logit-classifier 0.2.0__tar.gz → 0.3.0__tar.gz

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.5
 Name: logit-classifier
-Version: 0.2.0
-Summary: Local zero-shot classifier for text and images. Declares options, returns a calibrated probability for each, generates no text. Accepts the TypeSafe System One request format.
+Version: 0.3.0
+Summary: Local zero-shot classifier for text and images. Declares options and reads a calibrated probability for each from the logits. Accepts the TypeSafe System One request format, and ships a toolkit of image tag tools and ComfyUI node tools.
 Project-URL: Repository, https://github.com/Blakeem/logit-classifier
 Project-URL: Bug Tracker, https://github.com/Blakeem/logit-classifier/issues
 Author: Blake
@@ -88,8 +88,8 @@ beside your project instead, which is what the examples do.
 Config(models_dir=Path("models"))
 ```
 
-`LOGIT_MODELS_DIR` sets the same folder for every process, and `HF_HOME` moves the
-Hugging Face cache itself.
+`LOGIT_MODELS_DIR` sets the same folder for the service and for `Config.from_env()`.
+`HF_HOME` moves the Hugging Face cache itself.
 
 ## Models
 
@@ -98,8 +98,9 @@ Hugging Face cache itself.
 | reads images | yes | no |
 | better at | choice and score | noul |
 
-`Qwen3-VL-4B-Instruct` is the default. Set `LOGIT_MODEL_ID` to use the other one. Any Qwen
-chat model loads, and a model with no fitted temperature gets 2.5.
+`Qwen3-VL-4B-Instruct` is the default. Set `LOGIT_MODEL_ID` to use the other one in the
+service or through `Config.from_env()`. Any Qwen chat model loads, and a model with no
+fitted temperature gets 2.5.
 
 ## Python API
 
@@ -123,6 +124,14 @@ print(response.to_dict())
 `load_model` loads the weights through the `hf` extra and returns a `Backend`.
 `Classifier(Config())` calls it for you when you pass no backend.
 
+`classifier.nouls` asks a list of statements about one state and returns one probability
+per statement. It sends more than 256 statements as several requests.
+
+```python
+probabilities = classifier.nouls(["The message is urgent", "The message is polite"],
+                                 state="the payment failed again")
+```
+
 Loading each model yourself is what lets one script compare several. The fitted
 temperature follows the backend, so every model is scored with its own value.
 
@@ -169,7 +178,7 @@ the model.
 
 ```json
 {
-  "model": "logit-classifier-0.2.0",
+  "model": "logit-classifier-0.3.0",
   "answers": {
     "department": {
       "type": "choice",
@@ -230,8 +239,9 @@ Each script runs on its own.
 
 ## Configuration
 
-Every setting reads from the environment at startup. `logit-classifier config` prints what
-they produce.
+The service and `logit-classifier config` read these variables at startup through
+`Config.from_env()`. Library code gets them by calling `Config.from_env()`, since `Config()`
+reads none of them. `logit-classifier config` prints what they produce.
 
 | Variable | Default | Effect |
 |---|---|---|
@@ -242,6 +252,7 @@ they produce.
 | `LOGIT_PRIOR_DEBIAS` | `1` | set to `0` to skip the label prior |
 | `LOGIT_BATCH_BRANCHES` | `1` | set to `0` for one forward pass per branch |
 | `LOGIT_SCORE_METHOD` | `joint` | set to `independent` to judge each level alone |
+| `LOGIT_PERMUTATIONS` | `1` | letterings averaged per question, each one adds branches |
 | `LOGIT_ABSTAIN` | `1` | set to `0` to drop the `none of these` label below 52 options |
 | `LOGIT_CALIBRATION_PATH` | `calibration.json` | where the service stores the running prior |
 
@@ -269,10 +280,10 @@ Nothing is sampled, so the same request returns bitwise identical logits.
 
