serve(runtime) listens first, writes the state root only once the runtime
holds it, answers Idunn's route challenge ahead of the cross-site refusal,
and fences on lease loss. --health is deleted; the pid file and /proc
reclaim move behind the hand-run runtime. Adds scripts/idunn-entry.mjs,
the kimodo and SOMA-X gitlinks, encode_texts for the child, and tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The header-on vector now matches Kimodo's encoder loaded from the repo ids (cosine 1.0; the bare header gave
0.9974) because the wrapped text carries LLM2Vec's instruction/query separator, which keeps the header out of
the pooling. live_identity reads the installed versions itself, so the worker passes no version it could fake,
and the tests pin installed_versions against substituted modules, the prompt form and device in the identity,
and a text that differs past any prefix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The child wraps every text with text_encoding.chat_header and tokenizes it itself, so the prompt form no
longer rides on the model's _name_or_path. The identity names the assembly truthfully (header form
included, as a hash of the template) and re-keys the cache. The worker builds its identity through
live_identity and installed_versions, which the tests reach; the identity test now uses live-length
revisions, so a truncated identity or key fails. An env-gated test compares a fresh child vector with
Kimodo's own encoder loaded from the repo ids.
Supersedes encoder-assembly-b (motion-pipeline:ruling:encoder-chat-header-on).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured on Yggdrasil: loading by repo id leaves the model's _name_or_path
equal to the Llama repo id, so prepare_for_tokenization wraps the text in
the chat header and the vector matches assembly A (cosine 1.0, 0.909 to the
reference B). Loading from the snapshot paths reproduces the reference.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The encoder child is CPU only (GPU placement, GpuBusy and plan_placement
are deleted) and assembles LLM2Vec B: Llama base, MNTP merged, then the
supervised adapter. The cache identity names assembly, device, dtype,
torch/transformers/peft/accelerate versions, Kimodo commit and the child's
own sha256 beside the three revisions. VILI_THREADS (1..16) caps torch in
the worker and the child. The worker binds port 0 and reports it in its
ready line; errors name a class and place, never the request.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Replaces the 2026-06-12 timing paragraph with the cache, the child, and the numbers
measured on Raven on 2026-10-22. Adds the tests hand mutation showed missing: the
exact 2 GiB boundary, a child answer with too few vectors, and the 4096 dimension.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
text_encoding.py holds the CultCache embedding store (gamecult.vili.text_embedding.v0
through cultcache-py), the encoder identity and key, plan_placement with its typed
gpu-busy refusal, and CachedTextEncoder, which sends a request's misses to one child
call. kimodo-text-encode.py is that child: it assembles Kimodo's LLM2Vec encoder in
bf16 with its own loader (operator ruling encoder-own-loader; peft's offload path
raises TypeError in Kimodo's loader), encodes each text alone, writes a MessagePack
answer and exits. GPU reserve 3 GiB: four cold runs on Raven's 16 GB card produced
identical vectors with no CUDA error. The worker and daemon are not wired yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
