-open-webui-quadlet.sh

Deploy services independently by changed paths

2026-08-25 20:13:33 -07:00

deploy-pihole-quadlet.sh

Deploy services independently by changed paths

2026-08-25 20:13:33 -07:00

deploy-plan.py

frank-small: llama.cpp -> vLLM (gemma-4-12B-it-qat-w4a16-ct, w4a16 Marlin)

2026-09-18 21:06:46 -07:00

deploy-remote.sh

frank-small: llama.cpp -> vLLM (gemma-4-12B-it-qat-w4a16-ct, w4a16 Marlin)

2026-09-18 21:06:46 -07:00

deploy-searxng-quadlet.sh

Deploy services independently by changed paths

2026-08-25 20:13:33 -07:00

deploy.sh

alerting: Alertmanager + Signal relay deliver firing rules to Frank's phone

2026-09-17 13:09:59 -07:00

download-frank-medium-gguf.sh

frank-medium: 36:29 split, 327680 context, vision enabled

2026-09-17 01:37:55 -07:00

download-frank-small-vllm.sh

frank-small: llama.cpp -> vLLM (gemma-4-12B-it-qat-w4a16-ct, w4a16 Marlin)

2026-09-18 21:06:46 -07:00

frank-medium.env

frank-medium: double the host-RAM prompt cache to 16 GiB

2026-09-18 15:25:55 -07:00

frank-medium.service

frank-medium: llama.cpp layer split across both R9700s, DFlash2 drafting

2026-09-16 14:59:23 -07:00

frank-small.env

frank-small: int8 KV (per-token, per-head) — doubles the pool

2026-09-18 23:24:37 -07:00

generate-prompts.py

docs: rename docs/ -> serverAppConfigs-docs/ per system-wide PROJECTNAME-docs convention

2026-08-28 02:20:06 -07:00

homepage-status.py

Homepage (TDD slice 7): fix the /admin/ path + status.json perms

2026-08-23 16:49:33 -07:00

homepage-status.service

Homepage (TDD slice 6): the git-driven deploy

2026-08-23 12:59:56 -07:00

homepage-status.timer

Homepage (TDD slice 6): the git-driven deploy

2026-08-23 12:59:56 -07:00

mount-entropy.sh

Add CLI helper for entropy volume

2026-08-25 15:38:57 -07:00

serve-frank-medium.sh

frank-medium: pass --mmproj explicitly; this build has no autoload

2026-09-17 01:54:26 -07:00

serve-frank-small.sh

frank-small: llama.cpp -> vLLM (gemma-4-12B-it-qat-w4a16-ct, w4a16 Marlin)

2026-09-18 21:06:46 -07:00

signal-relay.py

relay: restore SIGNAL_CLI constant dropped by a boundary-repaired edit

2026-09-17 18:38:54 -07:00

signal-relay.service

alerting: Alertmanager + Signal relay deliver firing rules to Frank's phone

2026-09-17 13:09:59 -07:00

test-changed.sh

frank-medium: llama.cpp layer split across both R9700s, DFlash2 drafting

2026-09-16 14:59:23 -07:00

tmux-resurrect.sh

feat: add tmux session snapshot and restore utility

2026-08-25 17:04:15 -07:00