@1930dev/opencode-usage 0.3.3 → 0.3.5

package/README.md

@@ -58,6 +58,10 @@ opencode-usage usage --pct              # add the % BUDGET column (--by provider
 opencode-usage providers
 opencode-usage providers --no-net       # cached only
 
+# What the providers themselves answer about their budget
+opencode-usage probe                    # one free-model call per provider
+opencode-usage probe --json
+
 # Model ranking by intelligence per blended dollar
 opencode-usage top
 opencode-usage top --limit 10 --json
@@ -124,7 +128,11 @@ Normalized percentage of budget consumed per provider, from these sources (in pr
    - `opencode-go` (Zen): rolling 5h / weekly / monthly % (binding window)
    - `openrouter`: credits used / total credits
    - `orcarouter`: spend since top-up; a per-key credit cap turns into a %
-   - `zai`: coding plan quota (needs an active GLM coding plan)
+   - `zai`: coding plan quota when the key has an active GLM plan; otherwise the
+     wallet balance (fallback)
+   - `snowflake-cortex`: tokens in the last 30d from the account usage views
+     (SQL API); the percentage comes from a `snowflake-cortex` line in
+     `budgets.json`
 
 2. **Documented limits** — published quotas, used when the provider reports none.
    Each one is read against the period it resets on: a daily limit against today,
@@ -142,8 +150,8 @@ Normalized percentage of budget consumed per provider, from these sources (in pr
    - `cloudflare-workers-ai`: 10k neurons/day (free tier, resets 00:00 UTC)
    - `nvidia`: credit allowance retired — the trial is rate-limited per model
    - `opencode`: no spend API for API keys
-   - `snowflake-cortex`: 100 credits/month, billed per token
-   - `zai`: no coding plan on this key
+   - `zai`: a key without a coding plan shows its wallet balance (live), and the
+     percentage comes from a `zai` line in `budgets.json`
 
 3. **budgets.json** — your monthly USD per provider, read against what the
    provider cost so far this calendar month:
@@ -157,6 +165,23 @@ Normalized percentage of budget consumed per provider, from these sources (in pr
 
 Providers without any source show `—`.
 
+## Free-model probe (`probe`)
+
+`probe` sends one minimal request against a free model per connected provider
+and reports what the answer reveals about the budget: a rate-limit header, a
+"no credits" error, or the meter itself (`usage.neurons` for Cloudflare). It
+reads the key from opencode's auth store and never prints it. Providers covered:
+
+- `cloudflare-workers-ai` — the response carries neurons per request (the free
+  tier's unit), so the probe answers whether the account can still run
+- `nvidia` — an HTTP 200 means the free per-model allowance still works
+- `orcarouter` — the `free` model answers `rate_limit_error` with a
+  `retry-after`, and credits raise the cap
+- `zai` — an HTTP 429 with `code 1113` means the balance is empty
+
+`opencode` exposes no chat API for keys and `snowflake-cortex` is read through
+its SQL quota, so both are skipped.
+
 ## Ranking (`top`)
 
 Models ranked by intelligence index per blended dollar:

