# token-monitor

Modular LLM API quota and usage visibility tool. Extracts rate-limit and usage data from all configured LLM providers before failures happen.

**Why it exists:** team-vigilio hit its 7-day rate limit (9+ days of 429s). api-ateam ran out of credit mid-session. We kept flying blind. This tool surfaces quota health before the failure.

## Quick start

```bash
node monitor.js          # run now — human-readable output + log
node analyze.js          # analyze accumulated logs — burn rates, rotation
~/os/token-status.sh     # what Vigilio's wake prompt sees (automated path)
```

## Usage

```bash
node monitor.js                          # human-readable summary + log to file
node monitor.js --json                   # JSON output + log to file
node monitor.js --summary                # human-readable only (no log)
node monitor.js --provider team-nadja    # single provider
node monitor.js --no-log                 # suppress log file write
```

## Analysis

```bash
node analyze.js                         # full report
node analyze.js --burn-rate             # burn rate per account
node analyze.js --weekly                # weekly budget reconstruction
node analyze.js --stagger               # reset schedule (next 48h)
node analyze.js --rotation              # rotation recommendation
node analyze.js --json                  # JSON output (all sections)
node analyze.js --provider team-nadja   # filter to one provider
node analyze.js --prune [--dry-run]     # archive and prune logs > 30 days
```

**Burn Rate** — delta analysis of 7d utilization over time, projected exhaustion at current rate.

**Reset Schedule** — providers resetting within the next 48 hours, sorted ascending by time to reset.

**Weekly Reconstruction** — peak and average 7d utilization per provider per ISO week. Shows exhaustion events.

**Rotation Recommendation** — ranked provider list by headroom, deprioritizing maxed/rejected/invalid-key accounts.

**Underspend Alerts** — active accounts with ≥ 40% of 5h window unused and < 2h until reset.

## Example output

```
Token Monitor — 2026-04-04 16:59 UTC
════════════════════════════════════════════════════════════

team-vigilio    [CRITICAL]  MAXED — 7d: 100% | resets in 23h 0m
team-ludo       [UNKNOWN]   Invalid API key (401)
team-molto      [OK]        5h: 32% | 7d: 45% | resets in 4h 0m
team-nadja      [OK]        5h: 52% | 7d: 39% | resets in 3h 0m
team-buio       [UNKNOWN]   Invalid API key (401)
shelley-proxy   [OK]        tokens: 4,800,000/4,800,000 | cost: $0.000013/call
api-ateam       [UNKNOWN]   Anthropic API does not expose billing/quota via REST

────────────────────────────────────────────────────────────
Overall: 1 CRITICAL, 3 OK, 3 UNKNOWN
```

## JSON output schema

```json
{
  "timestamp": "2026-04-04T16:59:50.000Z",
  "providers": {
    "team-vigilio": {
      "type": "teams-direct",
      "status": "rejected",
      "utilization_5h": 0.94,
      "utilization_7d": 1.0,
      "representative_claim": "seven_day",
      "reset_timestamp": 1743800000,
      "reset_in_seconds": 82800,
      "organization_id": "1d7653ad-...",
      "severity": "critical",
      "probe_latency_ms": 210
    },
    "shelley-proxy": {
      "type": "shelley-proxy",
      "status": "ok",
      "tokens_remaining": 4800000,
      "tokens_limit": 4800000,
      "requests_remaining": 20000,
      "requests_limit": 20000,
      "tokens_reset": "2026-04-04T17:59:50Z",
      "cost_per_call_usd": 0.000013,
      "severity": "ok",
      "probe_latency_ms": 180
    },
    "api-ateam": {
      "type": "api-direct",
      "status": "no_billing_data",
      "message": "Anthropic API does not expose billing/quota via REST.",
      "severity": "unknown",
      "probe_latency_ms": 0
    }
  }
}
```

## Provider support

| Provider | Type | Data available |
|----------|------|---------------|
| team-vigilio, team-ludo, team-molto, team-nadja, team-buio | Anthropic Teams (direct) | 5h/7d utilization (0–100%), status, reset countdown, severity |
| shelley-proxy | Shelley/exe.dev proxy | Token headroom, request headroom, per-call USD cost |
| api-ateam | Anthropic API (pay-per-use) | Key validity only — no billing API exists |
| google-gemini | Gemini API (free tier) | Quota violation detail, retry delay, key validity |
| xai-face, xai-amy, xai-murdock, xai-ba, xai-vigilio | xAI/Grok | Request/token remaining counts, rate limit status |

