Fix invoke owner protocol inference

[SandroMaglione]

Summary

• contextually type Machine.invoke sources and lifecycle handlers with the owning machine public and parent protocols inside .handle(...)
• retain Effect output, failure, service, state, and child protocol inference while rejecting invalid parent or self sends
• teach direct event protocol construction in API docs, the agent guide, README, and examples

Changeset

• Added or updated for a library or package-metadata change
• Not required because this PR does not change src/ or package.json

Patch changeset: compatible inference and documentation fix.

Validation

• pnpm check
• Relevant example checks, when examples changed
• Automated type-performance measurement passed or was not required
• Automated runtime- and memory-performance measurement passed or was not required

Affected examples checked: playground, pokemon, and platformer. Local pnpm perf:types and pnpm perf:runtime smoke tests passed; CI base-versus-PR comparisons remain authoritative.

[github-actions]

Type performance

Measured with TypeScript 6.0.3 and skipLibCheck=true.

Scenario	Base	PR	Difference
Effect only	55	55	0 (0.0%)
Import effect-machine	55	55	0 (0.0%)
Machine.defineStates (3 states)	3,039	3,039	0 (0.0%)
Machine.make (3 states, 2 events)	10,439	10,439	0 (0.0%)
machine.handle (3 states, 2 transitions)	53,206	53,206	0 (0.0%)
Machine.transition (10 conditional cases)	116,110	116,110	0 (0.0%)
Machine.invoke (state-dependent Effect)	57,187	57,606	+419 (+0.7%)
machine.handle (depth 24)	188,289	188,289	0 (0.0%)
machine.handle (wide depth 16)	219,355	219,355	0 (0.0%)
machine.handle (parallel/history/choice)	132,085	132,085	0 (0.0%)
machine.handle (4 successive calls)	137,762	137,762	0 (0.0%)
machine exact input/output/error/services	114,832	114,832	0 (0.0%)
execution adapter readiness	124,664	124,748	+84 (+0.1%)

Marginal instantiations are measured against the matching setup without that API call:

Scenario	Base	PR	Difference
Import effect-machine	0	0	0
Machine.defineStates (3 states)	2,984	2,984	0 (0.0%)
Machine.make (3 states, 2 events)	7,392	7,392	0 (0.0%)
machine.handle (3 states, 2 transitions)	42,767	42,767	0 (0.0%)
Machine.transition (10 conditional cases)	106,181	106,181	0 (0.0%)
Machine.invoke (state-dependent Effect)	48,341	48,760	+419 (+0.9%)
machine.handle (depth 24)	168,599	168,599	0 (0.0%)
machine.handle (wide depth 16)	200,664	200,664	0 (0.0%)
machine.handle (parallel/history/choice)	112,409	112,409	0 (0.0%)
machine.handle (4 successive calls)	122,216	122,216	0 (0.0%)
machine exact input/output/error/services	100,741	100,741	0 (0.0%)
execution adapter readiness	62,727	62,811	+84 (+0.1%)

Check times (informational)

Scenario	Base	PR
Effect only	0.03s	0.04s
Import effect-machine	0.04s	0.03s
Machine.defineStates (3 states)	0.11s	0.11s
Machine.make (3 states, 2 events)	0.18s	0.18s
machine.handle (3 states, 2 transitions)	0.36s	0.34s
Machine.transition (10 conditional cases)	0.56s	0.58s
Machine.invoke (state-dependent Effect)	0.40s	0.42s
machine.handle (depth 24)	0.78s	0.80s
machine.handle (wide depth 16)	0.80s	0.81s
machine.handle (parallel/history/choice)	0.65s	0.66s
machine.handle (4 successive calls)	0.66s	0.64s
machine exact input/output/error/services	0.59s	0.61s
execution adapter readiness	0.58s	0.57s

Type instantiations are the comparison metric. Check time varies with runner load and is informational only.

[github-actions]

Runtime performance

Median of 5 independent benchmark processes on AMD EPYC 9V74 80-Core Processor with Node v24.19.0.