package/dist/cli.js

@@ -284,12 +284,39 @@ function parseZai(data) {
     raw: data
   };
 }
+function parseZaiWallet(data) {
+  const failed = data.success === false || data.code !== undefined && data.code !== 0 && data.code !== 200;
+  if (failed) {
+    return { provider: "zai", ok: false, detail: data.msg ?? "wallet endpoint error", windows: [] };
+  }
+  const balance = data.data?.availableBalance ?? data.data?.rechargeAmount ?? 0;
+  const spend = data.data?.totalSpendAmount ?? 0;
+  const windows = [
+    { label: "balance", percentUsed: 0, detail: `$${balance.toFixed(2)}` },
+    { label: "spend", percentUsed: 0, detail: `$${spend.toFixed(2)}` }
+  ];
+  const today = data.data?.todaySpendAmount;
+  if (typeof today === "number") {
+    windows.push({ label: "today", percentUsed: 0, detail: `$${today.toFixed(2)}` });
+  }
+  return {
+    provider: "zai",
+    ok: true,
+    detail: `balance $${balance.toFixed(2)}`,
+    windows,
+    raw: data
+  };
+}
+var QUOTA_URL = "https://api.z.ai/api/monitor/usage/quota/limit";
+var WALLET_URL = "https://api.z.ai/api/biz/account/query-customer-account-report";
 async function fetchZai(key) {
   try {
-    const data = await getJson("https://api.z.ai/api/monitor/usage/quota/limit", {
-      Authorization: `Bearer ${key}`,
-      Accept: "application/json"
-    });
+    const auth = { Authorization: `Bearer ${key}`, Accept: "application/json" };
+    const data = await getJson(QUOTA_URL, auth);
+    if (data.code !== undefined && data.code !== 0) {
+      const wallet = await getJson(WALLET_URL, auth);
+      return parseZaiWallet(wallet);
+    }
     return parseZai(data);
   } catch (err) {
     return { provider: "zai", ok: false, detail: err.message, windows: [] };
@@ -338,11 +365,13 @@ function parseOrcaRouter(usage, subscription) {
   const cap = hard > 0 && hard < NO_CAP ? hard : soft > 0 && soft < NO_CAP ? soft : 0;
   const capped = cap > 0;
   const pct = capped ? spent / cap * 100 : 0;
+  const hint = capped ? "" : " (no cap set)";
+  const detail = `$${spent.toFixed(2)} spent${hint}`;
   return {
     provider: "orcarouter",
     ok: true,
-    detail: `$${spent.toFixed(2)} spent`,
-    windows: [{ label: "spend", percentUsed: pct, detail: `$${spent.toFixed(2)}` }],
+    detail,
+    windows: [{ label: "spend", percentUsed: pct, detail: `$${spent.toFixed(2)}${hint}` }],
     budget: capped ? { percentUsed: pct, label: `$${cap}` } : undefined,
     raw: { usage, subscription }
   };
@@ -358,6 +387,56 @@ async function fetchOrcaRouter(key) {
   }
 }
 
+// packages/core/src/quota/snowflake.ts
+var SQL_API = (account) => `https://${account}.snowflakecomputing.com/api/v2/statements`;
+var STATEMENT = `SELECT
+  SUM(CASE WHEN START_TIME >= DATE_TRUNC('day', CURRENT_TIMESTAMP())
+      AND START_TIME < DATEADD('day', 1, DATE_TRUNC('day', CURRENT_TIMESTAMP()))
+      THEN TOKENS ELSE 0 END) AS today_tokens,
+  SUM(TOKENS) AS month_tokens
+FROM SNOWFLAKE.ACCOUNT_USAGE.CORTEX_REST_API_USAGE_HISTORY
+WHERE START_TIME >= DATEADD('day', -30, CURRENT_TIMESTAMP())`;
+var fmtTokens = (n) => n.toLocaleString("en-US");
+function parseSnowflake(row) {
+  const today = Number(row?.[0] ?? 0);
+  const month = Number(row?.[1] ?? 0);
+  return {
+    provider: "snowflake-cortex",
+    ok: true,
+    detail: `${fmtTokens(month)} tokens/30d`,
+    windows: [
+      { label: "tokens/mo", percentUsed: 0, detail: `${fmtTokens(month)}` },
+      { label: "today", percentUsed: 0, detail: `${fmtTokens(today)}` }
+    ],
+    raw: { row }
+  };
+}
+async function fetchSnowflake(key, metadata) {
+  const account = String(metadata?.account ?? "").toUpperCase();
+  if (!account) {
+    return { provider: "snowflake-cortex", ok: false, detail: "metadata.account is required", windows: [] };
+  }
+  try {
+    const res = await fetch(SQL_API(account), {
+      method: "POST",
+      headers: {
+        Authorization: `Bearer ${key}`,
+        "Content-Type": "application/json",
+        Accept: "application/json"
+      },
+      body: JSON.stringify({ statement: STATEMENT, timeout: 30 })
+    });
+    const body = await res.json().catch(() => null);
+    if (!res.ok) {
+      return { provider: "snowflake-cortex", ok: false, detail: body?.message ?? `HTTP ${res.status}`, windows: [] };
+    }
+    const rows = Array.isArray(body?.data) ? body.data : [];
+    return parseSnowflake(rows[0]);
+  } catch (err) {
+    return { provider: "snowflake-cortex", ok: false, detail: err.message, windows: [] };
+  }
+}
+
 // packages/core/src/providers.ts
 var TTL_MS = 15 * 60 * 1000;
 async function readCache() {
@@ -381,7 +460,8 @@ var FETCHERS = {
   "github-copilot": fetchCopilot,
   zai: fetchZai,
   amd: fetchAmd,
-  orcarouter: fetchOrcaRouter
+  orcarouter: fetchOrcaRouter,
+  "snowflake-cortex": fetchSnowflake
 };
 async function providerStatuses(opts) {
   const auth = await readAuth();
@@ -402,7 +482,7 @@ async function providerStatuses(opts) {
       const secret = authSecret(auth[p]);
       if (!fetcher || !secret)
         return null;
-      return fetcher(secret);
+      return fetcher(secret, auth[p].metadata);
     }));
     for (const q of live) {
       if (q)
@@ -425,7 +505,7 @@ async function providerStatuses(opts) {
   }));
 }
 // packages/core/src/report.ts
-function fmtTokens(n) {
+function fmtTokens2(n) {
   if (n >= 1e9)
     return `${(n / 1e9).toFixed(1)}B`;
   if (n >= 1e6)
@@ -446,6 +526,15 @@ function fmtAgo(ms, now = Date.now()) {
     return `${h}h ago`;
   return `${Math.floor(h / 24)}d ago`;
 }
+function fmtRetry(seconds) {
+  if (seconds < 60)
+    return `${seconds}s`;
+  if (seconds < 3600)
+    return `${Math.round(seconds / 60)}m`;
+  if (seconds < 86400)
+    return `${Math.ceil(seconds / 3600)}h`;
+  return `${Math.ceil(seconds / 86400)}d`;
+}
 function pad(s, n) {
   return s.length >= n ? s : s + " ".repeat(n - s.length);
 }
@@ -467,8 +556,8 @@ function usageTable(rows, totals, sinceLabel, groupBy, pct) {
     const row = [
       r.group,
       String(r.messages),
-      fmtTokens(r.tokensInput),
-      fmtTokens(r.tokensOutput),
+      fmtTokens2(r.tokensInput),
+      fmtTokens2(r.tokensOutput),
       fmtCost(r.cost)
     ];
     if (pct) {
@@ -485,7 +574,7 @@ function usageTable(rows, totals, sinceLabel, groupBy, pct) {
 [sources: ${[...pct.values()].map((p) => `${p.source}${p.label ? `(${p.label})` : ""}`).filter((v, i, a) => a.indexOf(v) === i).join(", ")}]` : "";
   return `${t}
 
