Top200 AI Papers Geo Publishing Pipeline

This folder is a standalone publish bundle extracted from the working branch. It is intended for sharing/releasing the publication flow separately.

What the pipeline does

The pipeline publishes the AI papers graph into a Geo space in 4 sequential scripts:

• 30a_foundation_people_papers1-30.ts
  • creates foundation entities: eras, domains, venues, datasets, concepts, organizations
  • creates missing people entities
  • publishes papers 1-30
• 30b_papers31-90.ts publishes papers 31-90
• 30c_papers91-150.ts publishes papers 91-150
• 30d_papers151-202.ts publishes the final papers batch
• 31_rollback_top200.ts rolls back entities and relations created by the 30a-30d layers using saved bounty_*.txt ops files
• 32_cleanup_top200_space.ts finds and deletes current Top200 entities in the target space by dataset names, including legacy papers created as Project

Data used

The scripts read JSON files from geo_publish_v2/. By default src/bounty_shared.ts auto-resolves the first existing directory from:

• ./geo_publish_v2

• ../geo_publish_v2

• entities: eras.json, domains.json, venues.json, datasets.json, concepts.json, organizations.json, persons.json, papers.json

• relations: rel_paper_person.json, rel_paper_venue.json, rel_paper_dataset.json, rel_paper_concept.json, rel_paper_org.json, rel_paper_era.json, rel_paper_domain.json

• images: paper_images_202/ (optional; papers still publish if an image is missing; auto-resolve order is ./paper_images_202 then ../paper_images_202)

If your data lives elsewhere, update DATA_DIR and IMAGES_DIR in src/bounty_shared.ts.

How publishing to Geo works

1. Each script builds Geo SDK operations (ops) for entities, relations, blocks, and images.
2. publishBatch writes ops snapshots to data_to_delete/ for audit/debug.
3. In live mode, publishOps resolves space type (PERSONAL or DAO) and submits/proposes the edit on testnet.

Dependencies

Install with Bun:

bun install

Main dependencies (from package.json):

• @geoprotocol/geo-sdk
• dotenv
• viem
• typescript (+ @types/node)

Environment

Use .env (or .env.example as a template):

DEMO_SPACE_ID=<target_space_id>
PK_SW=0x<private_key>

Optional:

DRY_RUN=1

Run order

bun run 30a_foundation_people_papers1-30.ts
bun run 30b_papers31-90.ts
bun run 30c_papers91-150.ts
bun run 30d_papers151-202.ts

Rollback dry-run / apply:

bun run 31_rollback_top200.ts
APPLY=true CONFIRM=ROLLBACK_TOP200 bun run 31_rollback_top200.ts

Cleanup current space data dry-run / apply:

bun run 32_cleanup_top200_space.ts
APPLY=true CONFIRM=DELETE_TOP200 bun run 32_cleanup_top200_space.ts

Both cleanup scripts default to BATCH_SIZE=1200. Override with BATCH_SIZE=<n> if you need smaller or larger transaction chunks.

Notes

• Keep this bundle and your exploratory scripts separate.