v2 roadmap: BibTeX, Beamer, direct OMML, run-merger, style map, S3/R2 storage, worker autoscale

[TheClazer]

Implements all seven v2 roadmap items (see README roadmap), each test-covered and behind graceful fallbacks so default behaviour is unchanged.

Converter features

• Full BibTeX extraction — app/converter/bibliography.py parses Word's managed Sources part into references.bib, resolves each in-text CITATION field to a \cite{key}, emits \bibliography{references}, and bundles references.bib in the download zip. Un-managed citations still fall back to plain text.
• Beamer slides template — new beamer template + frame-segmenting renderer (one frame per H1/H2, allowframebreaks, dedicated References frame); omits article geometry.
• Direct OMML→LaTeX parser — app/converter/handlers/omml_direct.py, a Pandoc-free fallback (fractions, scripts, radicals, delimiters, n-ary ops, accents, matrices, Greek/symbol maps). Used when Pandoc is absent or drops an equation, so equations no longer degrade to placeholders.
• Run-merger — app/converter/run_merger.py collapses Word's adjacent same-format run fragmentation; one wrapper per span.
• Style mapping config — app/converter/style_map.py + CORETEX_STYLE_MAP / app/style_map.json map custom Word style names to headings/code/lists. Empty config = pre-v2 behaviour.

Infra / scaling

• S3 / R2 figure storage — app/storage.py FigureStore abstraction. Redis backend is byte-identical to v1; S3FigureStore offloads to S3/Cloudflare R2/MinIO when FIGURE_STORAGE=s3, degrading back to Redis on misconfig.
• Worker autoscale — railway.worker.toml numReplicas=2; independent RQ workers on the shared queue remove single-worker head-of-line blocking.

Verification

• Backend: 102 passed (was 70), ruff clean.
• Frontend: tsc --noEmit, eslint, vitest all green.
• Docs: README roadmap marked v2 shipped; DEPLOY.md scaling section updated.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
core-tex	Ready	Preview, Comment	Jun 9, 2026 10:59pm

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f6142e68ae

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

This PR implements the project’s v2 roadmap items across the conversion pipeline and deployment setup, adding BibTeX/citation support, a Beamer output mode, a Pandoc-free OMML math fallback, run-merging/style-mapping improvements, and optional S3/R2-backed figure storage—while preserving v1 defaults via graceful fallbacks.

Changes:

• Add BibTeX extraction from Word managed sources, resolve in-text citations to \cite{...}, emit \bibliography{references}, and ship references.bib in downloads.
• Add a Beamer template and renderer mode that segments headings into frames, plus frontend template selection support.
• Add infra/scaling improvements: figure-store abstraction (Redis default, optional S3/R2) and Railway worker replica configuration.

Reviewed changes

Copilot reviewed 26 out of 26 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tests/test_v2_features.py	Adds tests for run-merging, BibTeX extraction, and citation rendering behavior.
tests/test_style_map.py	Adds tests validating style-map normalization, clamping, and file-loading behavior.
tests/test_storage.py	Adds tests for Redis figure-store behavior and S3 misconfig fallback.
tests/test_renderer.py	Expands renderer tests to include Beamer and bibliography frame behavior.
tests/test_omml_direct.py	Adds tests for direct OMML→LaTeX conversion coverage.
README.md	Updates limitations/roadmap documentation to mark v2 items as shipped.
railway.worker.toml	Sets worker replica count and documents scaling intent.
frontend/src/types.ts	Extends template union type to include beamer.
frontend/src/components/TemplateSelector.tsx	Adds “Beamer Slides” option to the template picker.
frontend/src/tests/TemplateSelector.test.tsx	Updates UI tests to expect the new template option.
DEPLOY.md	Updates scaling guidance to reflect worker replicas and S3 figure offload support.
app/templates/beamer.tex.j2	Introduces a Beamer LaTeX template.
app/style_map.example.json	Adds an example style-map configuration file for deployments.
app/storage.py	Introduces FigureStore abstraction with Redis + S3 implementations and fallback selection.
app/queue/worker.py	Switches figure persistence to FigureStore and threads extracted BibTeX into results/compile-check.
app/converter/style_map.py	Implements configurable style mapping and global STYLE_MAP singleton loader.
app/converter/run_merger.py	Adds adjacent-run merging to reduce LaTeX verbosity and wrapper fragmentation.
app/converter/renderer.py	Adds Beamer template support, frame segmentation rendering, and resolved citation rendering.
app/converter/parser.py	Integrates BibTeX extraction, style-map hooks, citation resolution, and run merging during parse.
app/converter/ir_schema.py	Extends IR/result schema to carry extracted BibTeX content.
app/converter/handlers/omml_direct.py	Adds dependency-free OMML→LaTeX converter implementation.
app/converter/handlers/equation_handler.py	Adds direct OMML fallback when Pandoc is unavailable or drops equations.
app/converter/compile_check.py	Writes references.bib for compile-check when BibTeX exists.
app/converter/bibliography.py	Implements Word Sources-part parsing and BibTeX emission.
app/config.py	Adds settings for figure storage backend selection and S3/R2 configuration.
app/api/routes.py	Updates template parameter, includes references.bib in zip output, and loads figures via FigureStore.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.