-TOTAL since ${sinceLabel}: ${totals.messages} msgs, ${fmtTokens(totals.tokensInput)} in / ${fmtTokens(totals.tokensOutput)} out, ${fmtCost(totals.cost)} est.${sourceNote}`;
+TOTAL since ${sinceLabel}: ${totals.messages} msgs, ${fmtTokens2(totals.tokensInput)} in / ${fmtTokens2(totals.tokensOutput)} out, ${fmtCost(totals.cost)} est.${sourceNote}`;
 }
 function providersTable(statuses, local) {
   const headers = ["PROVIDER", "QUOTA", "USAGE 7D (LOCAL)", "LAST USED"];
@@ -498,6 +587,18 @@ function providersTable(statuses, local) {
   });
   return table(headers, rows);
 }
+function probeTable(results) {
+  const headers = ["PROVIDER", "TAUGHT", "SIGNAL", "RETRY"];
+  const rows = results.map((r) => {
+    if (r.status === 0)
+      return [r.provider, "error", r.message ?? "no signal", "\u2014"];
+    const taught = r.ok ? "yes" : "no";
+    const signal = r.ok ? r.usage && Object.keys(r.usage).length > 0 ? Object.entries(r.usage).map(([k, v]) => `${k}=${v}`).join(", ") : "http ok" : r.message || (r.code ? `code ${r.code}` : `HTTP ${r.status}`);
+    const retry = r.retryAfterSeconds !== undefined ? fmtRetry(r.retryAfterSeconds) : "\u2014";
+    return [r.provider, taught, signal, retry];
+  });
+  return table(headers, rows);
+}
 // packages/core/src/ranking/fetch.ts
 import { mkdir as mkdir2, readFile as readFile3, writeFile as writeFile2 } from "fs/promises";
 import path3 from "path";
@@ -735,7 +836,7 @@ var PROVIDER_LIMITS = [
     limit: 100,
     tier: "paid",
     source: "https://docs.snowflake.com/en/sql-reference/account-usage/cortex_ai_functions_usage_history",
-    note: "Cortex AI Functions bill per token (credits per million tokens, Service Consumption Table). Usage lives in ACCOUNT_USAGE views, which the API key alone cannot read."
+    note: "Billed per token (Service Consumption Table). Live tokens come from ACCOUNT_USAGE.CORTEX_REST_API_USAGE_HISTORY through the SQL API with the stored JWT; the percentage needs a snowflake-cortex line in budgets.json."
   },
   {
     provider: "zai",
@@ -795,6 +896,89 @@ function resolvePct(provider, sources) {
   }
   return { pct: 0, source: "none" };
 }
+// packages/core/src/probe.ts
+function numericUsage(u) {
+  if (typeof u !== "object" || u === null)
+    return;
+  const out = {};
+  for (const [k, v] of Object.entries(u)) {
+    if (typeof v === "number")
+      out[k] = v;
+  }
+  return Object.keys(out).length > 0 ? out : undefined;
+}
+function redact(text, key) {
+  return key.length > 6 ? text.split(key).join("[redacted]") : text;
+}
+var chatBody = (model) => JSON.stringify({ model, messages: [{ role: "user", content: "ping" }], max_tokens: 4 });
+async function probeChat(p, key) {
+  let res;
+  try {
+    res = await fetch(p.url, {
+      method: "POST",
+      headers: {
+        "Content-Type": "application/json",
+        Accept: "application/json",
+        Authorization: `Bearer ${key}`
+      },
+      body: p.body
+    });
+  } catch (err) {
+    return { provider: p.name, status: 0, ok: false, message: redact(err.message, key) };
+  }
+  const text = await res.text();
+  let parsed = null;
+  try {
+    parsed = JSON.parse(text);
+  } catch {
+    parsed = null;
+  }
+  const obj = parsed ?? {};
+  const err = obj.error;
+  const firstErr = Array.isArray(obj.errors) ? obj.errors[0] : undefined;
+  const message = redact(err?.message ?? err?.detail ?? firstErr?.message ?? "", key);
+  const rawRetry = res.headers.get("retry-after");
+  const headerRetry = rawRetry !== null && rawRetry !== "" ? Number(rawRetry) : Number.NaN;
+  const retry = Number.isFinite(headerRetry) ? headerRetry : err?.metadata?.retry_after_seconds;
+  const usage = p.extractUsage ? p.extractUsage(obj) : numericUsage(obj.usage);
+  return {
+    provider: p.name,
+    status: res.status,
+    ok: res.status >= 200 && res.status < 300,
+    ...err?.code ?? firstErr?.code ? { code: err?.code ?? firstErr?.code } : {},
+    ...message ? { message } : {},
+    ...Number.isFinite(retry) ? { retryAfterSeconds: retry } : {},
+    ...usage ? { usage } : {}
+  };
+}
+var PROBERS = {
+  zai: (key) => probeChat({ name: "zai", url: "https://api.z.ai/api/paas/v4/chat/completions", body: chatBody("glm-4.5-air") }, key),
+  nvidia: (key) => probeChat({ name: "nvidia", url: "https://integrate.api.nvidia.com/v1/chat/completions", body: chatBody("meta/llama-3.2-11b-vision-instruct") }, key),
+  orcarouter: (key) => probeChat({ name: "orcarouter", url: "https://api.orcarouter.ai/v1/chat/completions", body: chatBody("orcarouter/free") }, key),
+  "cloudflare-workers-ai": (key, meta) => {
+    const account = String(meta?.accountId ?? "");
+    if (!account) {
+      return Promise.resolve({ provider: "cloudflare-workers-ai", status: 0, ok: false, message: "metadata.accountId is required" });
+    }
+    return probeChat({
+      name: "cloudflare-workers-ai",
+      url: `https://api.cloudflare.com/client/v4/accounts/${account}/ai/run/@cf/meta/llama-3.1-8b-fast-v2`,
+      body: JSON.stringify({ prompt: "ping", max_tokens: 4 }),
+      extractUsage: (parsed) => numericUsage(typeof parsed === "object" && parsed !== null && "result" in parsed ? parsed.result.usage : undefined)
+    }, key);
+  }
+};
+async function runProbes() {
+  const auth = await readAuth();
+  const results = await Promise.all(Object.entries(auth).sort().map(async ([name, entry]) => {
+    const prober = PROBERS[name];
+    const key = authSecret(entry);
+    if (!prober || !key)
+      return null;
+    return prober(key, entry.metadata);
+  }));
+  return results.filter((r) => r !== null);
+}
 // packages/core/src/snapshot.ts
 async function withConnectedProviders(rows) {
   let connected;
@@ -868,6 +1052,7 @@ var USAGE = `opencode-usage \u2014 usage tracking and model ranking for opencode
 USAGE
   opencode-usage usage  [--since 7d] [--by provider|model|day|project|agent] [--today] [--pct] [--json]
   opencode-usage providers [--no-net] [--json]
+  opencode-usage probe [--json]
   opencode-usage top [--limit 20] [--json]
 
