@eigenpal/automations 0.20.73 → 0.20.74

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +1 -1
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -148,7 +148,7 @@ agent identity; the display name can change without changing that identity.
148
148
  - `inputSchema`: optional object schema for run input. Prefer Zod 4. Raw JSON Schema remains accepted for existing contracts.
149
149
  - `outputSchema`: optional object schema for the validated final answer. Prefer `z.strictObject` to reject unexpected fields.
150
150
  - `tools`: optional map of names to `createTool(...)` definitions. Names become model-facing tool names. At most 16; built-in names such as `read`, `write`, `bash` and `finish` are reserved.
151
- - `guards`: optional array of positive limits. `maxTurns(n)` bounds model requests, including retries; `maxTimeMs(n)` bounds active execution time, including preparation but excluding time parked for human review; `maxCredits(n)` bounds model credits only. One credit is one cent. `maxTimeMs(n)` bounds active sandbox execution time; sandbox compute is billed separately. Repeated limits use the strictest value and continue across human-review resumes. Platform ceilings remain 128 turns, 20 active minutes and 10,000 model credits, with any tighter deployment settings still applying. Conservative request admission can stop before the exact limit when the remaining balance cannot cover the next bounded request.
151
+ - `guards`: optional array of positive limits. `maxTurns(n)` bounds model requests, including retries; `maxTimeMs(n)` bounds active execution time, including preparation but excluding time parked for human review; `maxCredits(n)` bounds model credits only. One credit is one cent. `maxTimeMs(n)` bounds active sandbox execution time; sandbox compute is billed separately. Repeated limits use the strictest value and continue across human-review resumes. Platform ceilings remain 128 turns, 20 active minutes and 10,000 model credits, with any tighter deployment settings still applying. Credit checks use actual usage, without reservations or estimates. A request may cross `maxCredits`; the next model request stops. Account balances similarly allow work while positive, charge actual usage even if it overdraws, and stop subsequent paid work once exhausted. Pay-as-you-go settings still apply.
152
152
  - `triggers`: optional `{ api, email: { enabled, aliases } }`. API defaults to enabled; absent email is disabled. Aliases may be addresses or objects with `address`, `allowlist`, `replyConfig`, `requireSenderAuth` and `enabled`. An omitted allowlist preserves platform ownership; `[]` denies all senders. Existing email routing rules continue to apply.
153
153
  - `resources`: optional map of imported knowledge, rules and templates. Keys are local aliases; TypeScript checks each imported definition. See shared resources below. Evaluators are registered through the chain, never a `createAgent` property.
154
154
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@eigenpal/automations",
3
- "version": "0.20.73",
3
+ "version": "0.20.74",
4
4
  "description": "TypeScript definitions for EigenPal agents.",
5
5
  "type": "module",
6
6
  "license": "Apache-2.0",