feat(host_api): Slice C.1 — Invocation payload vocabulary (#6168) by ilblackdragon · Pull Request #6223 · nearai/ironclaw

ilblackdragon · 2026-07-18T03:09:38Z

What

First sub-slice of Slice C — the capability-path DTO/dyn collapse from docs/reborn/2026-07-17-architecture-simplification-dto-dyn-local.md (§3, §5). Per the migration plan (§9), Slice C lands the kernel vocabulary in host_api first, ahead of any wiring. This PR adds the single data-plane payload that replaces the request side of the ~14-hop re-wrap (§1.1).

Stacked on #6222 (C3, deployment-mode ratchet).

Types added

ids.rs (via the crate's own uuid_id!/string_id! macros):
- ActivityId — the invocation's idempotency identity (§11.3); what §1.1's dead-future idempotency_key becomes, unified not deleted.
- ProductKind / RoutineId — validated string newtypes for the two non-loop origins (kept strings, not enums, while the product/routine sets are still evolving per §5.8; noted as future-enum candidates).
invocation.rs (new module):
- InvocationOrigin { LoopRun(RunId) | Product(ProductKind) | Automation(RoutineId) } — sealed-at-membrane origin (§5.2.1); snake_case wire-tagged; .kind() discriminant pinned to the tag for per-origin accounting views (§5.3.3).
- Invocation { activity_id, capability, input, scope, actor, origin, estimate } — the "one payload" (§3/§4.1) every layer will reference instead of re-declaring.

Design reconciliations (documented in the code)

The doc names LoopRun(TurnRunId), but host_api cannot depend on ironclaw_turns. Modeled as RunId — this crate's prompt-visible turn-run identity (= ExecutionContext::run_id); the alias relationship is noted in the doc comment.
Invocation.actor is a required UserId (vs CapabilityDispatchRequest.authenticated_actor_user_id: Option<UserId>) — the sealed-at-membrane intent.
mounts / resource_reservation are deliberately absent — they are outputs of authorize(), not request inputs, so they move into the sealed Authorized witness in a later slice.

Why additive is safe (§9)

Nothing is wired into the dispatch path. The five old request DTOs still exist and function; the doc is explicit that the type count rises before it falls (~14 → ~18 → ~11) while the new vocabulary and old shapes coexist. python3 scripts/check-type-duplicates.py does not flag Invocation against CapabilityDispatchRequest — it is a genuinely distinct state, not a mirror.

Testing

5 new crate-tier tests: origin serde round-trip + snake_case tags, .kind()↔tag agreement, id validation, ActivityId stable-carry (idempotency), and one-payload-per-origin construction.
Crate-tier only because the types aren't production-wired yet; integration coverage is owed when a later slice threads &Invocation through the four capability mediators (testing.md: crate-tier is appropriate when the integration harness cannot reach the path).

Checks

cargo test -p ironclaw_host_api — 38 + 57 + 5 green
cargo clippy -p ironclaw_host_api --all-targets --all-features -- -D warnings — clean
cargo test -p ironclaw_architecture — all boundary + the localdev/deployment-mode/inmemory ratchets green
scripts/pre-commit-safety.sh — clean

Next sub-slices (planned, stacked)

C.2 — Resolution/Blocked/Suspension/HostFailure/Outcome result channels (the §5.3 acceptance table maps all 10 CapabilityOutcome variants; kills §1.2's Ok(Failed)-vs-Err ambiguity).
C.3 — LoopRequest + resolve() membrane.
C.4 — sealed Authorized + authorize() (security milestone — the seal becomes load-bearing when the last policy check is inlined, §9).
Then thread &Invocation/&Authorized through the four mediators without merging crates (§9 step 3) and measure type/dyn counts (step 4).

🤖 Generated with Claude Code

ironloopai · 2026-07-18T03:09:42Z

🔎 IronLoop Review Status

Head: 2dfc33d66a50b1fe9a9c4495d6653dd8d02a2a4f
Result: One or more review results were superseded by a newer PR head.
Next: Run @ironloopai review on the latest PR head.
Updated: 2026-07-18T03:49:51.947Z

Current reviewers:

Reviewer	State	Verdict	Findings	Last update
ironloop/common-reviewer (`reviewer`)	Superseded	N/A	N/A	2026-07-18T03:18:45.343Z

Reviewer summaries

Reviewer	Detail
ironloop/common-reviewer (`reviewer`)	Superseded by a newer PR head. New head: `c70f7fd`. Previous verdict: Review declined.

Recent activity

Time	Reviewer	State	Detail
2026-07-18T03:09:41.752Z	ironloop/common-reviewer (`reviewer`)	Queued	Accepted review request for head `277f774`.
2026-07-18T03:09:41.752Z	ironloop/common-reviewer (`reviewer`)	Queued	Waiting for this reviewer lane to become available.
2026-07-18T03:09:42.000Z	ironloop/common-reviewer (`reviewer`)	Started	Reviewer worker started.
2026-07-18T03:09:44.521Z	ironloop/common-reviewer (`reviewer`)	Workspace ready	Prepared isolated checkout (head_ref) at `277f774`.
2026-07-18T03:11:16.283Z	ironloop/common-reviewer (`reviewer`)	Superseded	A newer PR head replaced this review (`50e1c2e`).
2026-07-18T03:15:27.533Z	ironloop/common-reviewer (`reviewer`)	Superseded	A newer PR head replaced this review (`c70f7fd`).
2026-07-18T03:18:45.343Z	ironloop/common-reviewer (`reviewer`)	Result captured	Skipped; 0 blocking findings.
2026-07-18T03:18:45.343Z	ironloop/common-reviewer (`reviewer`)	Superseded	A newer PR head replaced this review (`c70f7fd`).

Available commands

@ironloopai help
@ironloopai agents
@ironloopai review
@ironloopai review --agent <agent>

Run metadata

Admission: webhook accepted the request and IronLoop persisted reviewer state before this projection.

coderabbitai · 2026-07-18T03:09:55Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

🗂️ Base branches to auto review (2)

staging
reborn-integration

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 2c45bb76-d611-4376-a4f4-323ed91184d2

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

gemini-code-assist

Code Review

This pull request consolidates the outbound state stores by removing several parallel in-memory implementations (such as InMemoryOutboundStateStore and InMemoryDeliveredGateRouteStore) and replacing them with the production FilesystemOutboundStateStore running over an in-memory filesystem backend for tests. It also introduces Slice-C kernel capability vocabulary types (Invocation, InvocationOrigin, ProductKind, RoutineId, and ActivityId) in ironclaw_host_api as part of the capability-path DTO collapse. Feedback is provided to improve test assertions in invocation.rs by asserting against specific error variants instead of using generic is_err() calls.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

gemini-code-assist · 2026-07-18T03:11:19Z

+        assert!(ProductKind::new("").is_err());
+        assert!(RoutineId::new("").is_err());
+        // Uppercase-leading is rejected by the name-segment validator.
+        assert!(ProductKind::new("Settings").is_err());


According to the general rules, we should avoid using generic is_err() assertions in tests when verifying that an operation fails. Instead, assert against the specific expected error message or variant to prevent infrastructure or harness-level failures from causing false positives.

Suggested change

assert!(ProductKind::new("").is_err());

assert!(RoutineId::new("").is_err());

// Uppercase-leading is rejected by the name-segment validator.

assert!(ProductKind::new("Settings").is_err());

assert!(matches!(ProductKind::new(""), Err(crate::ids::IdError::InvalidFormat { .. })));

assert!(matches!(RoutineId::new(""), Err(crate::ids::IdError::InvalidFormat { .. })));

// Uppercase-leading is rejected by the name-segment validator.

assert!(matches!(ProductKind::new("Settings"), Err(crate::ids::IdError::InvalidFormat { .. })));

References

Avoid using generic is_err() assertions in tests when verifying that an operation fails. Instead, assert against the specific expected error message or variant to prevent infrastructure or harness-level failures from causing false positives.

Applied in the follow-up commit: the rejection tests now assert the specific kind + reason from HostApiError::InvalidId's message (product/routine + "must not be empty", and the uppercase-leading case pins the kind) instead of bare is_err().

