dsh-paper-reader
DeepSeek Harness paper-reading workbench whose PDF search reranks keyword candidates with Jev, falling back to keyword order.
Overview
dsh-paper-reader turns DeepSeek Harness, including the official DeepSeek desktop app, into a paper-reading workbench: a topic library, a PDF reader and an AI reading companion whose answers cite page numbers you can click to jump back to the source. It adds a coaching preset with step-by-step reading plans and quizzes, one-click Chinese translation with exportable bilingual PDFs, formula-preserving selection translation, side-by-side views and library management. Optional MinerU parsing and an embeddings endpoint improve scanned documents and search.
The decision-model part is narrow: according to the list summary, PDF transcription and search use keyword candidates reranked by Jev, with a fallback to keyword order when no key is set. Most of the plugin is unrelated UI and reading tooling. Documentation is in Chinese with an English README; the listing notes no licence file.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How relevant is this passage candidate to the reader's query?(inferred) | — |
How it works
- Upload PDFs into the library
- Keyword search finds candidate passages
- Jev reranks candidates (or keyword order without a key)
- Companion answers with clickable page citations
Strengths
- Works without a key via keyword fallback
- Page-cited answers
- Installable in the official desktop app
Caveats
- Jev plays a small role
- No licence file per the listing
- Chinese-first documentation
Getting started
dsh plugin --profile web add @ggboy123/dsh-paper-reader@latest (or add it from the desktop app's plugin page).At a glance
- Type
- Benchmark
- Status
- beta
- Language
- TypeScript
- Runtime
- DeepSeek Harness plugin
- Has code
- yes
Use cases and integrations
How curated lists describe it
DeepSeek Harness plugin that turns the workspace into a paper-reading bench: PDF transcription and search, with keyword candidates reranked by Jev and a fallback to keyword order without a key. Chinese and English; no licence file.
MrJev/awesome-jev
Frequently asked questions
Where is Jev used?
To rerank keyword search candidates over transcribed PDFs.
Does it need a Jev key?
No, without a key it falls back to keyword order.
Does it work in the DeepSeek desktop app?
Yes, it was tested on the official desktop app V0.2.0-rc.2.
Related projects
Assembles source-linked evidence for an AI task within a token budget using Jev relevance judgments and diversity-aware selection.
ProjectbetaPythonResearch agent where GPT plans and Jev finds evidence in parallel across document regions with Choice and Noul.
Browser extensionresearchTypeScriptMilvus bootcamp notebooks combining Gemini embeddings, Milvus retrieval and Jev judgments for reranking, filtering, routing and stopping.
ArticleproductionPythonLocal tool where Jev reads every accepted paper's abstract at NeurIPS, ICLR or ICML and judges relevance to your research topic.
ProjectbetaTypeScriptAgent skill that teaches coding agents to hand bulk semantic judgments to Jev and to find where a product could use Jev decisions.
Agent skillbetaWorked Jev examples for ticket triage and reranking, runnable through OpenRouter without an early-access key, with offline tests.
ProjectprototypePython