Critical: - nicobailon#1 pdf-extract: return full text inline instead of just file path - nicobailon#2 new semantic-scholar-search.ts + arxiv-search.ts providers (no API key needed) integrated into auto waterfall with academic query detection - nicobailon#3 brave: request up to 50 results when domain filter set (post-hoc filter fix) duckduckgo: inject site: operator for domain-include filters Major: - nicobailon#4 brave+searxng: build meaningful answer from snippets (not just 'Found N results') - nicobailon#5 searxng: massively expanded academic category inference regex + scholar category - nicobailon#6 search-filter: TIMEOUT_MS 15000→30000, MAX_TOKENS 2048→4096 - nicobailon#7 search-filter: QUALITY_TIERS now covers 15+ academic publisher domains updated DEFAULT_PROMPT with academic-specific synthesis rules - nicobailon#8 extract: arXiv abs pages → structured metadata via arXiv API (title/authors/abstract/categories/DOI) DOI URLs → Unpaywall open-access resolution before fallback Moderate: - nicobailon#9 index: updated tool descriptions to guide academic numResults (10-15) - nicobailon#10 perplexity: extract per-citation snippets from answer text via [N] markers - nicobailon#11 gemini-search: parallelize grounding chunk redirect resolution (Promise.all) - nicobailon#12 index: clarify curate=false recommended for automated agent pipelines - nicobailon#13 index: fetch_content description covers arXiv/PDF/DOI patterns - nicobailon#14 index: background fetch result instructions clarify get_search_content usage - nicobailon#15 extract: DOI → Unpaywall OA resolution before generic HTTP fallback - nicobailon#16 index: get_search_content description improved with per-parameter guidance

User Running in WSL #13

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions