Skip to content

workflows.md

Latest commit

 

History

History
189 lines (130 loc) · 6.38 KB

workflows.md

File metadata and controls

189 lines (130 loc) · 6.38 KB
Raw

alphaXiv CLI workflows

These are higher-level task workflows. They are not a second command reference.

Find important recent papers on a topic

Goal:

  • Find recent papers on a topic and rank candidates using alphaXiv's feed signals.

Recommended command order:

alphaxiv feed filters search "multi-agent systems"
alphaxiv feed list --interval 90-days --topic <topic-slug> --sort hot --limit 10
alphaxiv feed list --interval 90-days --topic <topic-slug> --sort likes --limit 10
alphaxiv feed list --interval 90-days --source github --topic <topic-slug> --sort most-stars --limit 10
alphaxiv paper summary <paper-id>

How to interpret the outputs:

  • feed filters search returns the actual topic slugs accepted by the feed.
  • feed list is the best surface for recent-paper discovery and rough importance ranking.
  • paper summary is the fastest way to validate whether a candidate is actually on-topic before deeper reading.

Common mistakes:

  • Using search papers when the real need is recent or ranked discovery.
  • Guessing topic slugs instead of discovering them with feed filters search.
  • Jumping into assistant before you have candidate paper ids.

Get the readable paper text or the PDF

Goal:

  • Get either the extracted paper body as text or the original PDF file.

Recommended command order:

alphaxiv context use paper <paper-id>
alphaxiv paper text
alphaxiv paper text --page 1 --page 2
alphaxiv paper pdf url
alphaxiv paper pdf download ./paper.pdf

How to interpret the outputs:

  • paper text returns readable text extracted from the PDF.
  • paper pdf url prints the resolved public PDF URL.
  • paper pdf download saves the original file locally.

Common mistakes:

  • Using paper pdf download when the user actually wants readable text.
  • Using paper summary when the task requires the paper body itself.

Inspect one paper deeply

Goal:

  • Understand what one paper is, what it claims, and what related resources exist.

Recommended command order:

alphaxiv context use paper <paper-id>
alphaxiv paper show
alphaxiv paper abstract
alphaxiv paper summary
alphaxiv paper overview
alphaxiv paper resources
alphaxiv paper similar --limit 5

How to interpret the outputs:

  • paper show confirms the ids, links, authors, and topics.
  • paper abstract gives the author-written abstract.
  • paper summary is the short AI digest.
  • paper overview is the long AI write-up when the summary is not enough.
  • paper resources surfaces BibTeX, transcript, links, and implementations.
  • paper similar is the best next step for adjacent papers.

Common mistakes:

  • Confusing abstract, summary, and overview.
  • Skipping paper show when the paper id or resolved identity is still uncertain.

Build and refine a shortlist

Goal:

  • Turn a rough set of candidates into a smaller, better-justified shortlist.

Recommended command order:

alphaxiv search papers "<keywords>"
alphaxiv paper summary <paper-id>
alphaxiv paper similar <paper-id> --limit 5
alphaxiv paper resources <paper-id>
alphaxiv context use paper <paper-id>
alphaxiv paper overview

How to interpret the outputs:

  • search papers works best when you already have a phrase or idea in mind.
  • paper similar expands one promising paper into a local neighborhood of related work.
  • paper resources is useful when code, implementations, or transcript artifacts affect the shortlist.
  • paper overview helps decide whether a paper belongs in the final cut.

Common mistakes:

  • Using only one signal or one query and never expanding with paper similar.
  • Treating assistant as the first-pass retrieval surface instead of using search, feed, and paper first.

Use the assistant after retrieval

Goal:

  • Use the authenticated assistant for comparison, synthesis, and follow-up after you already have paper ids or a specific paper in mind.

Auth setup note:

  • If assistant chat writes are restricted for the API key, run alphaxiv auth login-web once and keep ALPHAXIV_HOME and browser-profile stable. Do not rerun it before every assistant command.

Recommended command order:

alphaxiv assistant start "Find the main contribution and weakness of this paper" --paper <paper-id>
alphaxiv assistant reply "Compare it with the previous two papers we discussed"
alphaxiv assistant history
alphaxiv assistant list --paper <paper-id>

How to interpret the outputs:

  • assistant start begins a new assistant chat and saves it as the current session.
  • assistant reply continues the current or selected session.
  • assistant history is the transcript view for the saved session.
  • assistant list --paper is the best way to inspect the saved chats that belong to one paper.

Long-session note:

  • assistant reply keeps extending the same remote chat session.
  • Very long sessions can become slower to answer over time.
  • If latency starts climbing, start a fresh chat with assistant start and restate only the context, recap, or paper ids you still need.
  • Keep the older thread available through assistant history or assistant list.

Common mistakes:

  • Assuming assistant start automatically uses saved paper context. It does not; use --paper <paper-id> when you want paper grounding in a new chat.
  • Using assistant as the first-step discovery mechanism when search, feed, or paper would be more deterministic.
  • Repeating auth login-web on every use instead of keeping the persistent browser profile.

Work with folders and comments when authenticated

Goal:

  • Save papers into folders, inspect folder contents, and participate in paper comment threads.

Recommended command order:

alphaxiv auth set-api-key --api-key "$ALPHAXIV_API_KEY"
alphaxiv folders list
alphaxiv paper folders list <paper-id>
alphaxiv paper folders add <paper-id> "Want to read"
alphaxiv paper comments list <paper-id>
alphaxiv paper comments add <paper-id> --body "Helpful note"
alphaxiv paper comments reply <paper-id> <comment-id> --body "Thanks"

How to interpret the outputs:

  • folders list shows the user's authenticated folder set.
  • paper folders list shows which of those folders already contain the selected paper.
  • paper comments list is public read access for the thread.
  • paper comments add and reply are authenticated mutations.

Common mistakes:

  • Forgetting that folders and comment mutations require an API key.
  • Using paper comments delete or upvote without an actual comment id.