@automatalabs/workflows 0.37.0 → 0.38.0

package/README.md

@@ -6,7 +6,7 @@ The programmatic **SDK** for AgentPrism — run dynamic, multi-agent **workflow
 
 You author a small JavaScript **script** (a string), the engine runs it in a deterministic,
 journaled, resumable realm, and every `agent()` call inside it is fanned out to a pooled ACP
-backend — **Claude** (`claude-agent-acp`), **Codex** (`codex-acp`), **OpenCode** (`opencode acp`),
+backend — **Claude** (`claude-agent-acp`), **Codex** (`codex-acp`), **OpenCode** (`opencode acp`), **pi** (`pi-acp`),
 or a registered custom ACP agent — driving the actual subprocess to completion.
 
 This package is the **canonical SDK** that the stdio MCP server
@@ -28,7 +28,7 @@ through this SDK's workflow/runner APIs rather than MCP server schemas.
 pnpm add @automatalabs/workflows
 ```
 
-> The Claude and Codex ACP servers ship as transitive dependencies and are spawned on demand.
+> The Claude, Codex, and pi ACP servers ship as transitive dependencies and are spawned on demand.
 > OpenCode is host-resolved: install `opencode-ai` or make `opencode` available on `PATH` before
 > routing a call to it.
 
@@ -43,6 +43,7 @@ pnpm add @automatalabs/workflows
     environment.
   - **Codex** — a logged-in Codex install (`~/.codex`).
   - **OpenCode** — credentials configured for the provider OpenCode will use.
+  - **pi** — a selected provider API key or credentials in `~/.pi/agent/auth.json`.
 
 You only need auth for the backend(s) your scripts actually route to. The default backend is
 Claude (override with `AGENTPRISM_DEFAULT_BACKEND`; see [Backend selection](#backend-selection)).
@@ -296,8 +297,15 @@ checkpoint unless the author opts in. `WorkflowManagerOptions` lets you set a de
 `loadSavedWorkflow` resolver (enables nested `workflow('name')`), a custom `persistence`
 implementation, and per-agent timeout/retry defaults.
 
-Every terminal result may also carry `fallbacks` (a compatibility audit surface for non-resolution
-subsystems or third-party runners) and `checkpointsTaken` (one
+Usage-limit and auth resumes are continuation-aware by default on both `resumeFromRunId` and
+same-ID `resume()` / `resumeInBackground()`: when the interrupted root call's index, identity hash,
+full execution-input fingerprint, backend identity, cwd, and reopen support still agree, the manager
+reattaches its recorded ACP session and continues the unfinished turn. Worktree calls, changed or
+legacy inputs, missing cwd, and every uncertain/rejected reopen run fresh. This is internal manager →
+engine plumbing; the public SDK accepts no continuation option.
+
+Every terminal result may also carry `fallbacks` (including `kind: "continuation"` notices that
+record a reattached `resume`/`load` method or an exact skip reason) and `checkpointsTaken` (one
 resolved checkpoint per call with the journaled decision and `live` / `headless-default` /
 `journal-replay` / `injected` source). They are absent when empty, persist for cold reads, never
 enter replay hashes, and are not part of `WorkflowRunStatus` inspection.
@@ -389,7 +397,9 @@ runner names.
 Every live ACP-backed `agent()` call records a non-secret re-attach handle in
 `run.agentSessions`. Set `agent(..., { keepSession: true })` to skip release-time `session/close`,
 then use the runner's `loadSession()` or `resumeSession()` host API with that record. Session
-handles are additive and are also preserved in journals; they do not change the deterministic
+handles are additive and are also preserved in journals; pause-class failures skip `session/close`
+automatically, and eligible managed resumes consume the recorded handle internally. A continued
+result carries a diagnostic journal marker, but neither the handle nor marker changes deterministic
 resume identity.
 
 ### d) Bring your own backend — implement the `AgentRunner` seam
@@ -658,7 +668,7 @@ npx @automatalabs/workflows config codex opencode   # only the named harnesses
 npx @automatalabs/workflows config claude --json    # machine-readable report
 ```
 
-Harness names are the routing names: built-in `claude` / `codex` / `opencode` plus any custom
+Harness names are the routing names: built-in `claude` / `codex` / `opencode` / `pi` plus any custom
 backend registered via `AGENTPRISM_BACKENDS` (registered customs also join the no-argument
 default set). Each harness opens one session without a prompt — zero tokens — and reports its
 advertised config-option catalog verbatim: model ids (including bracket variants), effort
@@ -687,7 +697,7 @@ formatHarnessConfigReport(report); // the CLI's human table
 
 Pass a JSON Schema to `agent({ schema })` (in a script) or `runner.run(prompt, { schema })` (direct)
 and the result is a **validated object** instead of text. The backend constrains output natively
-(Claude `outputFormat`; Codex strict `outputSchema`; prompt/tool-assisted JSON for OpenCode and
+(Claude `outputFormat`; Codex strict `outputSchema`; pi native `_meta.outputSchema`; prompt/tool-assisted JSON for OpenCode and
 custom agents), then the value is coerced and validated client-side (typebox `Convert` → `Check`); on a miss the runner re-prompts a bounded number of
 times before failing with a non-recoverable `SCHEMA_NONCOMPLIANCE`.
 
@@ -725,7 +735,7 @@ toStrictJsonSchema(schema);  // OpenAI-strict-normalized (Codex outputSchema)
 
 The backend for each agent is chosen from its effective `model` (preferred) or `tier` string. Split
 the string on its first `/`. If the first segment, ASCII-case-insensitively, is `claude`, `codex`,
-`opencode`, or a registered custom backend name, it selects that harness and is stripped exactly
+`opencode`, `pi`, or a registered custom backend name, it selects that harness and is stripped exactly
 once; a custom registration wins on a built-in-name collision. A registered harness name alone is
 backend-only and preserves that harness's configured default model. Any other first segment sends
 the entire authored string unchanged to `AGENTPRISM_DEFAULT_BACKEND` (default `claude`). Omitting
@@ -737,6 +747,7 @@ import { selectBackend } from "@automatalabs/workflows";
 selectBackend({ model: "claude/opus[1m]" }).id;      // "claude"
 selectBackend({ model: "codex/gpt-5.6-sol" }).id;    // "codex"
 selectBackend({ model: "opencode/zai/glm-5.2" }).id; // "opencode"
+selectBackend({ model: "pi/openrouter/vendor/model" }).id; // "pi"
 ```
 
 When an id remains after routing, it becomes the exact ACP `session/set_config_option` value
@@ -779,7 +790,7 @@ createAcpRunner,              // () => AcpAgentRunner (the default AgentRunner;
 AcpAgentRunner,               // class — implements AgentRunner over ACP
 InteractiveSession,           // held-open multi-turn ACP session returned by openSession()
 selectBackend,                // pick a built-in/custom backend from a model/tier spec
-ClaudeBackend, CodexBackend, CustomAcpBackend,
+ClaudeBackend, CodexBackend, OpenCodeBackend, PiBackend, CustomAcpBackend,
 resolveBackendRegistry, BACKENDS_ENV,
 AGENT_METHODS, CLIENT_METHODS, ACP_AUTH_REQUIRED_ERROR_CODE,
 clientCapabilitiesFor, adaptPromptContent,

package/dist/cli.js

@@ -75,7 +75,7 @@ and every other session option — by opening one no-prompt session per harness.
 tokens. Run this BEFORE authoring a workflow so \`model\` / \`configOptions\` values come
 from the live catalog instead of guesswork (or a throwaway probe workflow).
 
-Harnesses: the built-in names (claude, codex, opencode) and any custom backend
+Harnesses: the built-in names (claude, codex, opencode, pi) and any custom backend
 registered via the AGENTPRISM_BACKENDS env var. Default: all of them. A harness that
 cannot spawn or authenticate reports \`probed: false\` with the reason and never blocks
 the others.

package/dist/config.d.ts

@@ -1,9 +1,9 @@
 import type { CustomBackendConfig } from "@automatalabs/acp-agents";
 import type { ValidateHarnessOptions } from "./validate.js";
 export interface ProbeHarnessConfigOptions {
-    /** Harness names to probe (built-in `claude` / `codex` / `opencode` or a registered
+    /** Harness names to probe (built-in `claude` / `codex` / `opencode` / `pi` or a registered
      *  custom name; any model spec routes like an agent() call's). Default: every routable
