Add Datacenter Monitor recipe

README.md

@@ -117,6 +117,7 @@ Recurring research, cron jobs, and webhook delivery.
 | [**Daily Insights**](typescript-recipes/parallel-daily-insights) | Cron-triggered daily research feed — runs Tasks on a schedule, persists to KV, publishes a public data feed. Includes a `SPEC.md` showing the task spec used. | `Task` `Webhooks` `Cron` | Cloudflare Workers · KV | – |
 | [**Vendor Intelligence**](typescript-recipes/parallel-vendor-intelligence) | Researches vendor risk, watches the structured report for changes, and runs follow-up research only when a change crosses the review threshold. | `Task` `Deep Research` `Monitors` | TypeScript · Local scripts | – |
 | [**Investor Signals + Sales Enrichment**](python-recipes/parallel-investor-signals) | Track VC funds *you* choose for new AI-native rounds — one daily Monitor per fund, each detection chain-verified by a follow-up Task, scored, CRM-checked, and posted to Slack — plus a cited company-enrichment app. Bring your own watchlist and CRM; includes an `AGENTS.md` one-command setup. | `Task` `Monitors` `Webhooks` `Cron` | Python · FastAPI · React · Vercel | – |
+| [**Datacenter Monitor**](typescript-recipes/parallel-datacenter-map) | Live map of ~2,700 US datacenters, each discovered, enriched across 25 fields, and AI-impact classified by the Task API, with 31 event-stream monitors plus 200 daily snapshot monitors watching the web. Per-field reasoning and citations via `output.basis`. Facility discovery uses shard-by-state plus `previous_interaction_id` pagination. | `Task` `Task Group` `Monitors` `Webhooks` `SSE` | Next.js · Vercel · Leaflet | [Live](https://datacenter-demo.app) |
 
 ### Deep Research & Notebooks
 

typescript-recipes/parallel-datacenter-map/.env.local.example

@@ -0,0 +1,14 @@
+# Parallel API key — required for monitors, facility discovery, and enrichment
+# Get one at https://platform.parallel.ai
+PARALLEL_API_KEY=your_parallel_api_key_here
+
+# Vercel Blob — stores per-facility enrichment basis (reasoning + citations)
+# Set after creating a Blob store in Vercel Dashboard → Storage → Blob
+ENRICHMENTS_BLOB_URL=https://your-store.private.blob.vercel-storage.com/enrichments.json
+BLOB_READ_WRITE_TOKEN=vercel_blob_rw_xxx
+
+# Anthropic API key — used by the weekly-brief writer (Claude)
+ANTHROPIC_API_KEY=your_anthropic_api_key_here
+
+# Optional: webhook URL for snapshot monitors (set to your deployed URL)
+# WEBHOOK_URL=https://your-app.vercel.app/api/webhook

typescript-recipes/parallel-datacenter-map/.gitignore

@@ -0,0 +1,62 @@
+# See https://help.github.com/articles/ignoring-files/ for more about ignoring files.
+
+# dependencies
+/node_modules
+/.pnp
+.pnp.*
+.yarn/*
+!.yarn/patches
+!.yarn/plugins
+!.yarn/releases
+!.yarn/versions
+
+# testing
+/coverage
+
+# next.js
+/.next/
+/out/
+
+# production
+/build
+
+# misc
+.DS_Store
+*.pem
+
+# debug
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+.pnpm-debug.log*
+
+# env files (can opt-in for committing if needed)
+.env*
+!.env.local.example
+
+# vercel
+.vercel
+
+# typescript
+*.tsbuildinfo
+next-env.d.ts
+
+# Large data files / run manifests (regenerated via scripts)
+public/data/enrichments.json
+src/data/backfill-results.json
+src/data/enrichment-runs.json
+src/data/enrichment-v2-runs.json
+src/data/ai-classification-runs.json
+.env*.local
+
+# Editor / OS
+*.code-workspace
+
+# Generated pipeline artifacts
+scripts/*.html
+scripts/newsletter-*.md
+scripts/newsletter-meta.json
+/datacenters.json
+/scripts/datacenters.json
+/datacenters.runs.jsonl
+/scripts/datacenters.runs.jsonl

typescript-recipes/parallel-datacenter-map/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Parallel Web Systems
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

typescript-recipes/parallel-datacenter-map/README.md

@@ -0,0 +1,159 @@
+# Datacenter Monitor
+
+A live map of U.S. datacenter infrastructure, built end-to-end on [Parallel](https://parallel.ai)'s **Task API** and **Monitor API**. Every facility on the map was *discovered*, *enriched*, and *classified* by Parallel — and 31 monitors watch the web for new developments, with each claim traceable to its sources.
+
+**Live demo:** https://datacenter-demo.app
+
+Everything here — the facility list, the 25 fields per facility, the AI-impact classifications, the weekly brief — is generated data. Nothing is hand-curated. This README documents exactly how it was produced so you can reproduce or extend it.
+
+> **Note on data:** this repo ships a **51-facility sample** (one per state) in `public/data/` so the app runs immediately after cloning. The [live demo](https://datacenter-demo.app) shows the full ~2,700-facility dataset. Run the pipeline below to regenerate the full set into `public/data/`. Monitor IDs in `src/data/` ship blank on purpose — run `setup-monitors.ts` and `create-snapshots.ts` to create your own against your account.
+
+---
+
+## How the data was built
+
+The whole dataset is a pipeline of Parallel Task API runs. Each stage is a script in [`scripts/`](./scripts).
+
+### 1. Discover the facilities — `build_datacenters_iterative.py`
+
+There is no clean public registry of U.S. datacenters, so we enumerate them with the Task API. The market is a long tail of ~1,800 colocation, enterprise, telecom, and edge operators, and enumerating them exhaustively is a decomposition problem: a broad, unscoped query surfaces the most prominent facilities first, so reaching the full tail takes two techniques:
+
+1. **Shard by geography.** Scope every query to *one state*. Enumerating a single state (reading facility directories) is the natural move, instead of dumping a global top-of-mind list. Ashburn alone goes 13 → ~90 this way.
+2. **Paginate via interactions (loop-until-dry).** After the first pass for a state, keep asking for *net-new* facilities while passing `previous_interaction_id`. The model carries its own memory of what it already returned in that thread, so we never paste a list of known facilities into the prompt. Loop until a page stops adding results.
+
+Everything is resumable: results checkpoint after each shard, and every `run_id` / `interaction_id` is logged so work that completed server-side is always recoverable.
+
+```bash
+export PARALLEL_API_KEY=...
+python scripts/build_datacenters_iterative.py                 # all 50 states + DC, until dry
+python scripts/build_datacenters_iterative.py --states "Texas,Ohio" --workers 8
+```
+
+This produced **2,811 unique facilities** (deduped to 2,694 shown on the map) across all 50 states + DC — each with a name, operator, owner, coordinates, and a source URL.
+
+### 2. Enrich each facility — `run-enrichment*.ts`
+
+Every facility is then deep-researched with the `ultra2x` processor against a **25-field JSON schema** (verified name/operator/owner, power capacity, sqft, cooling type, tier, fiber, utility provider, tax incentives, hazard zone, construction updates, recent news, tenants, and more). Runs are submitted in parallel via **Task Groups**, and Parallel's `output.basis` gives per-field **reasoning + citations** for every value — that's what powers the basis panel behind each cell.
+
+```bash
+npx tsx scripts/run-enrichment-v2.ts        # submit enrichment task group
+npx tsx scripts/collect-enrichments-v2.ts   # collect results
+npx tsx scripts/upload-per-facility.ts      # store per-facility basis in Vercel Blob
+```
+
+### 3. Classify AI impact — `classify-ai.ts`
+
+A second Task API pass classifies each facility's AI profile and community/resource impact against a structured schema: `ai_class` (ai-training / ai-inference / ai-mixed / cloud-hyperscale / not-ai), plus water impact, grid impact, and community-pushback levels — each with its own evidence and citations. ~710 facilities classified; ~520 flagged as AI/cloud infrastructure.
+
+### 4. Monitor the web — `setup-monitors.ts` + `create-snapshots.ts`
+
+- **31 event-stream monitors** watch for datacenter developments — power-grid changes, zoning decisions, ownership transfers, community opposition, new-site discovery — returning classified, cited events.
+- **200 daily snapshot monitors** re-verify facility fields once a day and surface field-level changes as diffs (with the re-verification's own reasoning + sources).
+
+---
+
+## What the app does
+
+- **Interactive map** of every facility, colored by lifecycle (operational / under construction / planned / decommissioned), with scope toggles (all vs. AI datacenters) and monitor-driven highlighting.
+- **Popups** with the enriched profile, AI-impact classification, and clickable primary sources loaded on demand.
