Automated README summary
Automatically summarized from the README on 18 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A vLLM endpoint plugin that exposes POST /v1/systemone, a jev-shaped system-one API, on the same server as /v1/chat/completions, so callers can send a state plus typed questions and get calibrated probabilities from a granite-switch decision adapter.
How it uses Jev
Jev-shaped requests carry an unstructured state and typed questions (choice, score, noul). The plugin validates the body, renders the state and option descriptions into chat messages and single-token label ids via a configurable renderer, calls openai_serving_chat.create_chat_completion, and converts top_logprobs into jev's answer shapes with calibrated probabilities and confidence.
Primitives:choicescorenoul
Technique worth stealing
Server-side translation of jev questions into a trained adapter's invocation marker, single-token label pool, and softmax over top_logprobs, reusing the same engine and KV cache.
Try it
docker build --platform linux/amd64 -t granite-switch-systemone:latest . then docker run with SYSTEMONE_INVOCATION_MARKER set.
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.13.0- Project signals
jev-1.13.0- Category
jev-1.13.0- Pattern tags
jev-1.13.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.89
- Shows a System One patternunclear0.56
- Describes handling uncertaintynot found in the excerpt0.04
- Reports measured numbersnot found in the excerpt0.04
- Is a runnable toolFound in material0.92
- Top-10 learning shortlist signalnot found in the excerpt0.29
- Is a replica of the modelunclear0.50
- Solves a common problem (0–2)score on a 0–2 scale1.00 / 2
- About JevFound in material0.98
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 18 Oct 2026.