Gemini note on #6223: pin the kind + reason of each expected rejection instead of bare is_err(), so an infrastructure failure can't masquerade as a validation pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

railway-app · 2026-07-18T03:21:37Z

🚅 Deployed to the ironclaw-pr-6223 environment in ironclaw-ci-preview

Service	Status	Web	Updated (UTC)
ironclaw	✅ Success (View Logs)	Web	Jul 18, 2026 at 4:07 am

ilblackdragon · 2026-07-18T03:28:50Z

✅ Ready for merge

Reviewed, finding fixed with reply, CI fully green (19 pass / 0 fail) on head c70f7fdac. Stacked on #6222 — merges last in the chain.

Review verdict: a clean additive Slice-C.1 vocabulary landing — correctly placed in ironclaw_host_api per its vocabulary-only guardrail (no wiring, no runtime behavior), the Invocation/InvocationOrigin/ActivityId/ProductKind/RoutineId shapes follow the crate's existing string_id!/uuid_id! conventions, the origin-sealing and authorize-fold design is documented against the plan's §3/§5.2.1/§11.3, and the coexistence-during-migration story (type count rises before it falls) is stated honestly. Tests pin the serde tags against the kind() discriminant so accounting keys can't drift.
Gemini finding fixed (c70f7fdac): validation-rejection tests assert the specific kind + reason from HostApiError::InvalidId instead of bare is_err().

🤖 Generated with Claude Code

Gemini note on #6223: pin the kind + reason of each expected rejection instead of bare is_err(), so an infrastructure failure can't masquerade as a validation pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Second §4.4 de-prefix slice (after C1's LocalDevRootFilesystem inline). Advances the doc's §4.4 enforcement endgame — "no public type name contains Local/LocalDev/Hosted/Enterprise" — one more type. `LocalDevOutboundStores` is a plain bundle struct (four outbound-store handle fields), NOT a cfg-switched policy/mode type. Its own constructor comment says it "works in both durable (libsql/postgres) and no-durable (in-memory backend) builds" — so the `LocalDev` prefix is factually wrong (it is used in libsql/ postgres PRODUCTION builds, not just local-dev). This is a bucket-2-style mis-prefix: a genuine composition type that only LOOKS like a deployment-mode leak, so the fix is a de-prefix rename, not the bucket-1 DeploymentConfig resolution the cfg-switched `LocalDev*Store` aliases need. Renamed the struct + its 3 use sites (all in factory.rs) to `OutboundStores` (name was free); trimmed it from the R2 `reborn_localdev_typename` ratchet allowlist. The `local_dev_outbound_store` builder fn keeps its name (fn names are not type-name leaks; the ratchet inventories types only). Pure type rename, semantically identical. Verified: localdev ratchet 4; `cargo build -p ironclaw_reborn_composition` (default + libsql+slack+telegram) clean; clippy -D warnings clean; fmt + pre-commit clean. Stacked on #6218. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…name ratchet (§4.4/§10) Completes the §4.4-mandated enforcement — "no public type name contains Local/LocalDev/Hosted/Enterprise" (the doc says: "Enforce it with an ironclaw_architecture test"). The existing `reborn_localdev_typename_ratchet` owns `LocalDev*` (shrinking to empty as Slice B lands) and explicitly scopes out "the broader Local*/Hosted* audit ... as a separate concern." This companion ratchet owns that separate concern — the OTHER three prefixes: - `Enterprise*` — NONE exist (achieved). Empty allowlist locks it in: a new `EnterpriseTierPolicy`-style mode leak fails the "no new" check. - `Hosted*` — all `HostedMcp*`/discovery/egress, a Bucket-3 FALSE POSITIVE ("hosted MCP" is a real domain concept — a platform-hosted MCP server — not a hosted-TIER deployment mode). Frozen/justified so a genuine `HostedTierRuntime` leak can't slip in behind them. - `Local*` (excluding `LocalDev*` → sibling ratchet, and `Locale*` → localization false positive) — the `LocalTriggerAccess*` family is genuine Bucket-1 debt (§4.4 folds `local_trigger_access` into "seed owner grant from config at boot," a policy value); `LocalInvocationServicesResolver` awaits a design-call rename. Reuses the shared `ratchet_support` scanner (comments/strings stripped, visibility-aware, skips tests/examples/benches). Same frozen-set contract as the sibling ratchets: no new type, no duplicate definition, trim on delete/rename. Predicate uses `starts_with` (not `contains`) so mid-word `Local` (`HookLocalId`) is not flagged; a self-test pins that + the `LocalDev`/`Locale` exclusions. Test-only, no production change. Verified: 2 tests pass (frozen allowlist + predicate self-test); clippy -D warnings clean; fmt + pre-commit clean. Stacked on #6220. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…efixes Addresses all three review findings on #6222: - IronLoop: §4.4's rule is "no public type name CONTAINS Local/Hosted/ Enterprise" — the predicate now matches the terms at any CamelCase word boundary (followed by uppercase/digit/underscore/end), which surfaces and freezes 16 previously-invisible mode-shaped mid-names: the RebornLocal* composition family, the Reborn*LocalTriggerAccess* backends, mid-name LocalDev types, and HookLocalId (justified-keep: hook-local id is a domain concept, annotated as such). - Gemini (both): localization names need no hand-listed exceptions under boundary matching — Locale*/Localization*/Localised* continue lowercase so the word is not "Local"; the self-test now pins those exclusions plus the mid-name positives (RebornLocalRuntimeServices, HookLocalId, SelfHostedMcpClient). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

First sub-slice of Slice C (arch-simplification §3/§5) — the capability-path DTO collapse. Per the migration plan (§9), the kernel vocabulary lands in `ironclaw_host_api` *first*, ahead of any wiring; this PR adds the single data-plane payload every layer will reference, replacing the request side of the ~14-hop re-wrap (§1.1). - ids.rs (via the crate's own `uuid_id!`/`string_id!` macros): - `ActivityId` — the invocation's idempotency identity (§11.3); what §1.1's dead-future `idempotency_key` becomes, unified rather than deleted. - `ProductKind` / `RoutineId` — validated string newtypes for the two non-loop origins (kept strings, not enums, while the product/routine sets evolve, §5.8). - invocation.rs (new): - `InvocationOrigin { LoopRun(RunId) | Product(ProductKind) | Automation(RoutineId) }` — sealed-at-membrane origin (§5.2.1); snake_case wire-tagged; `.kind()` discriminant pinned to the tag for per-origin accounting views (§5.3.3). - `Invocation { activity_id, capability, input, scope, actor, origin, estimate }` — the "one payload" (§3/§4.1). Reuses existing host_api types; `actor` is required (sealed), and authorization *outputs* (`mounts`, `resource_reservation`) are deliberately absent — they move into the sealed `Authorized` in a later slice. Additive only — nothing is wired into the dispatch path yet. The five old request DTOs still exist and function; per §9 the type count rises before it falls (~14 → ~18 → ~11) as the new vocabulary and old shapes coexist. `check-type-duplicates.py` does not flag `Invocation` against `CapabilityDispatchRequest` (it is a genuinely distinct state, not a mirror). Tests are crate-tier because the types are not production-wired yet; integration coverage is owed when a later slice threads `&Invocation` through the four capability mediators (testing.md: crate-tier when the harness cannot reach the path). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Gemini note on #6223: pin the kind + reason of each expected rejection instead of bare is_err(), so an infrastructure failure can't masquerade as a validation pass. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

ilblackdragon · 2026-07-18T04:08:42Z

Post-restack confirmation (after #6216/#6218 merges): rebased onto main via the chain, CI fully green — 19 pass / 0 fail on head 2dfc33d66. Still ready for merge — last in the chain after #6220 and #6222.

railway-app Bot temporarily deployed to ironclaw-ci-preview / ironclaw-pr-6223 July 18, 2026 03:09 Destroyed

github-actions Bot added the size: XL 500+ changed lines label Jul 18, 2026

github-actions Bot added scope: dependencies Dependency updates risk: medium Business logic, config, or moderate-risk modules contributor: core 20+ merged PRs and removed size: XL 500+ changed lines labels Jul 18, 2026

ilblackdragon force-pushed the refactor/reborn-slicec-invocation-vocabulary branch from 277f774 to 50e1c2e Compare July 18, 2026 03:11

railway-app Bot temporarily deployed to ironclaw-ci-preview / ironclaw-pr-6223 July 18, 2026 03:11 Destroyed

gemini-code-assist Bot reviewed Jul 18, 2026

View reviewed changes

github-actions Bot added size: L 200-499 changed lines risk: low Changes to docs, tests, or low-risk modules and removed risk: medium Business logic, config, or moderate-risk modules labels Jul 18, 2026

railway-app Bot temporarily deployed to ironclaw-ci-preview / ironclaw-pr-6223 July 18, 2026 03:15 Destroyed

ilblackdragon force-pushed the refactor/reborn-deployment-mode-typename-ratchet branch from b94edc5 to 361f06e Compare July 18, 2026 03:30

ilblackdragon force-pushed the refactor/reborn-slicec-invocation-vocabulary branch from c70f7fd to a581269 Compare July 18, 2026 03:30

railway-app Bot temporarily deployed to ironclaw-ci-preview / ironclaw-pr-6223 July 18, 2026 03:30 Destroyed

ilblackdragon and others added 5 commits July 18, 2026 03:49

ilblackdragon force-pushed the refactor/reborn-deployment-mode-typename-ratchet branch from 361f06e to 74e84c9 Compare July 18, 2026 03:49

ilblackdragon force-pushed the refactor/reborn-slicec-invocation-vocabulary branch from a581269 to 2dfc33d Compare July 18, 2026 03:49

railway-app Bot temporarily deployed to ironclaw-ci-preview / ironclaw-pr-6223 July 18, 2026 03:49 Destroyed

Base automatically changed from refactor/reborn-deployment-mode-typename-ratchet to main July 18, 2026 04:09

ilblackdragon merged commit 2a1e4be into main Jul 18, 2026
24 checks passed

ilblackdragon deleted the refactor/reborn-slicec-invocation-vocabulary branch July 18, 2026 04:36

ilblackdragon mentioned this pull request Jul 18, 2026

feat(host_api): Slice C.2 — HostFailure infra-error channel (#6168) #6225

Merged

BenKurrek mentioned this pull request Jul 18, 2026

feat(reborn): unified generic extension runtime + Option A honest state machine (reconcile main) #6116

Merged

7 tasks done

ilblackdragon mentioned this pull request Jul 19, 2026

docs(reborn): fold #6284 recoverability contract into the resolution model + implementation status (r8) #6291

Merged

Navigation Menu

Repository navigation

feat(host_api): Slice C.1 — Invocation payload vocabulary (#6168) - #6223

feat(host_api): Slice C.1 — Invocation payload vocabulary (#6168)#6223
ilblackdragon merged 5 commits into
mainfrom
refactor/reborn-slicec-invocation-vocabulary

ilblackdragon commented Jul 18, 2026

Uh oh!

ironloopai Bot commented Jul 18, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Jul 18, 2026 •

edited

Loading

Review skipped

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot Jul 18, 2026

Uh oh!

ilblackdragon Jul 18, 2026

Uh oh!

railway-app Bot commented Jul 18, 2026 •

edited

Loading

Uh oh!

ilblackdragon commented Jul 18, 2026

Uh oh!

ilblackdragon commented Jul 18, 2026

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

ilblackdragon commented Jul 18, 2026

What

Types added

Design reconciliations (documented in the code)

Why additive is safe (§9)

Testing

Checks

Next sub-slices (planned, stacked)

Uh oh!

ironloopai Bot commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔎 IronLoop Review Status

Uh oh!

coderabbitai Bot commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Jul 18, 2026

Choose a reason for hiding this comment

Uh oh!

ilblackdragon Jul 18, 2026

Choose a reason for hiding this comment

Uh oh!

railway-app Bot commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ilblackdragon commented Jul 18, 2026

✅ Ready for merge

Uh oh!

ilblackdragon commented Jul 18, 2026

Uh oh!

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

ironloopai Bot commented Jul 18, 2026 •

edited

Loading

coderabbitai Bot commented Jul 18, 2026 •

edited

Loading

railway-app Bot commented Jul 18, 2026 •

edited

Loading