 OPTIONS
@@ -957,6 +1142,18 @@ async function cmdProviders(flags, json) {
     db.close();
   }
 }
+async function cmdProbe(flags, json) {
+  const results = await runProbes();
+  if (json) {
+    console.log(JSON.stringify({ probes: results }, null, 2));
+    return;
+  }
+  const auth = await readAuth();
+  const withoutProbe = Object.entries(auth).filter(([, e]) => e.key || e.access).map(([p]) => p).filter((p) => !["cloudflare-workers-ai", "nvidia", "orcarouter", "zai"].includes(p)).sort();
+  console.log(probeTable(results));
+  if (withoutProbe.length > 0)
+    console.log(`no free-model probe: ${withoutProbe.join(", ")}`);
+}
 var TOP_HEADERS = ["MODEL", "NAME", "IQ", "CODING", "$/M", "IQ/$"];
 var TOP_WIDTHS = [28, 26, 5, 7, 8, 6];
 function topLine(cells) {
@@ -994,6 +1191,8 @@ async function run(argv) {
     await cmdUsage(flags, json);
   else if (cmd === "providers")
     await cmdProviders(flags, json);
+  else if (cmd === "probe")
+    await cmdProbe(flags, json);
   else if (cmd === "top")
     await cmdTop(flags, json);
   else

package/dist/tui.js

@@ -274,12 +274,39 @@ function parseZai(data) {
     raw: data
   };
 }
+function parseZaiWallet(data) {
+  const failed = data.success === false || data.code !== undefined && data.code !== 0 && data.code !== 200;
+  if (failed) {
+    return { provider: "zai", ok: false, detail: data.msg ?? "wallet endpoint error", windows: [] };
+  }
+  const balance = data.data?.availableBalance ?? data.data?.rechargeAmount ?? 0;
+  const spend = data.data?.totalSpendAmount ?? 0;
+  const windows = [
+    { label: "balance", percentUsed: 0, detail: `$${balance.toFixed(2)}` },
+    { label: "spend", percentUsed: 0, detail: `$${spend.toFixed(2)}` }
+  ];
+  const today = data.data?.todaySpendAmount;
+  if (typeof today === "number") {
+    windows.push({ label: "today", percentUsed: 0, detail: `$${today.toFixed(2)}` });
+  }
+  return {
+    provider: "zai",
+    ok: true,
+    detail: `balance $${balance.toFixed(2)}`,
+    windows,
+    raw: data
+  };
+}
+var QUOTA_URL = "https://api.z.ai/api/monitor/usage/quota/limit";
+var WALLET_URL = "https://api.z.ai/api/biz/account/query-customer-account-report";
 async function fetchZai(key) {
   try {
-    const data = await getJson("https://api.z.ai/api/monitor/usage/quota/limit", {
-      Authorization: `Bearer ${key}`,
-      Accept: "application/json"
-    });
+    const auth = { Authorization: `Bearer ${key}`, Accept: "application/json" };
+    const data = await getJson(QUOTA_URL, auth);
+    if (data.code !== undefined && data.code !== 0) {
+      const wallet = await getJson(WALLET_URL, auth);
+      return parseZaiWallet(wallet);
+    }
     return parseZai(data);
   } catch (err) {
     return { provider: "zai", ok: false, detail: err.message, windows: [] };
@@ -328,11 +355,13 @@ function parseOrcaRouter(usage, subscription) {
   const cap = hard > 0 && hard < NO_CAP ? hard : soft > 0 && soft < NO_CAP ? soft : 0;
   const capped = cap > 0;
   const pct = capped ? spent / cap * 100 : 0;
+  const hint = capped ? "" : " (no cap set)";
+  const detail = `$${spent.toFixed(2)} spent${hint}`;
   return {
     provider: "orcarouter",
     ok: true,
-    detail: `$${spent.toFixed(2)} spent`,
-    windows: [{ label: "spend", percentUsed: pct, detail: `$${spent.toFixed(2)}` }],
+    detail,
+    windows: [{ label: "spend", percentUsed: pct, detail: `$${spent.toFixed(2)}${hint}` }],
     budget: capped ? { percentUsed: pct, label: `$${cap}` } : undefined,
     raw: { usage, subscription }
   };
@@ -348,6 +377,56 @@ async function fetchOrcaRouter(key) {
   }
 }
 