 Batch composition and padding length both change the low bits of a bfloat16 forward pass,
 and both are pure functions of the request. So the same question asked inside two
-different requests can differ slightly. Set `LOGIT_BATCH_BRANCHES=0` to score each branch
-alone, which makes a question independent of the questions sent with it and costs one
-forward pass per branch. `ComfyClipBackend` reads no environment variable, so it takes
-`batch_branches=False` as a keyword instead.
+different requests can differ slightly. Scoring each branch alone makes a question
+independent of the questions sent with it and costs one forward pass per branch. The service
+turns it on with `LOGIT_BATCH_BRANCHES=0`. Library code passes `Config(batch_branches=False)`,
+and `ComfyClipBackend` takes `batch_branches=False` as a keyword.
 
 A forward pass needs several process-global torch settings held at known values. Torch
 exposes none of them as a call argument, so the backend sets them around each pass and
@@ -339,6 +350,72 @@ shape `[1, H, W, 3]`. An empty state asks about the image alone. Questions share
 pass of up to 4,096 tokens, and the image counts toward that limit. Any text encoder other
 than Qwen3-VL raises `UnsupportedModelError`.
 
+## Toolkit
+
+`logit_classifier.toolkit` holds tools built on the classifier with the settings our
+measurements found best. A project imports them to get good results without working out the
+wording, thresholds and host details itself. The toolkit uses the classifier's public API,
+and the classifier never imports the toolkit.
+
+| Package | Runs in | Holds |
+|---|---|---|
+| `logit_classifier.toolkit.tags` | any Python process | text tools for image tags and prompts |
+| `logit_classifier.toolkit.comfyui` | a ComfyUI node | a classifier over the workflow's CLIP, text generation, a progress display and an image tagger |
+
+### Tag Text Tools
+
+| Tool | Does |
+|---|---|
+| `parse_candidates` | splits a model's tag reply into clean tags with no duplicates |
+| `split_prompt` | splits an image prompt into its fragments |
+| `drop_subsets` | drops a tag whose words all appear in a longer tag |
+| `prompt_word_forms`, `in_prompt` | test whether a tag's words appear in a prompt, singular or plural |
+| `merge_candidates` | merges proposed tags with a prompt's tags under a cap |
+| `presence_question` | writes "Is there a car in this image?" with the article the thing takes |
+
+```python
+from logit_classifier.toolkit.tags import drop_subsets, parse_candidates
+
+tags = drop_subsets(parse_candidates("red car, car, wooden bench, bench"))
+# ['red car', 'wooden bench']
+```
+
+`drop_subsets(tags, rule="head-noun")` drops a tag only into a longer tag with the same head
+noun. The head noun is the last word before "of", "in", "on", "with" or "reading". So "cat"
+stays beside "cat ear", and "signage" is dropped beside "signage in chinese".
+
+### ComfyUI Tools
+
+These run inside a ComfyUI node over the CLIP the workflow loaded. Each one imports comfy and
+torch only when it is called.
+
+| Tool | Does |
+|---|---|
+| `comfy_classifier` | builds a classifier over the CLIP, with errors that name the node and the Load CLIP fix |
+| `generate_text` | generates one greedy reply with the Qwen chat template, and stops early on a condition |
+| `shared_vision_encode` | runs the vision tower once per picture for every generate and classify inside the block |
+| `skip_resident_loads` | skips core's model load while the CLIP is already on the GPU |
+| `routed_progress_bars`, `send_status`, `run_outcome` | show one progress bar per run and a status line under the node |
+| `fit_picture` | downscales an image to a pixel cap |
+| `prompt_tags` | lists the physical things a prompt names |
+| `tag_picture` | tags a picture, with a prompt's things added as candidates |
+| `TagSettings` | holds every wording, budget, cap and threshold of the tagger |
+
+The tagger generates candidate tags over the CLIP and verifies each one with the classifier.
+
+```python
+from logit_classifier.toolkit.comfyui import TagSettings, comfy_classifier, fit_picture, tag_picture
+
+classifier = comfy_classifier(clip, node="My Tagger")
+picture = fit_picture(image, 1024 * 1024)
+trace = tag_picture(clip, classifier, picture, settings=TagSettings(), known={})
+tags = trace.kept
+```
+
+`known` holds the thing check's answers across a run, so each tag is asked once. The
+`TagSettings` defaults are the current measured values, and a later release may change them.
+Pass every value your output depends on to keep it fixed.
+
 ## Differences From Jev
 