-     *  harness — the three built-ins plus each registered custom backend. */
+     *  harness — the four built-ins plus each registered custom backend. */
     harnesses?: string[];
     /** Programmatic custom-backend registry, merged over the AGENTPRISM_BACKENDS env var
      *  exactly like `createAcpRunner({ backends })`. */

package/dist/config.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../src/config.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,0BAA0B,CAAC;AAGpE,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,eAAe,CAAC;AAE5D,MAAM,WAAW,yBAAyB;IACxC;;6EAEyE;IACzE,SAAS,CAAC,EAAE,MAAM,EAAE,CAAC;IACrB;wDACoD;IACpD,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,mBAAmB,CAAC,CAAC;IAC/C;yEACqE;IACrE,GAAG,CAAC,EAAE,MAAM,CAAC;IACb;+CAC2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,mBAAmB;IAClC,6DAA6D;IAC7D,EAAE,EAAE,OAAO,CAAC;IACZ,qDAAqD;IACrD,QAAQ,EAAE,CAAC,GAAG,CAAC,CAAC;IAChB,2FAA2F;IAC3F,cAAc,EAAE,sBAAsB,EAAE,CAAC;CAC1C;AAKD;;;;;GAKG;AACH,wBAAsB,kBAAkB,CACtC,OAAO,GAAE,yBAA8B,GACtC,OAAO,CAAC,mBAAmB,CAAC,CAoC9B;AAED,+FAA+F;AAC/F,wBAAgB,yBAAyB,CAAC,MAAM,EAAE,mBAAmB,GAAG,MAAM,CAU7E"}
+{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../src/config.ts"],"names":[],"mappings":"AAUA,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,0BAA0B,CAAC;AAGpE,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,eAAe,CAAC;AAE5D,MAAM,WAAW,yBAAyB;IACxC;;4EAEwE;IACxE,SAAS,CAAC,EAAE,MAAM,EAAE,CAAC;IACrB;wDACoD;IACpD,QAAQ,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,mBAAmB,CAAC,CAAC;IAC/C;yEACqE;IACrE,GAAG,CAAC,EAAE,MAAM,CAAC;IACb;+CAC2C;IAC3C,SAAS,CAAC,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,mBAAmB;IAClC,6DAA6D;IAC7D,EAAE,EAAE,OAAO,CAAC;IACZ,qDAAqD;IACrD,QAAQ,EAAE,CAAC,GAAG,CAAC,CAAC;IAChB,2FAA2F;IAC3F,cAAc,EAAE,sBAAsB,EAAE,CAAC;CAC1C;AAKD;;;;;GAKG;AACH,wBAAsB,kBAAkB,CACtC,OAAO,GAAE,yBAA8B,GACtC,OAAO,CAAC,mBAAmB,CAAC,CAoC9B;AAED,+FAA+F;AAC/F,wBAAgB,yBAAyB,CAAC,MAAM,EAAE,mBAAmB,GAAG,MAAM,CAU7E"}

package/dist/config.js

@@ -9,7 +9,7 @@ import { redactText } from "@automatalabs/workflow-engine";
 import { resolveBackendRegistry } from "@automatalabs/acp-agents";
 import { createValidateProbeRunner } from "./validate-internal.js";
 import { renderHarnessOptionLines } from "./validate.js";
-const BUILTIN_HARNESSES = ["claude", "codex", "opencode"];
+const BUILTIN_HARNESSES = ["claude", "codex", "opencode", "pi"];
 const DEFAULT_PROBE_TIMEOUT_MS = 60_000;
 /**
  * Probe each requested harness's advertised config-option catalog. A per-harness

package/dist/validate.d.ts

@@ -67,7 +67,7 @@ export interface ValidatedAgentCall {
     mode?: string;
     /** The verbatim session config options authored for this call. */
     configOptions?: Record<string, string | boolean>;
-    /** Which concrete backend the spec routes to: "claude" | "codex" | "opencode" | a custom
+    /** Which concrete backend the spec routes to: "claude" | "codex" | "opencode" | "pi" | a custom
      *  backend name (suffixed " (script-declared)" when it comes from meta.backends). */
     backend: string;
     /** True when the call requested structured output. */

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@automatalabs/workflows",
-  "version": "0.37.0",
+  "version": "0.38.0",
   "license": "Apache-2.0",
   "engines": {
     "node": ">=22"
@@ -31,9 +31,9 @@
   },
   "dependencies": {
     "typebox": "1.3.2",
-    "@automatalabs/shared-types": "0.23.0",
-    "@automatalabs/acp-agents": "0.28.1",
-    "@automatalabs/workflow-engine": "0.25.0"
+    "@automatalabs/shared-types": "0.24.0",
+    "@automatalabs/workflow-engine": "0.26.0",
+    "@automatalabs/acp-agents": "0.30.0"
   },
   "scripts": {
     "build": "tsc -b",