ark by ptr.pet · Tangled

Commits on buer

buer: bring the 1 GB headscale VPS into

committed 6d ago

focalor: strata's vision encoder runs on the RTX 4060 (CUDA build for sm_89)

committed 1w ago

focalor: strata no longer needs /strata-fast mounted

committed 1w ago

focalor: strata back on Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS with niri's think1, 200K context, served as deepseek-v4.1-flash

committed 1w ago

tangled: master

committed 1w ago

trimounts: knot blob

committed 1w ago

trimounts: knot on tangled core master (1.0), admission

committed 1w ago

valefar: pia-exit tries every DC ovpnudp server, not one pinned

committed 1w ago

valefar: spindle allows cross-origin

committed 1w ago

valefar: spindle on tangled 1.0, served as

committed 1w ago

valefar: swap to zram first, the NVMe partition only as overflow

committed 1w ago

valefar: spindle metrics on the tailnet address for victoriametrics

committed 1w ago

focalor: strata keeps and splices conversations with pictures (niri back after another request: 94 s -> 0.4 s), slot folder 16 GB

committed 1w ago

focalor: nvidia-smi patched with patchelf 0.15 (the driver's copy segfaults, NixOS/patchelf#152)

committed 1w ago

focalor: strata streams an empty delta every 15 s while reading a long prompt (niri's harness idle timeout ignores SSE comments)

committed 1w ago

focalor: strata context 320K (niri's history is 224K tokens)

committed 1w ago

focalor: strata DeepSeek-V4.1 sees images (vision tower on the 4060)

committed 1w ago

focalor: strata keeps the 4060 expert GPU and its cache (3b5d464 dropped them by mistake)

committed 1w ago

focalor: strata Cache-Prior over the 5070 and 4060 only, swaps at window ticks (63K decode 10.9 -> 12.0 tok/s, 16K 12.7 -> 13.5, deterministic)

committed 1w ago

focalor: strata Cache-Prior favours RAM-store and 4060 experts too (63K decode 10.0 -> 10.9 tok/s, misses -40%, KL no worse)

committed 1w ago

focalor: strata 4060 experts follow the conversation (63K-context decode 9.0 -> 10.0 tok/s, 16K 10.9 -> 12.3)

committed 1w ago

focalor: strata SCR keeps conversation slots (a compaction after another conversation 79 -> 16 s, restart restores from disk)

committed 1w ago

focalor: strata reuses edited prompts (SCR: 63K compaction 77 -> 16.5 s, continuation KL ~0.015), exits on QUIT at once

committed 1w ago

focalor: strata Cache-Prior 0.15 (perplexity-neutral), the 4060 as expert GPU, faster prompts (63.5K cold 116 -> 70 s); prose 13.7, code 21.3 tok/s

committed 1w ago

focalor: strata suffix drafts up to 7 tokens (code 20.4 -> 22.2 tok/s, exact), store keeps failed pinned slots for their last holder

committed 1w ago

focalor: strata DeepSeek adds in-place store reads, top-3 read-ahead and suffix drafting (code rewrites 12.5 -> 20.4 tok/s, exact)

committed 1w ago

focalor: strata serves DeepSeek-V4.1-Flash (ISTA GSQ-RCO IQ2_XXS) on its own engine, AVX2 build, 200K context, Cache-Prior 0.3

committed 1w ago

chernobog: build udiskie without its keyring chernobog-openssh

committed 1w ago

nodejs-security: patched node as separate attributes, not a global

committed 1w ago

chernobog: sign the hand-installed kernel modules with

committed 1w ago

valefar: bonsai's NVMe prompt cache tier is 20 GB

committed 1w ago

chernobog: drop the depmod index files v4l2loopback 0.15.4

committed 1w ago

chernobog: drop

committed 1w ago

dzwonek: rate-limit headscale's unauthenticated paths, cap its

committed 1w ago

chernobog: OpenSSH for mayer instead of Tailscale

committed 1w ago

valefar: the nixos-large CI image gets 12 vCPUs

committed 1w ago

valefar: spindle microVM overlays on the NVMe

committed 1w ago

valefar: spindle microVM overlays on a sync=disabled dataset

committed 1w ago

valefar: a 12 GiB / 8 vCPU nixos-large CI image, bonsai keeps 8 GB of prompt cache in RAM, qBittorrent's page cache stops at 2 GB

committed 1w ago

focalor: strata serves niri on Swift 1.5 GSQ-RCO IQ3_S (held-out NLL 2.394 -> 2.344, decode 65 -> 51 t/s)

committed 1w ago

focalor: strata decodes without the routing-aware lookahead (+3% decode)

committed 1w ago

focalor: strata skips GPU-missed experts ranked 5th or lower (decode +10%)

committed 1w ago

valefar: pia-exit tunnel back to OpenVPN

committed 1w ago

valefar: cap the CPU at 3.8 GHz instead of

committed 1w ago

focalor: strata serves niri's think1 merged into Swift 1.5 (median thinking 924 -> 646 characters)

committed 2w ago

focalor: 128 KiB read-ahead on the strata array

committed 2w ago

focalor: drop the depmod index files v4l2loopback 0.15.4 ships, which clash with the kernel's

committed 2w ago

focalor: strata reads from a RAID0 over both NVMes, /nix moves to the SN7100, zram swap

committed 2w ago

valefar: nightly encrypted offsite backup to

committed 2w ago

dzwonek: lisya may use exit nodes

committed 2w ago

valefar: pia-exit clamps MSS and lets tailscaled's UDP out

committed 2w ago

nodejs: October 2026 security release (Brotli RCE

committed 2w ago

valefar: nvidia-smi patched with patchelf 0.15, which keeps .dynamic writable

committed 2w ago

improve deploy script

committed 2w ago

valefar: fetch searxng with

committed 2w ago

fix js service ca trust

committed 2w ago

disable inverter, block emudhra ca

committed 2w ago

chore(nix): update inputs [skip ci]

committed 2w ago

awa

committed 2w ago

focalor: niri matches the MSI monitor on the RTX 4060 too

committed 2w ago