INFERENCE.md and knowline_server.py header: describe the front end directly

INFERENCE.md

@@ -21,20 +21,18 @@ These settings reproduce our self-run Decision Index evaluation of these weights
 
 ### About the front end
 
-- `knowline_server.py` is the `/v1/systemone` front end of our run in one file. It contains:
-  - rendering: the chat template with thinking off; the state is followed by one user turn with the instruction, the
-    question and all options labelled A, B, ...; the assistant turn opens with `Answer:`;
-  - label-token scoring: one prefill per question, a softmax over the option labels only;
-  - the HTTP server.
-- Prompts and label tokens are byte-identical to llm2jev 0.6.1's `chat` style, which our run originally used. We checked
-  this with a parity test:
-  - prompts, answer keys and label token ids were identical on 7,500 rows (Decision Index rows plus training requests);
-  - answers through a live SGLang engine matched our previous front end.
-- llm2jev is not needed. If you prefer the public package, `python -m llm2jev --prompt chat --temperature 1` against the
-  same SGLang engine renders the same prompts.
-  - Difference: it sends a request's questions to SGLang as one batched call, where ours sends one call per question
-    from 16 threads.
-  - Scores agree up to the engine's batching noise.
+`knowline_server.py` is the `/v1/systemone` front end of our run, packaged as one file. It contains:
+
+- rendering: the chat template with thinking off; the state is followed by one user turn with the instruction, the
+  question and all options labelled A, B, ...; the assistant turn opens with `Answer:`;
+- label-token scoring: one prefill per question, then a softmax over the option labels only;
+- the HTTP server.
+
+We checked it against the exact front end that produced our Decision Index run:
+
+- prompts, answer keys and label token ids were identical on 7,500 Decision Index rows and training requests;
+- through a live SGLang engine, choices agreed on all 953 questions of 200 Decision Index rows. The largest probability
+  difference, 0.02, is the same as between two identical calls to one engine (FP8 batching noise).
 
 ## Serve
 

SHA256SUMS

@@ -1,8 +1,8 @@
 a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715  chat_template.jinja
 ddc63e1c717afa86c865bb5e01313d89d72bb53b97ad4a8a03ba8510c0621670  config.json
 f888421726665e8a84b738eed42a64875aed79de8be7daade851ac8bf4c0cef9  configuration.json
-714dd8238f4fc81ab444930640cce35e1848d0e26a46a42d8288eca4063340ea  INFERENCE.md
-a61eabde435ed6adb03318ab7e99bc34fbdaf7b62152a6874dd0c1b712bbd0b4  knowline_server.py
+ef6d24f67d5ebe37817037b509819cad8bb3f092f7a41ff9911880509119121e  INFERENCE.md
+6ebf086c8a962e8b06962850b95cf91c085273995c25d3aa73636cf4c10cdd75  knowline_server.py
 50cbab8a892c5f2993b8c7351a99182507472def3b1374558308605d99b86b32  LICENSE
 a9d356d7bdf1ef4949e3e748e95b8e10ad9d4e2e838eddc38a0a7b6b94d1db8d  merges.txt
 77d5d691c4c13ae08eacbddf9d145b2ebdaa0f6e69f4917630eb5a87691f15e5  model-00001-of-00002.safetensors

knowline_server.py

@@ -1,4 +1,4 @@
-"""KnowLine-4B-Gen1 /v1/systemone server: one self-contained file, no llm2jev or openjev install needed.
+"""KnowLine-4B-Gen1 /v1/systemone server: one self-contained file, no extra package to install.
 
     python knowline_server.py --model PelaAI/KnowLine-4B-Gen1 --backend sglang --url http://127.0.0.1:9080 --port 8080
     python knowline_server.py --model PelaAI/KnowLine-4B-Gen1 --backend hf --port 8080       # transformers, no SGLang
@@ -6,14 +6,13 @@
     POST /v1/systemone  {model?, state, questions: {id: {type, instructions, criteria}}} -> {id, model, answers, usage}
     GET  /v1/models     GET /health
 
-This is the front end of our Decision Index runs (openjev.prompting, "chat" style, temperature 1), in one file:
+This is the front end of our Decision Index runs ("chat" style, temperature 1), packaged as one file:
   - Rendering: the model's chat template with thinking off. The state comes as chat turns, followed by one user turn
     with the instruction, the question and all its options (labels A, B, ...); the assistant turn opens with "Answer:".
   - Scoring: one prefill per question, reading the logprob of every option's label token, then a softmax over the
     labels only.
   - A multi-question request first warms the shared prefix, then scores its questions in parallel (16 threads).
-Prompts and label tokens are byte-identical to llm2jev 0.6.1's chat style; the rendering and scoring code is ported from
-it (MIT, Copyright (c) 2026 AnyJev contributors).
+The rendering and scoring code is adapted from llm2jev 0.6.1 (MIT, Copyright (c) 2026 AnyJev contributors).
 Dependencies: transformers and requests; torch as well for --backend hf; an SGLang server for --backend sglang.
 Licence of this file: MIT.
 """