@aeon-ai-pay/aigateway 0.1.5 → 0.2.1

package/CHANGELOG.md

@@ -7,6 +7,28 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
 
 ## [Unreleased]
 
+## [0.1.6] — 2026-05-19
+
+### Changed
+- **`update-check.mjs` is now synchronous / foreground.** A newer published
+  version triggers `npm install -g` + `scripts/postinstall.mjs` *inline*,
+  with `npm`'s own progress streamed to stderr. After the upgrade the CLI
+  emits an envelope with `error.code === "UPDATE_APPLIED"` (and `error.from`
+  / `error.to`) and exits with code 2 **without running the original
+  command**. The caller / agent reruns the command on the new CLI + new
+  SKILL.md.
+- Replaces the previous detached-background upgrade, which could leave the
+  global package half-replaced (`bin/cli.mjs` already updated while
+  `src/commands/*` still old → `ERR_MODULE_NOT_FOUND` on the very next
+  invocation) and could fail silently (e.g. `ENOTEMPTY` on Windows / nvm).
+
+### Added
+- New error code `UPDATE_APPLIED` (exit 2) emitted by `update-check.mjs`
+  after a successful synchronous upgrade.
+- SKILL.md Hard Rules: new clause telling the agent how to react when it
+  sees `error.code === "UPDATE_APPLIED"` (briefly inform the user of the
+  version transition, then rerun the same command).
+
 ## [0.1.5] — 2026-05-19
 
 ### Changed

package/bin/cli.mjs

@@ -104,45 +104,33 @@ program
     return withdraw(opts);
   });
 
-program
-  .command("create-card")
-  .description("Create a one-time virtual card by paying with USDT on BSC via x402")
-  .requiredOption("--amount <usd>", "Card amount in USD ($0.6 ~ $800)")
-  .option("--app-id <id>", "Merchant app ID", DEFAULT_APP_ID)
-  .option("--topup-amount <usdt>", "USDT top-up amount when balance is insufficient (≥5)")
-  .option("--private-key <key>", "Override EVM private key")
-  .option("--poll", "Auto-poll status after creation", false)
-  .option("--dry-run", "Run all preflight checks but do not sign/transact", false)
-  .action(async (opts) => {
-    const { createCard } = await import("../src/commands/create-card.mjs");
-    return createCard(opts);
-  });
-
-program
-  .command("create-image")
-  .description("Generate an AI image via Skill Boss, paying with USDT on BSC via x402")
-  .requiredOption("--prompt <text>", "Image prompt text")
+// ─── x402 unified paid-call entry ──────────────────────────────────────────
+//   The only paid-call surface: `sb invoke --model <id> --inputs <json|@file>`.
+const sb = program
+  .command("sb")
+  .description("AI tools — invoke any of ~200+ endpoints via x402 (USDT on BSC)");
+
+sb
+  .command("invoke")
+  .description("Invoke any AI tool model (run `aigateway sb tools` first to populate the local catalog)")
+  .requiredOption("--model <id>", "Model id, e.g. replicate/black-forest-labs/flux-schnell")
+  .requiredOption("--inputs <json>", "Inputs as JSON literal or @path/to/file.json")
   .option("--app-id <id>", "Merchant app ID", DEFAULT_APP_ID)
-  .option("--aspect-ratio <ratio>", "Image aspect ratio", "16:9")
-  .option("--output-format <fmt>", "Image format (png/jpeg/webp)", "png")
-  .option("--model <id>", "Model identifier")
-  .option("--output <dir>", "Output directory (default ~/aigateway-images)")
+  .option("--output <dir>", "Output directory for binary downloads (image/video/audio)")
+  .option("--raw", "Skip auto-download and emit the upstream response as-is", false)
   .option("--topup-amount <usdt>", "USDT top-up amount when balance is insufficient (≥5)")
   .option("--private-key <key>", "Override EVM private key")
   .action(async (opts) => {
-    const { createImage } = await import("../src/commands/create-image.mjs");
-    return createImage(opts);
+    const { sbInvokeCommand } = await import("../src/commands/sb-invoke.mjs");
+    return sbInvokeCommand(opts);
   });
 
-program
-  .command("create-card-status")
-  .description("Query the status of a virtual card created via create-card")
-  .requiredOption("--order-no <orderNo>", "Order number returned by create-card")
-  .option("--app-id <id>", "Merchant app ID", DEFAULT_APP_ID)
-  .option("--poll", "Poll until terminal status", false)
+sb
+  .command("tools")
+  .description("Fetch and display the AI tool catalog from the server (no caching, always live)")
   .action(async (opts) => {
-    const { status } = await import("../src/commands/create-card-status.mjs");
-    return status(opts);
+    const { sbTools } = await import("../src/commands/sb-tools.mjs");
+    return sbTools(opts);
   });
 
 program

package/docs/exit-codes.md

@@ -38,8 +38,9 @@ The full set of `error.code` values, grouped by exit code, is defined in [`src/e
 | ---- | ---- |
 | `PAYMENT_TIMEOUT` | WalletConnect / signature request timed out (5 minutes). |
 | `WC_SESSION_EXPIRED` | WalletConnect session dropped mid-flow. |
-| `POLL_TIMEOUT` | `status --poll` exhausted attempts. Card may still be provisioning. |
+| `POLL_TIMEOUT` | `create-card-status --poll` exhausted attempts. Card may still be provisioning. |
 | `TX_TIMEOUT` | On-chain receipt wait exceeded 60 s. |
+| `UPDATE_APPLIED` | The CLI just upgraded itself synchronously to a newer version. The previous command was not executed — the caller (or the agent) must rerun it on the new version. Envelope carries `error.from` / `error.to` showing the version transition. |
 
 ### Exit 3 — Service / Network
 

package/docs/release-process.md

@@ -59,15 +59,17 @@ We follow [SemVer](https://semver.org/):
 
 ## What happens for users after publish
 
-`src/update-check.mjs` runs on every CLI invocation. Within a few seconds of the next `aigateway <cmd>` call by any user, they will:
+`src/update-check.mjs` runs at the start of every CLI invocation. On the **next** `aigateway <cmd>` call by any user with a stale version, they will:
 
-1. See `[update] @aeon-ai-pay/aigateway 0.1.0 → 0.1.1, upgrading in background...` on stderr.
-2. The current command still completes normally.
-3. A detached child process runs `npm install -g @aeon-ai-pay/aigateway@0.1.1` + `node scripts/postinstall.mjs`, which calls `npx skills add ... -g -y --copy`.
-4. The new `SKILL.md` is re-installed into every detected IDE skill folder (Cursor / Claude Code / Codex / Windsurf / Cline / …).
-5. Next invocation uses the new CLI and the new agent reads the new skill.
+1. See `[update] @aeon-ai-pay/aigateway 0.1.0 → 0.1.1, upgrading (foreground)...` on stderr.
+2. **`npm install -g @aeon-ai-pay/aigateway@<latest>` runs synchronously**, with `npm`'s own progress streamed to stderr.
+3. **`scripts/postinstall.mjs` runs synchronously**, calling `npx skills add ... -g -y --copy` to refresh every detected IDE skill folder (Cursor / Claude Code / Codex / Windsurf / Cline / …).
+4. The CLI emits an envelope with `error.code === "UPDATE_APPLIED"` and **exits with code 2** without executing the original command.
+5. The caller (or the agent — `error.code` is stable, agents should match on it) **reruns the original command**, which now executes on the new CLI + new SKILL.md.
 
-Upgrade results land in `~/.aigateway/update.log`.
+Why foreground instead of detached background: a backgrounded `npm install -g` mid-command can leave the global package in a half-replaced state (`bin/cli.mjs` already updated while `src/commands/*` still old, or vice versa), causing `ERR_MODULE_NOT_FOUND` on the very next invocation. Synchronous upgrade keeps the package consistent.
+
+If the upgrade itself fails (network, permissions, `ENOTEMPTY` on Windows / nvm), the failure is logged to stderr and the command continues on the current version — the user is never silently left on a half-installed package.
 
 ## Rolling back
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aeon-ai-pay/aigateway",
-  "version": "0.1.5",
+  "version": "0.2.1",
   "description": "AI Agents discover, invoke, and settle paid LLMs, APIs, and Skills — starting with Skill Boss. No manual key setup. No prepayment. Pay-per-call via x402 or Agent Card.",
   "type": "module",
   "bin": {