 This project matches the System One request and response format and the documented limits.
@@ -348,8 +425,8 @@ Jev is trained for calibrated probabilities. This project reads them from a gene
 so the choices agree more often than the confidences do.
 
 Jev judges a score level without its number or its neighbours. This project judges all
-levels together by default. Set `LOGIT_SCORE_METHOD=independent` for the documented
-behavior.
+levels together by default. The service follows the documented behavior with
+`LOGIT_SCORE_METHOD=independent`, and library code with `Config(score_method="independent")`.
 
 Jev publishes status codes but no error body. The error shape here is our own.
 

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/README.md

@@ -55,8 +55,8 @@ beside your project instead, which is what the examples do.
 Config(models_dir=Path("models"))
 ```
 
-`LOGIT_MODELS_DIR` sets the same folder for every process, and `HF_HOME` moves the
-Hugging Face cache itself.
+`LOGIT_MODELS_DIR` sets the same folder for the service and for `Config.from_env()`.
+`HF_HOME` moves the Hugging Face cache itself.
 
 ## Models
 
@@ -65,8 +65,9 @@ Hugging Face cache itself.
 | reads images | yes | no |
 | better at | choice and score | noul |
 
-`Qwen3-VL-4B-Instruct` is the default. Set `LOGIT_MODEL_ID` to use the other one. Any Qwen
-chat model loads, and a model with no fitted temperature gets 2.5.
+`Qwen3-VL-4B-Instruct` is the default. Set `LOGIT_MODEL_ID` to use the other one in the
+service or through `Config.from_env()`. Any Qwen chat model loads, and a model with no
+fitted temperature gets 2.5.
 
 ## Python API
 
@@ -90,6 +91,14 @@ print(response.to_dict())
 `load_model` loads the weights through the `hf` extra and returns a `Backend`.
 `Classifier(Config())` calls it for you when you pass no backend.
 
+`classifier.nouls` asks a list of statements about one state and returns one probability
+per statement. It sends more than 256 statements as several requests.
+
+```python
+probabilities = classifier.nouls(["The message is urgent", "The message is polite"],
+                                 state="the payment failed again")
+```
+
 Loading each model yourself is what lets one script compare several. The fitted
 temperature follows the backend, so every model is scored with its own value.
 
@@ -136,7 +145,7 @@ the model.
 
 ```json
 {
-  "model": "logit-classifier-0.2.0",
+  "model": "logit-classifier-0.3.0",
   "answers": {
     "department": {
       "type": "choice",
@@ -197,8 +206,9 @@ Each script runs on its own.
 
 ## Configuration
 
-Every setting reads from the environment at startup. `logit-classifier config` prints what
-they produce.
+The service and `logit-classifier config` read these variables at startup through
+`Config.from_env()`. Library code gets them by calling `Config.from_env()`, since `Config()`
+reads none of them. `logit-classifier config` prints what they produce.
 
 | Variable | Default | Effect |
 |---|---|---|
@@ -209,6 +219,7 @@ they produce.
 | `LOGIT_PRIOR_DEBIAS` | `1` | set to `0` to skip the label prior |
 | `LOGIT_BATCH_BRANCHES` | `1` | set to `0` for one forward pass per branch |
 | `LOGIT_SCORE_METHOD` | `joint` | set to `independent` to judge each level alone |
+| `LOGIT_PERMUTATIONS` | `1` | letterings averaged per question, each one adds branches |
 | `LOGIT_ABSTAIN` | `1` | set to `0` to drop the `none of these` label below 52 options |
 | `LOGIT_CALIBRATION_PATH` | `calibration.json` | where the service stores the running prior |
 
@@ -236,10 +247,10 @@ Nothing is sampled, so the same request returns bitwise identical logits.
 
 Batch composition and padding length both change the low bits of a bfloat16 forward pass,
 and both are pure functions of the request. So the same question asked inside two
-different requests can differ slightly. Set `LOGIT_BATCH_BRANCHES=0` to score each branch
-alone, which makes a question independent of the questions sent with it and costs one
-forward pass per branch. `ComfyClipBackend` reads no environment variable, so it takes
-`batch_branches=False` as a keyword instead.
+different requests can differ slightly. Scoring each branch alone makes a question
+independent of the questions sent with it and costs one forward pass per branch. The service
+turns it on with `LOGIT_BATCH_BRANCHES=0`. Library code passes `Config(batch_branches=False)`,
+and `ComfyClipBackend` takes `batch_branches=False` as a keyword.
 
 A forward pass needs several process-global torch settings held at known values. Torch
 exposes none of them as a call argument, so the backend sets them around each pass and
@@ -306,6 +317,72 @@ shape `[1, H, W, 3]`. An empty state asks about the image alone. Questions share
 pass of up to 4,096 tokens, and the image counts toward that limit. Any text encoder other
 than Qwen3-VL raises `UnsupportedModelError`.
 
+## Toolkit
+
+`logit_classifier.toolkit` holds tools built on the classifier with the settings our
+measurements found best. A project imports them to get good results without working out the
+wording, thresholds and host details itself. The toolkit uses the classifier's public API,
+and the classifier never imports the toolkit.
+
+| Package | Runs in | Holds |
+|---|---|---|
+| `logit_classifier.toolkit.tags` | any Python process | text tools for image tags and prompts |
+| `logit_classifier.toolkit.comfyui` | a ComfyUI node | a classifier over the workflow's CLIP, text generation, a progress display and an image tagger |
+
+### Tag Text Tools
+
+| Tool | Does |
+|---|---|
+| `parse_candidates` | splits a model's tag reply into clean tags with no duplicates |
+| `split_prompt` | splits an image prompt into its fragments |
+| `drop_subsets` | drops a tag whose words all appear in a longer tag |
+| `prompt_word_forms`, `in_prompt` | test whether a tag's words appear in a prompt, singular or plural |
+| `merge_candidates` | merges proposed tags with a prompt's tags under a cap |
+| `presence_question` | writes "Is there a car in this image?" with the article the thing takes |
+
+```python
+from logit_classifier.toolkit.tags import drop_subsets, parse_candidates
+
+tags = drop_subsets(parse_candidates("red car, car, wooden bench, bench"))
+# ['red car', 'wooden bench']
+```
+
+`drop_subsets(tags, rule="head-noun")` drops a tag only into a longer tag with the same head
+noun. The head noun is the last word before "of", "in", "on", "with" or "reading". So "cat"
+stays beside "cat ear", and "signage" is dropped beside "signage in chinese".
+
+### ComfyUI Tools
+
+These run inside a ComfyUI node over the CLIP the workflow loaded. Each one imports comfy and
+torch only when it is called.
+
+| Tool | Does |
+|---|---|
+| `comfy_classifier` | builds a classifier over the CLIP, with errors that name the node and the Load CLIP fix |
+| `generate_text` | generates one greedy reply with the Qwen chat template, and stops early on a condition |
+| `shared_vision_encode` | runs the vision tower once per picture for every generate and classify inside the block |
+| `skip_resident_loads` | skips core's model load while the CLIP is already on the GPU |
+| `routed_progress_bars`, `send_status`, `run_outcome` | show one progress bar per run and a status line under the node |
+| `fit_picture` | downscales an image to a pixel cap |
+| `prompt_tags` | lists the physical things a prompt names |
+| `tag_picture` | tags a picture, with a prompt's things added as candidates |
+| `TagSettings` | holds every wording, budget, cap and threshold of the tagger |
+
+The tagger generates candidate tags over the CLIP and verifies each one with the classifier.
+
+```python
+from logit_classifier.toolkit.comfyui import TagSettings, comfy_classifier, fit_picture, tag_picture
+
+classifier = comfy_classifier(clip, node="My Tagger")
+picture = fit_picture(image, 1024 * 1024)
+trace = tag_picture(clip, classifier, picture, settings=TagSettings(), known={})
+tags = trace.kept
+```
+
+`known` holds the thing check's answers across a run, so each tag is asked once. The
+`TagSettings` defaults are the current measured values, and a later release may change them.
+Pass every value your output depends on to keep it fixed.
+
 ## Differences From Jev
 
 This project matches the System One request and response format and the documented limits.
@@ -315,8 +392,8 @@ Jev is trained for calibrated probabilities. This project reads them from a gene
 so the choices agree more often than the confidences do.
 
 Jev judges a score level without its number or its neighbours. This project judges all
-levels together by default. Set `LOGIT_SCORE_METHOD=independent` for the documented
-behavior.
+levels together by default. The service follows the documented behavior with
+`LOGIT_SCORE_METHOD=independent`, and library code with `Config(score_method="independent")`.
 
 Jev publishes status codes but no error body. The error shape here is our own.
 

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/examples/compare_models.py

@@ -19,7 +19,8 @@ from pathlib import Path
 from logit_classifier import Classifier, Config, load_model, parse_request
 
 # Weights land beside the project instead of in the global Hugging Face cache.
-# Set LOGIT_MODELS_DIR, or HF_HOME, to keep them somewhere shared across projects.
+# Edit MODELS_DIR to share one folder across projects.
+# Neither LOGIT_MODELS_DIR nor HF_HOME reaches this script.
 MODELS_DIR = Path(__file__).resolve().parent.parent / "models"
 
 MODELS = ["Qwen/Qwen3-VL-4B-Instruct", "Qwen/Qwen3-4B-Instruct-2507"]

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/examples/images.py

@@ -23,7 +23,8 @@ from logit_classifier import (
 )
 
 # Weights land beside the project instead of in the global Hugging Face cache.
-# Set LOGIT_MODELS_DIR, or HF_HOME, to keep them somewhere shared across projects.
+# Edit MODELS_DIR to share one folder across projects.
+# Neither LOGIT_MODELS_DIR nor HF_HOME reaches this script.
 MODELS_DIR = Path(__file__).resolve().parent.parent / "models"
 QUESTIONS = {
     "subject": {

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/examples/question_types.py

@@ -14,7 +14,8 @@ from pathlib import Path
 from logit_classifier import Classifier, Config, load_model, parse_request
 
 # Weights land beside the project instead of in the global Hugging Face cache.
-# Set LOGIT_MODELS_DIR, or HF_HOME, to keep them somewhere shared across projects.
+# Edit MODELS_DIR to share one folder across projects.
+# Neither LOGIT_MODELS_DIR nor HF_HOME reaches this script.
 MODELS_DIR = Path(__file__).resolve().parent.parent / "models"
 
 REVIEW = "Shipped two days late and the box was crushed, but the product itself works fine."

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/examples/quickstart.py

@@ -16,7 +16,8 @@ from pathlib import Path
 from logit_classifier import Classifier, Config, load_model, parse_request
 
 # Weights land beside the project instead of in the global Hugging Face cache.
-# Set LOGIT_MODELS_DIR, or HF_HOME, to keep them somewhere shared across projects.
+# Edit MODELS_DIR to share one folder across projects.
+# Neither LOGIT_MODELS_DIR nor HF_HOME reaches this script.
 MODELS_DIR = Path(__file__).resolve().parent.parent / "models"
 
 SAMPLE_TEXT = "I have been trying to connect my Stripe account for 3 days and it keeps failing."

