Skip to content

Add ISGT Europe source via the open DBLP API#15

Merged
picaultj merged 1 commit into
mainfrom
claude/paper-browser-llm-classify-20e9rk
Jul 10, 2026
Merged

Add ISGT Europe source via the open DBLP API#15
picaultj merged 1 commit into
mainfrom
claude/paper-browser-llm-classify-20e9rk

Conversation

@picaultj

Copy link
Copy Markdown
Owner

Assessment of the three requested venues

All three are IEEE PES conferences whose full text lives on IEEE Xplore (paywalled, no API key, scraping disallowed). I checked the open alternatives:

Venue Open source? Result
ISGT Europe ✅ indexed by DBLP (editions 2017–2024) Added (title-based)
PowerTech ❌ zero DBLP records; IEEE Xplore only Not added
PES General Meeting ❌ not on DBLP; IEEE Xplore only Not added

PowerTech and PES-GM can't be supported without an IEEE Xplore API key. (Both requested editions — PowerTech 2027, and any 2026 ISGT — are also future/unpublished; the DBLP adapter will pick up ISGT Europe 2026 automatically once it's indexed.)

What's added

A generic DBLPSource that lists a conference edition through DBLP's public search API:

  • resolves "<venue> <year>" (e.g. isgteurope 2024), a DBLP proceedings key, or a DBLP URL to a toc: query;
  • pages the results at 100/request (DBLP's per-request cap) with a politeness delay, deduping by DBLP key;
  • extracts title, authors (DBLP "0001"-style disambiguation suffixes stripped) and a DOI link.

DBLP exposes no abstracts, so classification is title-based (same as PSCC). Registered as the isgteurope source (default isgteurope 2024); because the adapter is generic, any DBLP-indexed venue works by passing <venue> <year>.

Also hardened _robust_get to retry ConnectionError (peer resets, e.g. DBLP rate-limiting) — a general robustness win for all sources.

Testing

  • uv run ruff check . — clean
  • uv run pytest -q60 passed (new tests/test_dblp_source.py: proc-key normalisation, hit parsing across author shapes + cleanup, ee-list, dedup/untitled, registry wiring)
  • uv build — builds
  • Live against the real DBLP API: fetched all 423 ISGT Europe 2024 papers (5 pages, deduped, correct authors/DOIs); instant cache hit on re-run.
  • Boot smoke test — HTTP 200, the ISGT Europe option is served in the UI.

Docs

README (intro, diagram/table, Configuration + a "title-based sources" note explaining why PowerTech / PES-GM can't be added), AGENTS, and docs/ARCHITECTURE.md updated.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NSHHDVFvrPSbRj4bcnKwUq


Generated by Claude Code

Assessed the three requested IEEE PES venues: only ISGT Europe is reachable
through an open source (the public DBLP search index) — PowerTech and the PES
General Meeting aren't indexed by DBLP and are IEEE-Xplore-only (paywalled, no
API key), so they can't be added without IEEE Xplore access.

Adds a generic DBLPSource that lists a conference edition via DBLP's publ search
API (toc: query, paged at 100/request with a politeness delay), extracting
title, authors (disambiguation suffixes stripped) and a DOI link. DBLP has no
abstracts, so classification is title-based (like PSCC). Registered as the
"isgteurope" source (default "isgteurope 2024"); the adapter is generic and
works for any DBLP-indexed venue via "<venue> <year>".

Also hardened _robust_get to retry ConnectionError (peer resets, e.g. DBLP
rate-limiting), and added an isgteurope no-results hint.

Tests cover proc-key normalisation, hit parsing (list/single author shapes,
disambiguation + trailing-dot cleanup, ee-list), dedup/untitled handling, and
registry wiring. README, AGENTS and ARCHITECTURE document the source and note
that PowerTech / PES GM can't be added.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NSHHDVFvrPSbRj4bcnKwUq
@picaultj picaultj merged commit 981a417 into main Jul 10, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants