Admin: Sources

Source registry management, health monitoring, and manual ingestion controls. ← Admin home

Phase 2: Source data is loaded from the Cloudflare D1 database via the ingestion Worker API. In local dev, sources display from the seed data in db/seed.sql.

Source Health

Total sources

Degraded

Failing

Tier A — Auto-Ingest Sources

47 sources polled every 30 min–3 hours. Eligible for automatic publication.

Source	Type	Cadence	Status
OpenAI News	`rss`	1h	healthy
Anthropic Newsroom	`rss`	1h	healthy
Google DeepMind	`rss`	2h	healthy
Meta AI Blog	`rss`	2h	healthy
Mistral News	`rss`	1–2h	healthy
Hugging Face Blog	`rss`	2h	healthy
GitHub — llama.cpp	`github_api`	3h	healthy
GitHub — Ollama	`github_api`	3h	healthy
GitHub — MCP Spec	`github_api`	3h	healthy

Tier B — Research & Benchmarks

14 sources polled daily. Requires classification before publication.

Source	Type	Cadence	Status
arXiv cs.AI	`arxiv_api`	6h	healthy
arXiv cs.CL	`arxiv_api`	6h	healthy
arXiv cs.LG	`arxiv_api`	6h	healthy
MLCommons	`manual`	daily	healthy
Epoch AI Benchmarks	`manual`	daily	healthy
LMSYS Chatbot Arena	`manual`	daily	healthy
SWE-bench	`manual`	daily	healthy

Tier C — Discovery Signals

3 sources. Manual review only. Never auto-published as evidence.

Source	Type	Cadence	Status
Hacker News	`hackernews_api`	2h	healthy
Specialist publications	`rss/manual`	daily	healthy
Reddit (manual)	`manual`	daily	active

Manual Actions

Controls will be active when connected to the live Worker API. Full functionality in Phase 2 deployment.

View ingestion jobs → · Review queue →