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "logit-classifier"
 dynamic = ["version"]
-description = "Local zero-shot classifier for text and images. Declares options, returns a calibrated probability for each, generates no text. Accepts the TypeSafe System One request format."
+description = "Local zero-shot classifier for text and images. Declares options and reads a calibrated probability for each from the logits. Accepts the TypeSafe System One request format, and ships a toolkit of image tag tools and ComfyUI node tools."
 requires-python = ">=3.12"
 license = "GPL-3.0-or-later"
 license-files = ["LICENSE"]
@@ -136,8 +136,8 @@ show_error_codes = true
 [[tool.mypy.overrides]]
 # transformers publishes no complete stubs, and accelerate none at all. torch is
 # listed so the type gate still runs where the hf extra is absent, such as CI.
-# The ComfyUI host supplies comfy, and neither this venv nor CI has it.
-module = ["transformers.*", "accelerate.*", "torch.*", "comfy.*"]
+# The ComfyUI host supplies comfy, server and comfy_execution, and neither this venv nor CI has them.
+module = ["transformers.*", "accelerate.*", "torch.*", "comfy.*", "server", "server.*", "comfy_execution.*"]
 ignore_missing_imports = true
 
 [[tool.mypy.overrides]]

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/_version.py

@@ -1,3 +1,3 @@
 """The package version, apart from __init__ so config.py can read it without a cycle."""
 
-__version__ = "0.2.0"
+__version__ = "0.3.0"

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/backends/comfy_clip.py

@@ -101,6 +101,29 @@ def _check_image(image: Any) -> None:
         raise ImageError(f"a ComfyUI IMAGE of shape [1, H, W, 3] is required, got shape {shape}")
 
 
+def _resident(clip: Any) -> bool:
+    """Whether the CLIP sits where its own load_model call leaves it, so loading again is a no-op.
+
+    Core's load_models_gpu has no fast path for a loaded model and costs about 0.1 s per call.
+    """
+    try:
+        import comfy.model_management as model_management
+
+        patcher = clip.patcher
+        loaded = model_management.current_loaded_models
+        # Another model's load always inserts at the head, and a changed LoRA patch set or an
+        # offloaded CLIP must reload.
+        return bool(
+            loaded
+            and loaded[0].model is patcher
+            and patcher.model.device == patcher.load_device
+            and patcher.model.current_weight_patches_uuid == patcher.patches_uuid
+            and patcher.model.model_loaded_weight_memory > 0
+        )
+    except (ImportError, AttributeError):
+        return False
+
+
 def _pack_passes(prefix_length: int, suffix_lengths: list[int], batch_branches: bool) -> list[list[int]]:
     """Group suffixes in request order into passes that stay under PACKED_TOKEN_CEILING.
 
@@ -284,7 +307,8 @@ class ComfyClipBackend:
             prefix_tokens[vision["pad_index"]] = {"type": "image", "data": vision["image"], "original_type": "image"}
         with _determinism(), torch.inference_mode():
             encoder.reset_clip_options()
-            clip.load_model({self._tokens_key: [prefix_tokens]})
+            if not _resident(clip):
+                clip.load_model({self._tokens_key: [prefix_tokens]})
             device = clip.patcher.load_device
             encoder.set_clip_options({"layer": None, "execution_device": device})
             # BaseGenerate.generate picks its execution dtype this way, at llama.py:1115-1120.