Pull request baseline

Scenario	Effect Machine
Plan counter transitions	145,659 transitions/s
Drain burst with terminal fence	428,964 increments/s
Drain burst with a change observer	402,293 increments/s
Lookup and send to one child	369,223 increments/s
Start and stop a machine	175,162 machines/s
Start and stop a parent with one child	53,769 families/s
Plan transitions through a compound state	126,738 transitions/s
Plan transitions through parallel regions	104,384 transitions/s
Drain burst through a compound state	353,965 events/s
Drain burst through two parallel regions	358,329 events/s
Drain a compound-state burst with a change observer	336,618 events/s

Process runtime reference points

Scenario	Effect Machine
Start and stop a raw generic process	34,731 processes/s
Start and stop a raw compiled process	84,260 processes/s

Memory profile	Effect Machine
Idle machine	1.8 KiB
Raw generic managed process	13.9 KiB
Raw compiled process	3.1 KiB
Two independent idle machines	3.4 KiB
Idle parent with one child	5.5 KiB
Parent with observed child registry	9.7 KiB
Parent with observed invoked child snapshots	6.1 KiB

Effect Machine change from base

Metric	Base	Base variability	PR	PR variability	Difference
Plan counter transitions	143,416 transitions/s	0.8% MAD	145,659 transitions/s	2.0% MAD	+1.6%
Drain burst with terminal fence	420,644 increments/s	1.4% MAD	428,964 increments/s	1.1% MAD	+2.0%
Drain burst with a change observer	392,940 increments/s	1.8% MAD	402,293 increments/s	1.6% MAD	+2.4%
Lookup and send to one child	359,208 increments/s	0.8% MAD	369,223 increments/s	0.6% MAD	+2.8%
Start and stop a machine	176,710 machines/s	0.6% MAD	175,162 machines/s	1.2% MAD	-0.9%
Start and stop a parent with one child	53,000 families/s	0.5% MAD	53,769 families/s	1.0% MAD	+1.5%
Plan transitions through a compound state	125,761 transitions/s	0.7% MAD	126,738 transitions/s	1.1% MAD	+0.8%
Plan transitions through parallel regions	102,659 transitions/s	1.1% MAD	104,384 transitions/s	1.4% MAD	+1.7%
Drain burst through a compound state	351,285 events/s	1.4% MAD	353,965 events/s	0.4% MAD	+0.8%
Drain burst through two parallel regions	357,944 events/s	2.7% MAD	358,329 events/s	0.4% MAD	+0.1%
Drain a compound-state burst with a change observer	333,756 events/s	0.8% MAD	336,618 events/s	0.5% MAD	+0.9%
Idle machine heap per unit	1.8 KiB	0.4% MAD	1.8 KiB	0.0% MAD	-0.2%
Raw generic managed process heap per unit	13.9 KiB	0.0% MAD	13.9 KiB	0.0% MAD	-0.0%
Raw compiled process heap per unit	3.1 KiB	0.2% MAD	3.1 KiB	0.1% MAD	+0.1%
Two independent idle machines heap per unit	3.4 KiB	0.0% MAD	3.4 KiB	0.0% MAD	+0.7%
Idle parent with one child heap per unit	5.5 KiB	0.0% MAD	5.5 KiB	0.0% MAD	+0.0%
Parent with observed child registry heap per unit	9.7 KiB	0.0% MAD	9.7 KiB	0.0% MAD	-0.0%
Parent with observed invoked child snapshots heap per unit	6.1 KiB	0.0% MAD	6.1 KiB	0.0% MAD	-0.0%

Process runtime reference change from base

Metric	Base	Base variability	PR	PR variability	Difference
Start and stop a raw generic process	34,826 processes/s	3.1% MAD	34,731 processes/s	0.6% MAD	-0.3%
Start and stop a raw compiled process	83,907 processes/s	1.2% MAD	84,260 processes/s	0.7% MAD	+0.4%

Regression guard

No large, noise-adjusted throughput or heap regressions detected.

Versions and interpretation

• Effect Machine: 0.14.1

Higher throughput is better; lower heap is better. Variability is the median absolute deviation across independent processes, relative to their median. Small differences on shared GitHub-hosted hardware remain informational; the required guard rejects only large changes beyond the measured noise allowance.