AnyJev operator
This skill targets the pinned AnyJev checkout at commit
d458d13968eec0e2cfa8c81b64629b96cf5cb4a9 (package 0.3.0, audited
2026-10-07). Detailed commands, limits, backend contracts, and citations live
in references/upstream.md.
When to use this skill
- Inspect or pin an AnyJev checkout without downloading weights or datasets.
- Build a training-free
Decider readout and distinguish raw from L0.
- Run an approved Tacit checkpoint with transformers, vLLM, or a vLLM server.
- Configure HF, vLLM, SGLang, or GGUF/llama.cpp backends and tokenizer rules.
- Evaluate adaptive rotation calibration or the local
anyjev.serve gateway.
- Record a bounded research readout with model, backend, and source evidence.
Use external TypeSafe AI documentation or its own skill for hosted System
One/Jev APIs, credentials, billing, or a separate Jev harness. AnyJev is an
independent self-hosted open-model library; its manifest has no typesafe-ai
dependency.
Instructions
1. Choose one route
Select exactly one route before changing anything: inspect, decider,
tacit, backend, gateway, or research. Read the corresponding pinned
details in references/upstream.md; do not infer newer main behavior.
2. Run the offline preflight first
From the repository root:
python3 .agent-skills/anyjev/scripts/preflight.py --json
From an installed skills root:
python3 "$SKILLS_ROOT/anyjev/scripts/preflight.py" --json
Add --repo PATH --expect-commit d458d13968eec0e2cfa8c81b64629b96cf5cb4a9
to inspect a literal local checkout path. The helper only reads Python and
distribution metadata, host/tool presence, public checkout shape, and a local
Git commit. runtime_verified=false is always honest: detected packages are
not working inference, activated servers, loaded models, or quality evidence.
Optional backend extras may be absent; base AnyJev still requires Python
>=3.10, AnyJev metadata, and NumPy >=1.24.
3. Execute the approved route
Install only the selected optional extra and make model, dataset, server, and
credential side effects explicit. Preserve the Question option/level
contracts, require L0 when it is part of the acceptance condition, and never
describe Tacit's one-forward output as an L0-averaged result. For HTTP
backends, verify tokenizer availability and required log-probability settings;
keep the gateway loopback-bound unless external authentication and network
policy are in place.
4. Verify and report
For no-download API plumbing, the pinned FakeBackend smoke path is allowed
but is fixture evidence, not model-quality evidence. For real inference report
the AnyJev commit, model revision, backend/engine, tokenizer, dtype/device,
question layout, prior/adaptive settings, and dataset license. Report package
metadata and PATH detections separately from successful inference.
Examples
- Inspect only: run the offline preflight with
--repo and the pinned
--expect-commit; then read the pinned metadata and limitations.
- Decider: use a causal-LM backend and
Decider(level="L0"); inspect the
returned distribution and explicit level before acting.
- Tacit service: approve the checkpoint, vLLM engine, loopback endpoint,
and compute first; then verify
/health, /v1/stats, and one /v1/decide.
- No-download smoke: use
FakeBackend only to exercise prompt/readout
plumbing and state clearly that it says nothing about accuracy.
Best practices
- Keep model/dataset downloads, optional installs, remote-server calls, and
provider credentials opt-in; local inference still consumes compute.
- Prefer
Decider(level="L0") when option-order bias matters and enforce the
level; choose prior="content_free" or prior="none" deliberately.
- Calibrate adaptive shifts with
canonical_order=True; the certificate is
agreement with full-K AnyJev output, not external accuracy.
- Treat Decider
answer_mass as prompt-landed evidence; treat Tacit's margin
and route as diagnostics, not universal confidence; keep its escalation cap
enabled for shared services.
- Treat GGUF/Hugging Face tokenizer disagreement as a safe failure and keep
non-loopback gateways behind authentication and network policy.
- Preserve Apache-2.0 notices and identify benchmark dataset licenses; do not
generalize isolated decision evidence into agent-loop claims.
References
references/upstream.md — pinned commands, backend
contracts, tokenizer/calibration limits, gateway boundary, and source index.
scripts/preflight.py — offline metadata and checkout
shape/commit preflight; it never imports AnyJev or invokes providers.
- AnyJev pinned repository
- AnyJev Apache-2.0 license
1---2name: anyjev3description: >4 Operate nokia-applied-research/AnyJev, the Apache-2.0 Python library for5 training-free typed choice, yes/no, and score decisions from next-token6 probabilities. Use when pinning or inspecting a checkout, selecting raw/L07 Decider readouts, running approved Tacit checkpoints, configuring HF/vLLM/8 SGLang/llama.cpp backends, checking tokenizer or calibration limits, or using9 the local gateway. Triggers on AnyJev, Tacit, L0, permutation debiasing,10 llama.cpp logprob, SGLang /generate, or comparisons with hosted TypeSafe11 AI/Jev. Do not use for TypeSafe hosted services or a generic Jev harness.12allowed-tools: Bash Read Write Edit Glob Grep13compatibility: >14 Python >=3.10. Model weights and datasets are opt-in downloads. HF/transformers,15 vLLM, SGLang, and llama.cpp have separate runtime and hardware requirements;16 remote servers and model providers may require credentials or incur compute cost.17license: Apache-2.018metadata:19 version: "1.1.0"20 source: https://skillmd.com/skills/akillness/anyjev21 upstream: https://github.com/nokia-applied-research/AnyJev22 upstream_commit: d458d13968eec0e2cfa8c81b64629b96cf5cb4a923---2425# AnyJev operator2627This skill targets the pinned AnyJev checkout at commit28`d458d13968eec0e2cfa8c81b64629b96cf5cb4a9` (package `0.3.0`, audited292026-10-07). Detailed commands, limits, backend contracts, and citations live30in [`references/upstream.md`](references/upstream.md).3132## When to use this skill3334- Inspect or pin an AnyJev checkout without downloading weights or datasets.35- Build a training-free `Decider` readout and distinguish `raw` from `L0`.36- Run an approved Tacit checkpoint with transformers, vLLM, or a vLLM server.37- Configure HF, vLLM, SGLang, or GGUF/llama.cpp backends and tokenizer rules.38- Evaluate adaptive rotation calibration or the local `anyjev.serve` gateway.39- Record a bounded research readout with model, backend, and source evidence.4041Use external TypeSafe AI documentation or its own skill for hosted System42One/Jev APIs, credentials, billing, or a separate Jev harness. AnyJev is an43independent self-hosted open-model library; its manifest has no `typesafe-ai`44dependency.4546## Instructions4748### 1. Choose one route4950Select exactly one route before changing anything: `inspect`, `decider`,51`tacit`, `backend`, `gateway`, or `research`. Read the corresponding pinned52details in `references/upstream.md`; do not infer newer `main` behavior.5354### 2. Run the offline preflight first5556From the repository root:5758```bash59python3 .agent-skills/anyjev/scripts/preflight.py --json60```6162From an installed skills root:6364```bash65python3 "$SKILLS_ROOT/anyjev/scripts/preflight.py" --json66```6768Add `--repo PATH --expect-commit d458d13968eec0e2cfa8c81b64629b96cf5cb4a9`69to inspect a literal local checkout path. The helper only reads Python and70distribution metadata, host/tool presence, public checkout shape, and a local71Git commit. `runtime_verified=false` is always honest: detected packages are72not working inference, activated servers, loaded models, or quality evidence.73Optional backend extras may be absent; base AnyJev still requires Python74`>=3.10`, AnyJev metadata, and NumPy `>=1.24`.7576### 3. Execute the approved route7778Install only the selected optional extra and make model, dataset, server, and79credential side effects explicit. Preserve the `Question` option/level80contracts, require `L0` when it is part of the acceptance condition, and never81describe Tacit's one-forward output as an L0-averaged result. For HTTP82backends, verify tokenizer availability and required log-probability settings;83keep the gateway loopback-bound unless external authentication and network84policy are in place.8586### 4. Verify and report8788For no-download API plumbing, the pinned `FakeBackend` smoke path is allowed89but is fixture evidence, not model-quality evidence. For real inference report90the AnyJev commit, model revision, backend/engine, tokenizer, dtype/device,91question layout, prior/adaptive settings, and dataset license. Report package92metadata and PATH detections separately from successful inference.9394## Examples9596- **Inspect only:** run the offline preflight with `--repo` and the pinned97 `--expect-commit`; then read the pinned metadata and limitations.98- **Decider:** use a causal-LM backend and `Decider(level="L0")`; inspect the99 returned `distribution` and explicit `level` before acting.100- **Tacit service:** approve the checkpoint, vLLM engine, loopback endpoint,101 and compute first; then verify `/health`, `/v1/stats`, and one `/v1/decide`.102- **No-download smoke:** use `FakeBackend` only to exercise prompt/readout103 plumbing and state clearly that it says nothing about accuracy.104105## Best practices1061071. Keep model/dataset downloads, optional installs, remote-server calls, and108 provider credentials opt-in; local inference still consumes compute.1092. Prefer `Decider(level="L0")` when option-order bias matters and enforce the110 level; choose `prior="content_free"` or `prior="none"` deliberately.1113. Calibrate adaptive shifts with `canonical_order=True`; the certificate is112 agreement with full-K AnyJev output, not external accuracy.1134. Treat Decider `answer_mass` as prompt-landed evidence; treat Tacit's `margin`114 and `route` as diagnostics, not universal confidence; keep its escalation cap115 enabled for shared services.1165. Treat GGUF/Hugging Face tokenizer disagreement as a safe failure and keep117 non-loopback gateways behind authentication and network policy.1186. Preserve Apache-2.0 notices and identify benchmark dataset licenses; do not119 generalize isolated decision evidence into agent-loop claims.120121## References122123- [`references/upstream.md`](references/upstream.md) — pinned commands, backend124 contracts, tokenizer/calibration limits, gateway boundary, and source index.125- [`scripts/preflight.py`](scripts/preflight.py) — offline metadata and checkout126 shape/commit preflight; it never imports AnyJev or invokes providers.127- [AnyJev pinned repository](https://github.com/nokia-applied-research/AnyJev/tree/d458d13968eec0e2cfa8c81b64629b96cf5cb4a9)128- [AnyJev Apache-2.0 license](https://github.com/nokia-applied-research/AnyJev/blob/d458d13968eec0e2cfa8c81b64629b96cf5cb4a9/LICENSE)