+// packages/core/src/quota/snowflake.ts
+var SQL_API = (account) => `https://${account}.snowflakecomputing.com/api/v2/statements`;
+var STATEMENT = `SELECT
+  SUM(CASE WHEN START_TIME >= DATE_TRUNC('day', CURRENT_TIMESTAMP())
+      AND START_TIME < DATEADD('day', 1, DATE_TRUNC('day', CURRENT_TIMESTAMP()))
+      THEN TOKENS ELSE 0 END) AS today_tokens,
+  SUM(TOKENS) AS month_tokens
+FROM SNOWFLAKE.ACCOUNT_USAGE.CORTEX_REST_API_USAGE_HISTORY
+WHERE START_TIME >= DATEADD('day', -30, CURRENT_TIMESTAMP())`;
+var fmtTokens = (n) => n.toLocaleString("en-US");
+function parseSnowflake(row) {
+  const today = Number(row?.[0] ?? 0);
+  const month = Number(row?.[1] ?? 0);
+  return {
+    provider: "snowflake-cortex",
+    ok: true,
+    detail: `${fmtTokens(month)} tokens/30d`,
+    windows: [
+      { label: "tokens/mo", percentUsed: 0, detail: `${fmtTokens(month)}` },
+      { label: "today", percentUsed: 0, detail: `${fmtTokens(today)}` }
+    ],
+    raw: { row }
+  };
+}
+async function fetchSnowflake(key, metadata) {
+  const account = String(metadata?.account ?? "").toUpperCase();
+  if (!account) {
+    return { provider: "snowflake-cortex", ok: false, detail: "metadata.account is required", windows: [] };
+  }
+  try {
+    const res = await fetch(SQL_API(account), {
+      method: "POST",
+      headers: {
+        Authorization: `Bearer ${key}`,
+        "Content-Type": "application/json",
+        Accept: "application/json"
+      },
+      body: JSON.stringify({ statement: STATEMENT, timeout: 30 })
+    });
+    const body = await res.json().catch(() => null);
+    if (!res.ok) {
+      return { provider: "snowflake-cortex", ok: false, detail: body?.message ?? `HTTP ${res.status}`, windows: [] };
+    }
+    const rows = Array.isArray(body?.data) ? body.data : [];
+    return parseSnowflake(rows[0]);
+  } catch (err) {
+    return { provider: "snowflake-cortex", ok: false, detail: err.message, windows: [] };
+  }
+}
+
 // packages/core/src/providers.ts
 var TTL_MS = 15 * 60 * 1000;
 async function readCache() {
@@ -371,7 +450,8 @@ var FETCHERS = {
   "github-copilot": fetchCopilot,
   zai: fetchZai,
   amd: fetchAmd,
-  orcarouter: fetchOrcaRouter
+  orcarouter: fetchOrcaRouter,
+  "snowflake-cortex": fetchSnowflake
 };
 async function providerStatuses(opts) {
   const auth = await readAuth();
@@ -392,7 +472,7 @@ async function providerStatuses(opts) {
       const secret = authSecret(auth[p]);
       if (!fetcher || !secret)
         return null;
-      return fetcher(secret);
+      return fetcher(secret, auth[p].metadata);
     }));
     for (const q of live) {
       if (q)
@@ -500,7 +580,7 @@ var PROVIDER_LIMITS = [
     limit: 100,
     tier: "paid",
     source: "https://docs.snowflake.com/en/sql-reference/account-usage/cortex_ai_functions_usage_history",
-    note: "Cortex AI Functions bill per token (credits per million tokens, Service Consumption Table). Usage lives in ACCOUNT_USAGE views, which the API key alone cannot read."
+    note: "Billed per token (Service Consumption Table). Live tokens come from ACCOUNT_USAGE.CORTEX_REST_API_USAGE_HISTORY through the SQL API with the stored JWT; the percentage needs a snowflake-cortex line in budgets.json."
   },
   {
     provider: "zai",
@@ -655,7 +735,7 @@ function columnsFor(inner) {
   const budget = Math.max(BUDGET_MIN, inner - used);
   return { cols: [...base, { title: "BUDGET", width: budget, align: "left" }], bar: wide ? WIDE_BAR : COMPACT_BAR };
 }
-function fmtTokens(n) {
+function fmtTokens2(n) {
   if (n >= 1e6)
     return `${(n / 1e6).toFixed(1)}M`;
   if (n >= 1000)
@@ -821,8 +901,8 @@ function leadCells(cols, r) {
   return row(cols, [
     r.provider,
     String(r.messages),
-    fmtTokens(r.tokensInput),
-    fmtTokens(r.tokensOutput),
+    fmtTokens2(r.tokensInput),
+    fmtTokens2(r.tokensOutput),
     `$${r.cost.toFixed(2)}`
   ]).trimEnd().padEnd(upToBudget) + " ";
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@1930dev/opencode-usage",
-  "version": "0.3.3",
+  "version": "0.3.5",
   "description": "Usage tracking, budget percentages and model ranking for opencode — CLI plus a /usage TUI plugin",
   "type": "module",
   "bin": {