+- **Dataset table** with all 25 fields per facility. Click any cell to open its **basis panel** — the reasoning and citations Parallel used to produce that value.
+- **Live monitor feed** with a chart-as-filter (break down by time / category / severity) and cited events.
+- **Weekly brief** — a newsletter deep-researched and written by the Task API across all monitors, every claim linked to a source.
+
+## Parallel APIs used
+
+**Task API**
+- Iterative facility discovery via interaction chaining (`previous_interaction_id`)
+- 25-field facility enrichment (`ultra2x`, Task Groups, structured output)
+- AI-impact classification (structured output)
+- Weekly brief research + writing
+- `output.basis` → per-field reasoning + citations everywhere
+
+**Monitor API**
+- 31 event-stream monitors (classified events with severity + citations)
+- 200 daily snapshot monitors (field-level change detection with diffs)
+- Webhooks + SSE for real-time updates
+
+## Tech stack
+
+- **Next.js 16** (App Router, TypeScript)
+- **react-leaflet** + Leaflet (map)
+- **Tailwind CSS** (Parallel design system)
+- **Vercel** (hosting, Blob storage, serverless, cron)
+
+---
+
+## Setup
+
+**Prerequisites:** Node.js 18+, Python 3.9+, a [Parallel API key](https://platform.parallel.ai), and a Vercel account (Blob storage + deploy).
+
+```bash
+git clone https://github.com/parallel-web/parallel-cookbook.git
+cd parallel-cookbook/typescript-recipes/parallel-datacenter-map
+npm install
+cp .env.local.example .env.local   # add your keys
+```
+
+Then run the pipeline to regenerate the full dataset (the repo ships only a 51-facility sample in `public/data/`; each stage is optional):
+
+```bash
+# 1. discover facilities
+pip install parallel-web
+python scripts/build_datacenters_iterative.py
+
+# 2. enrich + classify
+npx tsx scripts/run-enrichment-v2.ts
+npx tsx scripts/collect-enrichments-v2.ts
+npx tsx scripts/upload-per-facility.ts
+npx tsx scripts/classify-ai.ts
+
+# 3. set up monitors
+npx tsx scripts/setup-monitors.ts
+WEBHOOK_URL=https://your-app.vercel.app/api/webhook npx tsx scripts/create-snapshots.ts
+
+# 4. run
+npm run dev       # local
+vercel --prod     # deploy
+```
+
+## Architecture
+
+```
+Map view                         Monitor panel (right rail)
+- Leaflet map, 2,694 facilities  - Chart-as-filter (time / category / severity)
+- lifecycle + AI scope toggles   - Cited event feed
+- popups: profile, AI impact,    - Weekly brief reader
+  on-demand sources
+
+Dataset view
+- 25 enriched fields per facility
+- per-cell basis panel (reasoning + citations, from Vercel Blob)
+- monitor signals + snapshot diffs
+
+API routes
+  /api/monitors            live events from the 31 event-stream monitors
+  /api/snapshots           daily snapshot change detection (with basis)
+  /api/basis               per-facility reasoning + citations (Vercel Blob)
+  /api/webhook             receives monitor events (SSE to the client)
+  /api/newsletter/*        weekly brief generate / preview / issue list
+  /api/cron/newsletter     weekly brief cron
+```
+
+## Scripts
+
+| Stage | Script | Purpose |
+|-------|--------|---------|
+| Discover | `build_datacenters_iterative.py` | Enumerate U.S. facilities (shard-by-state + interaction pagination) |
+| Enrich | `run-enrichment.ts` / `run-enrichment-v2.ts` | Submit facility enrichment task groups (`ultra2x`) |
+| Enrich | `collect-enrichments*.ts` | Collect enrichment results |
+| Enrich | `upload-per-facility.ts` / `upload-enrichments.ts` | Store per-facility basis in Vercel Blob |
+| Enrich | `backfill-basis.ts` | Re-fetch a run's basis from its `run_id` onto a stored facility |
+| Classify | `classify-ai.ts` | AI-impact classification pass |
+| Monitor | `setup-monitors.ts` / `monitor-configs.ts` | Create the 31 event-stream monitors |
+| Monitor | `create-snapshots.ts` / `snapshots-to-daily.ts` | Create/tune daily snapshot monitors |
+| Monitor | `set-webhooks.ts` / `check-events.ts` | Register webhooks / inspect events |
+| Brief | `generate-issue.ts` / `seed-issue.ts` / `regenerate-brief.ts` | Generate the weekly brief |
+
+## License
+
+MIT — see [LICENSE](./LICENSE).

typescript-recipes/parallel-datacenter-map/eslint.config.mjs

@@ -0,0 +1,18 @@
+import { defineConfig, globalIgnores } from "eslint/config";
+import nextVitals from "eslint-config-next/core-web-vitals";
+import nextTs from "eslint-config-next/typescript";
+
+const eslintConfig = defineConfig([
+  ...nextVitals,
+  ...nextTs,
+  // Override default ignores of eslint-config-next.
+  globalIgnores([
+    // Default ignores of eslint-config-next:
+    ".next/**",
+    "out/**",
+    "build/**",
+    "next-env.d.ts",
+  ]),
+]);
+
+export default eslintConfig;

typescript-recipes/parallel-datacenter-map/next.config.ts

@@ -0,0 +1,7 @@
+import type { NextConfig } from "next";
+
+const nextConfig: NextConfig = {
+  /* config options here */
+};
+
+export default nextConfig;