feat(search): add --summarise flag for AI-powered abstract synthesis #14

Summary

Add a --summarise flag to mosaic search that generates a concise AI-written synthesis of all returned abstracts using the Claude API, printed above the results table.

Motivation

MOSAIC already integrates with Google NotebookLM for AI analysis of downloaded papers. A lightweight --summarise flag would provide immediate value without requiring a download step — especially useful when scanning many results quickly.

Proposed API

mosaic search "CRISPR off-target effects" --summarise
mosaic search "diffusion models" -n 20 --oa-only --summarise

Example output:

── AI Summary ────────────────────────────────────────────
The 18 returned papers span 2019–2024 and focus on three
main themes: (1) detection methods for off-target edits
using high-throughput sequencing, (2) engineered Cas9
variants with improved specificity, and (3) computational
prediction models. Recent work (2023–2024) increasingly
combines base editing with off-target profiling.
──────────────────────────────────────────────────────────

Implementation notes

• Optional dependency: anthropic SDK (pip install 'mosaic-search[ai]')
• Use the Claude API (model: claude-haiku-4-5 for speed and cost)
• Collect non-null abstracts from results, truncate to fit context window
• Gracefully skip if ANTHROPIC_API_KEY is not set, printing a hint
• Add --summarise flag only when the [ai] extra is installed