## Severity levels

| Level | Condition |
|-------|-----------|
| `critical` | Teams provider rejected (rate limited / budget exhausted) |
| `warning` | Teams 7d utilization > 85%, or 5h utilization > 70% |
| `ok` | Healthy |
| `unknown` | Invalid key (401), no billing data, or probe error |

## Rate-limit header schemas

**Teams direct** (oat01 keys — all `team-*` providers):
Anthropic Teams uses the **unified** schema. Headers are present on every response (200 and 429).
- `anthropic-ratelimit-unified-5h-utilization` — 5-hour window utilization (0.0–1.0)  
- `anthropic-ratelimit-unified-7d-utilization` — 7-day budget utilization (0.0–1.0)
- `anthropic-ratelimit-unified-status` — `allowed` | `rejected`
- `anthropic-ratelimit-unified-representative-claim` — which window is binding (`five_hour` | `seven_day`)
- `anthropic-ratelimit-unified-reset` — Unix timestamp when binding window resets

**Shelley proxy** (exe.dev gateway):
Uses classic absolute-count headers plus an exe.dev-injected cost header:
- `Anthropic-Ratelimit-Tokens-Remaining` / `Limit` — absolute token counts
- `Anthropic-Ratelimit-Requests-Remaining` / `Limit` — request counts
- `Exedev-Gateway-Cost` — USD cost per call (unique to exe.dev gateway)

## Log files

Each run appends to `~/.logs/token-monitor/YYYY-MM-DD.jsonl`:

```bash
# View today's log
cat ~/.logs/token-monitor/$(date +%Y-%m-%d).jsonl | \
  python3 -c "import sys,json; [print(json.loads(l)['ts'], json.loads(l)['providers']['team-vigilio']['severity']) for l in sys.stdin]"

# Check when team-vigilio was last healthy
grep '"team-vigilio"' ~/.logs/token-monitor/*.jsonl | \
  python3 -c "import sys,json; [print(l[:40]) for l in sys.stdin if json.loads(l.split(':',1)[1])['providers']['team-vigilio']['severity']=='ok']"
```

## Architecture

```
monitor.js              — CLI entrypoint, orchestrates probes
analyze.js              — analysis CLI (burn rates, weekly, stagger, rotation)
providers/
  index.js              — reads ~/.pi/agent/models.json, returns typed provider list
  anthropic-teams.js    — unified schema parser (oat01 keys, all team-* providers)
  anthropic-api.js      — pay-per-use (api03 keys) — reports "no billing data"
  shelley-proxy.js      — classic schema + Exedev-Gateway-Cost header
  gemini.js             — Gemini API (free tier, quota via response body)
  xai.js                — x.ai/Grok (rate-limit headers)
logger.js               — JSONL log to ~/.logs/token-monitor/
report.js               — human-readable summary + severity logic
test.js                 — test suite (run: node test.js)
docs/
  analyze.md            — analysis CLI full reference
```

## How to add a new provider

1. Add provider to `~/.pi/agent/models.json` with `"api": "anthropic-messages"`
2. If it uses Teams unified schema and its name starts with `team-`: picked up automatically
3. For a custom schema: create `providers/yourprovider.js`, implement `probeYourProvider()`, classify it in `providers/index.js`, add probe dispatch in `monitor.js`

## Probe behavior

The tool makes one minimal API call per provider to extract headers:
- Model: `claude-haiku-4-5-20251001` (cheapest available)
- Max tokens: 1 (minimizes cost)
- Rate-limit headers are returned on **every** response (200 and 429) for Teams providers
- A 429 from a maxed provider is expected and treated as valid quota data

**Stealth:** Read-only, one call per run, no hammering. Run at most once per session.

## Related

- `~/os/token-status.sh` — wake-prompt integration (calls monitor.js, formats for beat.sh)
- `~/projects/provider-check/` — predecessor (liveness only, no quota depth)
- `~/.pi/agent/models.json` — provider configuration source
- Forgejo issue: trentuna/token-monitor#1

## Wake integration

`~/os/token-status.sh` is the automated interface. It runs `monitor.js --json`
and formats the output into a compact summary block for injection into
Vigilio's wake prompt via `beat.sh`.

```bash
# Manual invocation (same as what the wake prompt sees)
~/os/token-status.sh

# Output format:
## Token Economics
Anthropic Teams (5 seats):
  team-vigilio     ✗ MAXED 7d:100% resets 23h
  team-molto       ✓ 5h:32% 7d:45% resets 4h
  ...
→ Current recommendation: use team-molto | avoid team-vigilio
```

File location: `~/os/token-status.sh`  
Called by: `~/os/beat.sh` (Vigilio wake script)  
Uses: `monitor.js --json` with 20-minute cache guard (won't double-probe within a session)