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/classifier.py

@@ -4,6 +4,7 @@ from __future__ import annotations
 
 import hashlib
 import random
+from collections.abc import Sequence
 from dataclasses import dataclass, field, replace
 from typing import Any
 
@@ -11,7 +12,7 @@ import numpy as np
 
 from .backends.base import Backend, BackendContractError, BranchLogits
 from .calibrate import PriorStore
-from .config import ANSWER_PREFILL, Config, fitted_temperature
+from .config import ANSWER_PREFILL, JEV_MAX_QUESTIONS, Config, fitted_temperature
 from .deps import MissingDependencyError
 from .prompt import (
     PROMPT_VERSION,
@@ -25,6 +26,7 @@ from .schema import (
     Answer,
     ChoiceAnswer,
     ChoiceQuestion,
+    JSONContent,
     NoulAnswer,
     NoulQuestion,
     Question,
@@ -180,13 +182,12 @@ class Classifier:
         rounds = [questions]
 
         for seed in range(1, max(1, self.config.permutations)):
-            shuffler = random.Random(seed)
             reordered: dict[str, Question] = {}
             for qid, question in questions.items():
                 if not isinstance(question, ChoiceQuestion):
                     continue
                 names = list(question.criteria)
-                shuffler.shuffle(names)
+                random.Random(seed).shuffle(names)
                 criteria = {name: question.criteria[name] for name in names}
                 reordered[qid] = replace(question, criteria=criteria)
             if reordered:
@@ -280,6 +281,28 @@ class Classifier:
         )
         return response, diagnostics
 
+    def nouls(self, statements: Sequence[str], *, image: Any = None, state: JSONContent = "") -> list[float]:
+        """Return each statement's noul about the state, in statement order.
+
+        The statements are split into requests of JEV_MAX_QUESTIONS, one classify call each, and no
+        statements make no request. An empty state with an image is the image-only state.
+        """
+        keys = [f"s{index}" for index in range(len(statements))]
+        questions: dict[str, Question] = {}
+        nouls: list[float] = []
+
+        for start in range(0, len(statements), JEV_MAX_QUESTIONS):
+            window = slice(start, start + JEV_MAX_QUESTIONS)
+            questions = {key: NoulQuestion(instructions=statement)
+                         for key, statement in zip(keys[window], statements[window], strict=True)}
+            response, _diagnostics = self.classify(SystemOneRequest(state=state, questions=questions), image=image)
+            for key in questions:
+                answer = response.answers[key]
+                if not isinstance(answer, NoulAnswer):
+                    raise TypeError(f"question {key} got a {answer.type} answer, not a noul")
+                nouls.append(answer.noul)
+        return nouls
+
     def _round_answers(self, questions: dict[str, Question], branches: list[Branch],
                        probabilities: list[np.ndarray]) -> dict[str, Answer]:
         answers: dict[str, Answer] = {}

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/config.py

@@ -141,9 +141,9 @@ class Config:
 
     # "joint" scores all levels in one branch, "independent" judges each alone.
     score_method: str = "joint"
-    # Averaging over several letterings cancels the model's preference for a label
-    # position. Measured on Banking77, four letterings moved accuracy from 0.554 to
-    # 0.693. Off by default because it multiplies the branch count.
+    # Relettering varies which group each option lands in once a question splits above 52 options.
+    # On Banking77's 77 options, four letterings moved accuracy from 0.554 to 0.693. At 10 options
+    # in one branch, accuracy did not move. Off by default because it multiplies the branch count.
     permutations: int = 1
     # An escape label absorbs the mass the model would otherwise spread over wrong
     # options, so offering one raised accuracy from 0.881 to 0.887 as well as scoring

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/prompt.py

@@ -52,6 +52,12 @@ class Branch:
     suffix_text: str
 
 
+def inert(text: str) -> str:
+    # The backends' tokenizers read a <|...|> spelling in plain text as a control token, so
+    # client text could otherwise forge a chat turn or a second image pad.
+    return text.replace("<|", "<\u200b|")
+
+
 def _option_lines(entries: list[tuple[str, str]]) -> str:
     lines = []
     for index, (name, description) in enumerate(entries):
@@ -61,7 +67,7 @@ def _option_lines(entries: list[tuple[str, str]]) -> str:
 
 
 def _question_block(prompt_line: str, entries: list[tuple[str, str]]) -> str:
-    return f"{prompt_line}\nOptions:\n{_option_lines(entries)}"
+    return inert(f"{prompt_line}\nOptions:\n{_option_lines(entries)}")
 
 
 def _choice_branches(qid: str, question: ChoiceQuestion, plan: BranchPlan,
@@ -146,7 +152,7 @@ def build_branches(questions: dict[str, Question], score_method: str = "joint",
 def prefix_content(state: Any, has_image: bool = False) -> str:
     """Build the user-message body every branch shares, image marker included."""
     marker = f"{IMAGE_MARKER}\n" if has_image else ""
-    return f"Context:\n{marker}{render_content(state)}\n\n"
+    return f"Context:\n{marker}{inert(render_content(state))}\n\n"
 
 
 def branch_content(state: Any, branch: Branch, has_image: bool = False) -> str:

{logit_classifier-0.2.0 → logit_classifier-0.3.0}/src/logit_classifier/schema.py

@@ -25,6 +25,8 @@ _REQUEST_FIELDS = frozenset({"state", "model", "questions"})
 _QUESTION_FIELDS = frozenset({"type", "instructions", "criteria"})
 _NOUL_CRITERIA_FIELDS = frozenset({"true", "false"})
 _QUESTION_TYPES = ("choice", "score", "noul")
+# Keeps render_content's recursive json.dumps(indent=2) far below the interpreter recursion limit.
+MAX_CONTENT_DEPTH = 64
 
 
 class SchemaError(LogitClassifierError, ValueError):
@@ -178,9 +180,19 @@ def _reject_unknown(mapping: dict[str, Any], allowed: frozenset[str], where: str
 
 
 def _content(value: Any, where: str) -> JSONContent:
-    if isinstance(value, str | dict | list):
+    pending: list[tuple[Any, int]] = [(value, 1)]
+
+    if isinstance(value, str):
         return value
-    raise SchemaError(f"expected a string, object or array, got {type(value).__name__}", where)
+    if not isinstance(value, dict | list):
+        raise SchemaError(f"expected a string, object or array, got {type(value).__name__}", where)
+    while pending:
+        node, depth = pending.pop()
+        if depth > MAX_CONTENT_DEPTH:
+            raise SchemaError(f"content nests deeper than {MAX_CONTENT_DEPTH} levels", where)
+        children = node.values() if isinstance(node, dict) else node
+        pending.extend((child, depth + 1) for child in children if isinstance(child, dict | list))
+    return value
 
 
 def _optional_content(value: Any, where: str) -> JSONContent | None:

logit_classifier-0.3.0/src/logit_classifier/tags.py

@@ -0,0 +1,36 @@
+"""The tag text tools, which moved to logit_classifier.toolkit.tags in 0.3.0.
+
+This path stays for code built on 0.2.x, which PyPI published. It raises no warning, so that code keeps a quiet log.
+"""
+
+from __future__ import annotations
+
+from .toolkit.tags import (
+    MAX_CANDIDATES,
+    MAX_REPEAT_BLOCK,
+    MAX_TAG_CHARS,
+    MAX_TAG_WORDS,
+    clean_item,
+    complete_tags,
+    drop_subsets,
+    drop_unfinished_tag,
+    normalize_item,
+    parse_candidates,
+    repeated_block,
+    split_prompt,
+)
+
+__all__ = [
+    "MAX_CANDIDATES",
+    "MAX_REPEAT_BLOCK",
+    "MAX_TAG_CHARS",
+    "MAX_TAG_WORDS",
+    "clean_item",
+    "complete_tags",
+    "drop_subsets",
+    "drop_unfinished_tag",
+    "normalize_item",
+    "parse_candidates",
+    "repeated_block",
+    "split_prompt",
+]

logit_classifier-0.3.0/src/logit_classifier/toolkit/__init__.py

@@ -0,0 +1,5 @@
+"""Tools built on the classifier core that follow its measured best practice.
+
+A caller gets good results from these tools without working the details out from the base API.
+The tools use the core and never mix into it, so the core never imports this package.
+"""