Skip to content
Skills Library
DevOps & cloud · Automation

wave

/wave

Farm wave management — create Railway accounts, spawn workers, deploy configs, track wave progress. Use for scaling training farm.

Repository
trinity ↗
Licence
MIT
Last repository change
27.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 27.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Farm wave management — create Railway accounts, spawn workers, deploy configs, track wave progress. Use for scaling training farm.

Arguments

/wave [status|spawn <account> <count>|deploy|progress]
Original author description

Farm wave management — create Railway accounts, spawn workers, deploy configs, track wave progress. Use for scaling training farm.

How it works

  1. Mode Detection
  2. Data Collection (all modes)
  3. MODE=STATUS
  4. MODE=SPAWN
  5. MODE=DEPLOY
  6. MODE=PROGRESS
  7. Telegram Notification
Original SKILL.md headings · 184 lines, 651 words
SKILL.md preview
---
name: wave
description: Farm wave management — create Railway accounts, spawn workers, deploy configs, track wave progress. Use for scaling training farm.
argument-hint: [status|spawn <account> <count>|deploy|progress]
allowed-tools: Bash(tri *), Bash(curl *), Bash(set *), Bash(source *), Bash(grep *), Bash(echo *), Bash(sleep *), Read, Grep, Glob
context: fork
---

For system state collection, follow `.claude/skills/_shared/data_collection.md`.
For Telegram output, follow `.claude/skills/_shared/telegram.md`.

## Mode Detection

Check $ARGUMENTS:
- "status" or empty → **MODE=STATUS** — show all accounts, slots used/free, wave progress
- "spawn <account> <count>" → **MODE=SPAWN** — create N services on account with phi-grid configs
- "deploy" → **MODE=DEPLOY** — trigger deploys on all pending services
- "progress" → **MODE=PROGRESS** — track wave completion (building/running/failed counts)

## Data Collection (all modes)

```bash
# Source tokens
set -a && source .env && set +a

# Read farm state
cat .trinity/railway_farm.json 2>/dev/null || echo "NO_FARM_STATE"

# Read evolution state
cat .trinity/evolution_state.json 2>/dev/null || echo "NO_EVOLUTION"
```

## MODE=STATUS

Show all accounts with slot utilization:

```bash
python3 << 'STATUS_EOF'
import json, os

farm_file = ".trinity/railway_farm.json"
if not os.path.exists(farm_file):
    print("No farm state. Run: tri farm scan")
    exit()

with open(farm_file) as f:
    farm = json.load(f)

accounts = farm.get("accounts", [])
cap = farm.get("capacity", {})

print("=" * 60)
print("  🌊 WAVE STATUS — HSLM Training Farm")
print("=" * 60)

total_active = 0
total_slots = 0
for acct in accounts:
    name = acct.get("name", "unknown")
    active = acct.get("active_services", 0)
    slots = acct.get("total_slots", 10)
    free = slots - active
    total_active += active
    total_slots += slots
    bar = "█" * active + "░" * free
    status = "🟢" if active > 0 else "⚪"
    print(f"  {status} {name:20s} [{bar}] {active}/{slots}")

print(f"\n  Total: {total_active}/{total_slots} ({total_active*100//total_slots}%)")
print(f"  Free slots: {total_slots - total_active}")

# Show wave progress from evolution state
evo_file = ".trinity/evolution_state.json"
if os.path.exists(evo_file):
    with open(evo_file) as f:
        evo = json.load(f)
    gen = evo.get("generation", 0)
    print(f"\n  Generation: {gen}")
    pop = evo.get("population", [])
    if pop:
        best = min(pop, key=lambda x: x.get("ppl", 999))
        print(f"  Best: {best.get('name', '?')} PPL={best.get('ppl', '?')}")
STATUS_EOF
```

## MODE=SPAWN

Create N training services on specified account with golden config mutations.

**Golden config baseline:** LAMB 1e-3, cosine, batch=66, ctx=81, grad_clip=1.0

**Sacred mutations (phi-grid):**
- LR: [3e-4, 5e-4, 1e-3, 1.618e-3, 3e-3] (phi-spaced)
- Optimizer: [adam, adamw, lamb]
- Batch: [33, 66, 99] (multiples of 3)
- Context: [27, 81, 243] (powers of 3)
- Schedule: cosine ONLY (flat = BANNED)

**Safeguards:**
- NEVER use flat LR schedule
- NEVER set startCommand — use Dockerfile ENTRYPOINT
- ALWAYS set builder: NIXPACKS via serviceInstanceUpdate
- ALWAYS set dockerfilePath: "Dockerfile.hslm-train"
- ALWAYS set minimum env vars: HSLM_OPTIMIZER, HSLM_LR, HSLM_LR_SCHEDULE
- ALWAYS cosine schedule — zero exceptions

## MODE=DEPLOY

Trigger redeployments on all services that are pending/failed:

```bash
# Query all accounts for services needing deploy
python3 << 'DEPLOY_EOF'
import json, os, subprocess

farm_file = ".trinity/railway_farm.json"
if not os.path.exists(farm_file):
    print("No farm state")
    exit()

with open(farm_file) as f:
    farm = json.load(f)

pending = []
for acct in farm.get("accounts", []):
    token_var = acct.get("token_var", "RAILWAY_API_TOKEN")
    token = os.environ.get(token_var, "")
    if not token:
        continue
    for svc in acct.get("services", []):
        if svc.get("status") in ("FAILED", "NOT_DEPLOYED", "CRASHED"):
            pending.append({"name": svc["name"], "id": svc["id"], "token": token})

if not pending:
    print("✅ No services need deployment")
else:
    print(f"🔨 {len(pending)} services need deployment")
    for svc in pending:
        print(f"  • {svc['name']} ({svc['id'][:8]}...)")
DEPLOY_EOF
```

## MODE=PROGRESS

Track wave completion metrics:

```bash
python3 << 'PROGRESS_EOF'
import json, os

farm_file = ".trinity/railway_farm.json"
if not os.path.exists(farm_file):
    print("No farm state")
    exit()

with open(farm_file) as f:
    farm = json.load(f)

statuses = {"ACTIVE": 0, "BUILDING": 0, "DEPLOYING": 0, "FAILED": 0, "CRASHED": 0, "OTHER": 0}
total = 0
for acct in farm.get("accounts", []):
    for svc in acct.get("services", []):
        total += 1
        s = svc.get("status", "OTHER")
        if s in statuses:
            statuses[s] += 1
        else:
            statuses["OTHER"] += 1

print("=" * 60)
print("  🌊 WAVE PROGRESS")
print("=" * 60)
print(f"  🟢 Training: {statuses['ACTIVE']}")
print(f"  🔨 Building: {statuses['BUILDING'] + statuses['DEPLOYING']}")
print(f"  🔴 Failed:   {statuses['FAILED'] + statuses['CRASHED']}")
print(f"  Total:       {total}")
pct = statuses['ACTIVE'] * 100 // total if total else 0
print(f"\n  Completion: {pct}%")
bar = "█" * (pct // 5) + "░" * (20 - pct // 5)
print(f"  [{bar}]")
PROGRESS_EOF
```

## Telegram Notification

After rendering output, if significant state change detected:

Set `TG_TEXT` to wave summary (active/total, best PPL, generation).
Set `TG_MODE=dedup`, `TG_DEDUP_FILE=.trinity/tg_dedup_wave.hash`.
Then execute the shared Telegram template from `.claude/skills/_shared/telegram.md`.

Versions

RepositoryScoreLicence
This page · trinity-s-ai-dna/trinity100MIT
apideck-libraries/api-skills100Apache-2.0

Identical content appears in 1 other GitHub folders. We show one copy with the highest score.

Database snapshot: 13.11.2026 · package version 7e96d810b9d9 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 95railway-deploy
    Deploy code to Railway using "railway up". Use when user wants to push code, says "railway up", "deploy", "ship", or "push". For initial setup or creating services, use railway-new skill. For Docker images, use railway-environment skill.
  • 100railway-deploy
    Deploys and manages apps on Railway with the Railway CLI: project setup, deploys, services, databases, environment variables, logs and scaling. Use when a user asks to deploy to Railway, check deployment status, set environment variables on Railway, view Railway logs, link a…
  • 100railway-deploy
    Deploy projects to Railway — Dockerfile creation, railway.toml config, environment variables, multi-service setups (web + worker), and deployment via CLI or GraphQL API. Trigger on "deploy to Railway", "ship it on Railway", or "set up Railway hosting", and on questions about…
  • 95railway-new
    Create Railway projects, services, and databases with proper configuration. Use when user says "setup", "deploy to railway", "initialize", "create project", "create service", or wants to deploy from GitHub. Handles initial setup AND adding services to existing projects. For…
  • 95memstack-deployment-railway-deploy
    Use this skill when the user says 'deploy to Railway', 'Railway setup', 'railway-deploy', or needs to deploy a Node.js, Python, or Docker application to Railway with environment variables, custom domains, and monitoring. Do NOT use for Netlify, Vercel, or Hetzner deployments.
  • 90devops/deploy-railway
    Railway 部署方法论,使用 railway CLI 部署后端服务(Node.js、Python、Go、Java、Docker)和托管数据库(PostgreSQL、MySQL、Redis)

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.