release: v0.2.0 async OpenAI workflow, docs/tutorial refresh, and CI updates

.Rbuildignore

@@ -1,6 +1,7 @@
 ^vignettes/\.quarto$
 ^vignettes/.*_files$
 ^IMPLEMENTATION_NOTES\.md$
+^DEVELOPMENT_CONTINUITY\.md$
 ^doc$
 ^Meta$
 ^inst/ASR$

.github/workflows/pr-checks.yml

@@ -45,4 +45,5 @@ jobs:
       - name: Run R CMD check
         uses: r-lib/actions/check-r-package@v2
         with:
-          args: 'c("--no-manual", "--no-build-vignettes", "--no-multiarch")'
+          build_args: 'c("--no-manual", "--no-build-vignettes")'
+          args: 'c("--no-manual", "--no-multiarch", "--no-examples", "--ignore-vignettes")'

DESCRIPTION

@@ -1,18 +1,19 @@
 Package: openalexVectorComp
 Type: Package
-Title: Auto-tagging via TEI Embeddings and Qdrant (Prototype-Margin + Ridge Logistic)
-Version: 0.2.0
+Title: Embedding Vectorization and Distance-Based Scoring Workflows
+Version: 0.3.0
 Authors@R: c(
     person(given = "Rainer", family = "Krug", role = c("aut", "cre"), email = "you@example.org"),
     person(given = "ChatGPT", family = "Assistant", role = "ctb")
   )
 Author: Rainer Krug [aut, cre],
   ChatGPT Assistant [ctb]
 Maintainer: Rainer Krug <you@example.org>
-Description: R-first orchestration for auto-tagging based on text embeddings served by
-    a TEI (Text Embeddings Inference) server and vector search in Qdrant.
-    Provides prototype-margin scoring, ridge logistic classification, simple ensembling,
-    calibration/threshold selection, and utilities to ingest/query Qdrant.
+Description: R-first orchestration for text vectorization (embeddings),
+    embedding distance computation, and distance-based scoring workflows.
+    Supports backend-neutral embedding providers (HF, OpenAI, TEI),
+    prototype cosine-distance scoring, reference-area distance scoring,
+    and threshold calibration utilities.
 License: MIT + file LICENSE
 Encoding: UTF-8
 LazyData: true

inst/DEVELOPMENT_CONTINUITY.md → DEVELOPMENT_CONTINUITY.md

@@ -43,20 +43,20 @@ Core flow:
 5. Optional threshold calibration (`calibrate_threshold()`).
 
 OpenAI batch flow:
-1. Submit (`embed_corpus_submit_openai_batch()`).
-2. Refresh status (`embed_corpus_status_openai_batch()`).
-3. Collect completed jobs (`embed_corpus_collect_openai_batch()`).
+1. Submit (`batch_submit_openai()`).
+2. Refresh status (`batch_status_openai()`).
+3. Collect completed jobs (`batch_collect_openai()`).
 4. Demo convenience wrapper:
-- `finalize_demo_openai_batch()` = status + collect + direct-vs-batch compare.
+- `demo_finalize_openai_batch()` = status + collect + direct-vs-batch compare.
 
-## 3. Current Demo Conventions (0.1.4)
+## 3. Current Demo Conventions (0.3.0)
 
 Default demo locations:
 - `demos/openalex`
 - `demos/openai`
 
 OpenAI demo behavior:
-- `run_demo_openai_quarto(..., render = TRUE)` may complete before batch does.
+- `run_demo_openai(..., render = TRUE)` may complete before batch does.
 - User is given explicit follow-up commands for status/finalize.
 - Batch comparison outputs are written to:
   `project/openai_batch_comparison/label=corpus_batch/`.
@@ -79,7 +79,7 @@ Template:
 - Date: 2026-04-01
 - Scope: OpenAI demo and batch comparison robustness
 - Decision: Implement two-phase OpenAI batch demo flow with
-  `finalize_demo_openai_batch()`.
+  `demo_finalize_openai_batch()`.
 - Why: Batch completion is asynchronous; render should not fail on pending jobs.
 - Alternatives considered: long blocking poll in render; hard-fail on timeout.
 - Impact: Clearer async semantics; stable demo render; persisted comparison