Paper Curation

Run automated curation

# Curate papers based on your config (topics, keywords, thresholds)
wikirecall papers curate

# Curate with custom topic filter
wikirecall papers curate --topics "LLM,agents,RAG"

Ingest papers into knowledge wiki

# Ingest a specific paper by ID
wikirecall papers ingest arxiv-2301-07041

# Batch ingest from last curation run
wikirecall papers ingest --all

The Curation → Ingest → Compound Loop

1. Discover — Search arXiv and Semantic Scholar for papers matching your topics and keywords
2. Score — Rank papers by relevance (topic match, keyword match, recency, citation count)
3. Filter — Apply minimum relevance threshold and daily paper limit
4. Ingest — Convert top papers into knowledge entities with YAML frontmatter
5. Link — Auto-detect related entities and create backlinks
6. Compound — Over time, your knowledge wiki grows richer with interconnected concepts

Tips for High-Signal Paper Finding

• Be specific with topics — "transformer attention" beats "machine learning"
• Use multiple keywords — combine technical terms: "flash attention", "efficient inference", "KV cache"
• Set a high min relevance — start with 0.3 and adjust based on signal-to-noise ratio
• Limit papers per day — 5-10 papers/day prevents information overload
• Review the ingestion log — check knowledge/ingestion-log.md to track what's been processed
• Cross-reference sources — papers found in both arXiv and Semantic Scholar tend to be higher quality
• Watch citation counts — high citations signal influential work, but don't ignore new papers

Configuration

Set up your curation config in your scenario or use CLI flags: