fix: await trajectory subscriber delivery - #727
Conversation
Signed-off-by: Ajay Thorve <athorve@nvidia.com>
github-actions
Bot
added
size:M
WalkthroughThe CLI now detects trajectory-finalizing events and flushes queued subscribers before closing hook responses. The runtime exposes the flush barrier. Tests cover event classification and Codex ChangesTrajectory boundary delivery
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant CodexHook
participant apply_hook_outcome
participant flush_queued_subscribers
participant ATIFSubscribers
CodexHook->>apply_hook_outcome: apply normalized hook events
apply_hook_outcome->>flush_queued_subscribers: flush after trajectory-finalizing event
flush_queued_subscribers->>ATIFSubscribers: deliver queued callbacks
ATIFSubscribers-->>flush_queued_subscribers: complete or return failure
flush_queued_subscribers-->>apply_hook_outcome: barrier result
apply_hook_outcome-->>CodexHook: return hook response or launch error
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 8
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@crates/cli/src/server/mod.rs`:
- Around line 1231-1235: Change the error mapping around the subscriber delivery
barrier spawned by spawn_blocking to use the request-specific CliError variant
instead of CliError::Launch, while preserving the existing message text and
inner FlowError propagation.
- Around line 1227-1236: Bound the await around flush_queued_subscribers in the
await_delivery branch using a timeout such as TRAJECTORY_BARRIER_TIMEOUT,
defined with the other server tunables. Preserve existing task and subscriber
errors, but when the timeout expires log the condition and return the hook
response without propagating a barrier failure, restoring asynchronous degraded
behavior and preventing blocking-pool starvation.
In `@crates/cli/tests/coverage/shared/events_tests.rs`:
- Around line 17-30: Extend trajectory_boundaries_require_subscriber_delivery
with negative assertions that NormalizedEvent::SubagentEnded, LlmEnded, and
ToolEnded return false from finalizes_trajectory. Construct each variant using
the existing test-module constructors and required payload types, while
preserving the current positive and HookMark assertions.
In `@crates/cli/tests/coverage/shared/server_tests.rs`:
- Line 1175: Replace the unwrap on release_tx.send in the surrounding test flow
with an ignored send result, allowing execution to reach the descriptive
response-ordering assertion when the receiver has already been dropped.
- Line 1201: Remove the exact "ATIF-v1.7" assertion from the
response-to-snapshot ordering test, keeping the test focused on ordering
behavior and aligned with the sibling test’s content-based assertions.
- Around line 1164-1167: Increase the timeout used by the readiness wait around
started_rx in the sanitizer test from one second to a few seconds, while leaving
the later 100 ms release window unchanged because it still detects a genuine
defect.
In `@crates/core/src/api/runtime/subscriber_dispatcher.rs`:
- Around line 1265-1269: Update the doc comment for flush_queued_subscribers to
explicitly state that it must not be called from subscriber, event-sanitizer,
guardrail, or intercept callbacks, because it does not establish a barrier there
and may create a wait cycle.
In `@docs/about-nemo-relay/concepts/subscribers.mdx`:
- Around line 150-155: Move the gateway barrier paragraph from before the
milestones table to after the table’s final row and before the “Use the
subscriber flush API” paragraph. Keep the paragraph text unchanged so the table
lead-in directly introduces the table.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Enterprise
Run ID: 98627723-4061-47cb-b8ce-d7bf3ecc3b0a
📒 Files selected for processing (9)
crates/cli/src/events/mod.rscrates/cli/src/server/mod.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/core/src/api/runtime.rscrates/core/src/api/runtime/subscriber_dispatcher.rsdocs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxintegrations/coding-agents/codex/README.md
📜 Review details
⏰ Context from checks skipped due to timeout. (2)
- GitHub Check: Changes / Detect
- GitHub Check: Preview docs
🧰 Additional context used
📓 Path-based instructions (34)
**/*.rs
📄 CodeRabbit inference engine (.agents/skills/prepare-pr/SKILL.md)
**/*.rs: Any Rust change must runjust test-rust
Any Rust change must runcargo fmt --all
Any Rust change must runcargo clippy --workspace --all-targets -- -D warnings
**/*.rs: Runcargo fmt --allfor all FFI work since it is Rust work
Runjust test-rustto validate FFI changes
Runcargo clippy --workspace --all-targets -- -D warningsto enforce strict linting on FFI workWhen Rust files changed as part of Go work, also run
cargo fmt --all,just test-rust, andcargo clippy --workspace --all-targets -- -D warnings
**/*.rs: Runcargo fmt --allwhen Rust files are changed as part of Node work
Runcargo clippy --workspace --all-targets -- -D warningswhen Rust files are changed as part of Node work
Runjust test-rustwhen Rust files are changed as part of Node work
**/*.rs: If any Rust code changed, always runjust test-rust.
If any Rust code changed, also runcargo fmt --all.
If any Rust code changed, also runcargo clippy --workspace --all-targets -- -D warnings.
For Rust changes headed for review, runcargo fmt --allandcargo clippy --workspace --all-targets -- -D warningseven if relying on pre-commit.
**/*.rs: Format Rust code with rustfmt defaults usingcargo fmt.
Runcargo clippy -- -D warnings; all Rust warnings must be treated as errors.
Use Rustsnake_casenaming conventions.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
{crates/**/src/**/*.rs,python/**/*.py}
📄 CodeRabbit inference engine (.agents/skills/maintain-dynamic-plugins/SKILL.md)
Do not add tests under
src; Rust tests belong in cratetests/trees, and Python SDK tests belong underpython/tests.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/src/server/mod.rs
**/*
📄 CodeRabbit inference engine (.agents/skills/validate-change/SKILL.md)
**/*: Format changed files with the language-native formatter before the final lint/test pass.
If dynamic plugin behavior changed, usemaintain-dynamic-pluginsand include the native SDK, worker protocol, Python SDK, docs, packaging, and Codecov surfaces in the validation plan.
If code changes alter APIs, bindings, commands, paths, packaging behavior, observability/adaptive semantics, or documented best practices, update any dependent maintainer or consumer skills in the same branch.
During iteration, preferuv run pre-commit run --files <changed files...>.
Before review or handoff, runuv run pre-commit run --all-files.
**/*: Use release tags in raw Rust-compatible SemVer without a leadingv; tags such asv0.1.0are prohibited.
Use branch prefixesfeat/,fix/,docs/,test/, orrefactor/according to the change purpose.
Every commit in a pull request must include a DCOSigned-off-by:sign-off.
Before submitting a pull request, ensure pre-commit hooks, relevant tests, target-specific builds, documentation updates, and a rebase on the latestmainare complete.
Use commit messages in the formtype: short description, with a valid type and a first line under 72 characters.
Files:
crates/cli/src/events/mod.rsdocs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxcrates/core/src/api/runtime/subscriber_dispatcher.rsintegrations/coding-agents/codex/README.mdcrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,go,js,ts}
📄 CodeRabbit inference engine (.agents/skills/validate-change/SKILL.md)
If a language surface changed, always run that language's test target even when Rust core did not change.
**/*.{rs,py,go,js,ts}: When observability configuration or lifecycle is exposed, keep FFI and Python, Go, and Node.js binding-native config objects and subscriber/exporter methods aligned in logical knobs and semantics.
Require every OpenTelemetry endpoint to have a type and nonblank destination; resolveheader_envvalues at activation and reject missing, blank, or duplicate headers.
Concatenate layered ATOF sink, ATIF storage, and OpenTelemetry endpoint lists with higher-precedence entries first.
Preserve correct handling of mark events, start/end events, orphan cases, and span or trajectory fields derived from intended event data.
Run affected Rust tests andjust test-rustwhen event fields change; runjust test-python,just test-go, andjust test-nodewhen binding-native configuration or lifecycle changes.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,js,ts,tsx,go,java,kt,swift}
📄 CodeRabbit inference engine (.agents/skills/add-middleware/SKILL.md)
Add tests covering registration and duplicate names, deregistration and missing names, priority ordering, callback failure policy, scope-local inheritance and cleanup, event payload semantics, immutable mark and scope fields, and parity across affected bindings.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,js,jsx,ts,tsx,go,c,h,cc,cpp,md,toml,yml,yaml,sh}
📄 CodeRabbit inference engine (AGENTS.md)
Keep SPDX headers on source, documentation, scripts, and configuration files; the project is Apache-2.0.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rsintegrations/coding-agents/codex/README.mdcrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py}
📄 CodeRabbit inference engine (AGENTS.md)
Use
snake_casenaming in Rust and Python.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
crates/**/*.rs
📄 CodeRabbit inference engine (AGENTS.md)
crates/**/*.rs: UseJson = serde_json::Valuein Rust-facing runtime APIs where existing code expects JSON payloads.
Treat Rust as the source of truth for runtime behavior; binding APIs should mirror Rust semantics unless a language-specific wrapper intentionally improves ergonomics.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,js,mjs,ts,tsx}
📄 CodeRabbit inference engine (AGENTS.md)
Preserve the existing Tokio-based asynchronous model and callback/future lifetimes; do not unexpectedly block or hide async work in bindings.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,go,js,jsx,ts,tsx,c,h}
📄 CodeRabbit inference engine (CONTRIBUTING.md)
**/*.{rs,py,go,js,jsx,ts,tsx,c,h}: Run tests for every language affected by a change; changes to the core Rust crate require tests across all bindings.
UseSONAR_IGNORE_START/SONAR_IGNORE_ENDonly for documented false positives, keep ignored blocks minimal, explain them with a comment, and obtain reviewer sign-off.
Preserve the layered architecture in which Rust provides the core runtime and C FFI, PyO3, and NAPI provide bindings that mirror the full API surface.
Files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.{rs,py,go,js,jsx,ts,tsx,c,h,html,md,mdx,toml}
📄 CodeRabbit inference engine (CONTRIBUTING.md)
Include the appropriate SPDX copyright and Apache-2.0 license header in every source file.
Files:
crates/cli/src/events/mod.rsdocs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxcrates/core/src/api/runtime/subscriber_dispatcher.rsintegrations/coding-agents/codex/README.mdcrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
**/*.mdx
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/SKILL.md)
MDX top-of-file SPDX comments must use {/* ... */} delimiters instead of HTML comment delimiters (Must-Fix)
**/*.mdx: In MDX files, top-of-file comments must use JSX comment delimiters ({/*and*/}); do not use HTML comments for MDX SPDX headers.
New or regenerated MDX files must use{/* ... */}for top-of-file SPDX comments.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdx
{docs,examples}/**/*
📄 CodeRabbit inference engine (.agents/skills/rename-surfaces/SKILL.md)
Update docs and examples.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdx
docs/**/*
📄 CodeRabbit inference engine (.agents/skills/validate-change/SKILL.md)
If documentation examples or commands under
docs/change, run the targeted docs checks appropriate to the change.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdx
**/*.{md,mdx}
📄 CodeRabbit inference engine (.agents/skills/validate-change/SKILL.md)
If links in documentation change, run
just docs-linkcheck.Use documented public APIs and stable wrapper commands in examples and user-facing documentation; do not rely on internal helpers.
**/*.{md,mdx}: Prefer the documented public API over internal shortcuts in documentation and examples.
Keep package names, repository references, and build commands current.
Contribution workflow documentation must require an issue before external contribution pull requests and note that NVIDIA contributors may use a GitHub or Linear issue.
Update entry-point documentation when examples or reading paths change.
Keep release-process and release-notes guidance in maintainer documentation such asRELEASING.md, rather than user-facing documentation pages orCHANGELOG.md.
Use stable user-facing wrappers at thescripts/root in documentation and examples; reference namespaced helper paths only for internal maintenance documentation.
When detailed dynamic plugin guides exist, keep Rust native plugin examples, Python worker plugin examples, andgrpc-v1protocol details on separate pages.
Dynamic plugin manifests in documentation and examples should usecompat.relay = ">=0.5,<1.0"unless deliberately narrower.
Render images, diagrams, tables, and other visual content at representative page widths, ensuring legibility and complete access without clipping; use responsive scaling, reflow, or overflow as appropriate and scope visual styling narrowly.
Dynamic plugin entry pages should link to native, worker, Rust example, Python example, and protocol pages when those pages exist.
Images, diagrams, tables, and custom visual content must remain legible and fully accessible at representative desktop and narrow page widths.
Release-policy documentation must point to GitHub Releases as the only release-history source of truth.
Runjust docswhen the documentation site changes; retain./scripts/build-docs.sh htmlas the compatibility wrapper.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxintegrations/coding-agents/codex/README.md
**/*.{md,mdx,py,go,js,ts}
📄 CodeRabbit inference engine (.agents/skills/maintain-observability/SKILL.md)
Examples and documentation must use each exporter's documented flush/deregister order before shutdown.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxintegrations/coding-agents/codex/README.md
docs/{about-nemo-relay/concepts/subscribers.mdx,configure-plugins/observability/**/*.mdx}
📄 CodeRabbit inference engine (.agents/skills/maintain-observability/SKILL.md)
Update observability documentation and examples alongside implementation changes, including configuration version 3 with one
opentelemetrysection containing typed endpoints and no standalone public OpenInference surface.
Files:
docs/about-nemo-relay/concepts/subscribers.mdx
docs/**/*.{md,mdx}
📄 CodeRabbit inference engine (CONTRIBUTING.md)
Update relevant reference documentation when public behavior or APIs change.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdx
**/*.{md,mdx,rst}
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/assets/nvidia-style-technical-docs.md)
**/*.{md,mdx,rst}: Use title case consistently for technical documentation headings and table headers; avoid quotation marks, ampersands, and exclamation marks in headings, while preserving official product, event, research, and whitepaper title case.
Format code elements, commands, parameters, package names, expressions, directories, file names, and paths in monospace; represent path placeholders with angle brackets inside monospace.
Format UI buttons, menus, fields, and labels in bold, and separate consecutive UI navigation labels with>.
Use quotation marks for error messages and strings when appropriate, italics for newly introduced terms and publication titles, and plain text for keyboard shortcuts.
Represent GitHub repositories with owner/repository link text, such as[NVIDIA/NeMo](link), rather than generic repository wording.
Introduce every code block with a complete sentence; do not let a code block complete or interrupt the grammar of surrounding prose; use syntax highlighting when supported.
Keep inline method, function, and class references consistent with nearby documentation; omit empty parentheses in prose when no call is shown.
Use descriptive link text matching the destination title when possible; avoid raw URLs, generic anchors, long-sentence links, and unnecessary links that distract from procedures.
Ensure lists have a complete lead-in sentence, more than one item, no more than two levels, parallel construction, one idea or action per item, and appropriate punctuation; use bullets for unordered items and numbers for ordered tasks.
Format definition lists with a bold term followed by a complete, parallel, punctuated definition.
Use tables for reference information, decision support, compatibility matrices, and comparable choices; flag one-row tables, missing captions or lead-ins, sentence-case headers where title case is expected, unexplained empty cells, and code or links that would be clearer as prose.
Write procedure steps as imperative ...
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdxintegrations/coding-agents/codex/README.md
{docs/**,README.md,CONTRIBUTING.md,RELEASING.md,SECURITY.md}
⚙️ CodeRabbit configuration file
{docs/**,README.md,CONTRIBUTING.md,RELEASING.md,SECURITY.md}: Review documentation for technical accuracy against the current API, command correctness, and consistency across language bindings.
Flag stale examples, missing SPDX headers where required, and instructions that no longer match CI or pre-commit behavior.
Files:
docs/about-nemo-relay/concepts/subscribers.mdxdocs/nemo-relay-cli/codex.mdx
{crates/core,crates/adaptive}/**/*
📄 CodeRabbit inference engine (.agents/skills/prepare-pr/SKILL.md)
Changes to
crates/coreorcrates/adaptivemust run the full language matrix
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
crates/core/**/*.rs
📄 CodeRabbit inference engine (.agents/skills/test-go-binding/SKILL.md)
If the change touched
crates/coreor shared runtime semantics, also usevalidate-changefor broader validationUse
Result<T>withFlowErrorin core runtime paths, keeping wrapper-layer errors explicit and binding-appropriate.
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
crates/core/src/{api/**/*.rs,api/runtime/**/*.rs,codec/**/*.rs,json.rs}
📄 CodeRabbit inference engine (.agents/skills/add-binding-feature/SKILL.md)
Implement the new or changed public runtime behavior first in the Rust core, especially under
crates/core/src/api/and related core modules such ascrates/core/src/api/runtime/,crates/core/src/codec/, andcrates/core/src/json.rs.
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
crates/{core,adaptive}/**/*
📄 CodeRabbit inference engine (.agents/skills/validate-change/SKILL.md)
If
crates/coreorcrates/adaptivechanged, run the full validation matrix across Rust, Python, Go, and Node.js.
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
crates/core/src/api/**/*.rs
📄 CodeRabbit inference engine (.agents/skills/add-middleware/SKILL.md)
Preserve the documented pipeline order: conditional guardrails, request intercepts, request sanitization, execution intercepts, and response sanitization for tool and LLM execution; specialized sanitization, event creation, and dispatch for mark and scope events.
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
crates/{core,adaptive}/**/*.rs
⚙️ CodeRabbit configuration file
crates/{core,adaptive}/**/*.rs: Review the Rust runtime for async correctness, scope isolation, middleware ordering, and event lifecycle regressions.
Pay close attention to task-local/thread-local scope propagation, callback lifetimes, stream finalization, and root_uuid isolation.
Public API changes should preserve existing behavior unless tests and docs show the intended migration path.
Files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
**/*.{md,rst,html,txt}
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/assets/nvidia-style-brand-terminology.md)
**/*.{md,rst,html,txt}: Always spellNVIDIAin all caps. Do not useNvidia,nvidia,nVidia,nVIDIA, orNV.
Usean NVIDIAbefore a noun because the name starts with an 'en' sound.
Do not add a registered trademark symbol afterNVIDIAwhen referring to the company.
Use trademark symbols with product names only when the document type or legal guidance requires them.
Verify official capitalization, spacing, and hyphenation for product names.
Precede NVIDIA product names withNVIDIAon first mention when it is natural and accurate.
Do not rewrite product names for grammar or title-case rules.
Preserve third-party product names according to the owner's spelling.
Include the company name and full model qualifier on first use when it helps identify the model.
Preserve the official capitalization and punctuation of model names.
Use shorter family names only after the full name is established.
Spell out a term on first use and put the acronym in parentheses unless the acronym is widely understood by the intended audience.
Use the acronym on later mentions after it has been defined.
For long documents, reintroduce the full term if readers might lose context.
Form plurals of acronyms withs, not an apostrophe, such asGPUs.
In headings, common acronyms can remain abbreviated. Spell out the term in the first or second sentence of the body.
Common terms such asCPU,GPU,PC,API, andUIusually do not need to be spelled out for developer audiences.
Files:
integrations/coding-agents/codex/README.md
**/*.{md,rst,html}
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/assets/nvidia-style-brand-terminology.md)
Link the first mention of a product name when the destination helps the reader.
Files:
integrations/coding-agents/codex/README.md
**/*.{md,rst,txt}
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/assets/nvidia-style-guide.md)
Spell
NVIDIAin all caps. Do not useNvidia,nvidia, orNV.
Files:
integrations/coding-agents/codex/README.md
**/*.{md,rst}
📄 CodeRabbit inference engine (.agents/skills/review-doc-style/assets/nvidia-style-guide.md)
**/*.{md,rst}: Format commands, code elements, expressions, package names, file names, and paths as inline code.
Use descriptive link text. Avoid raw URLs and weak anchors such as "here" or "read more."
Use title case consistently for technical documentation headings.
Introduce code blocks, lists, tables, and images with complete sentences.
Write procedures as imperative steps. Keep steps parallel and split long procedures into smaller tasks.
Prefer active voice, present tense, short sentences, contractions, and plain English.
Usecanfor possibility and reservemayfor permission.
Useafterfor temporal relationships instead ofonce.
Preferrefer tooverseewhen the wording points readers to another resource.
Avoid culture-specific idioms, unnecessary Latinisms, jokes, and marketing exaggeration in technical docs.
Spell out months in body text, avoid ordinal dates, and use clear time zones.
Spell out whole numbers from zero through nine unless they are technical values, parameters, versions, or UI values.
Use numerals for 10 or greater and include commas in thousands.
Do not add trademark symbols to learning-oriented docs unless the source, platform, or legal guidance explicitly requires them.
Files:
integrations/coding-agents/codex/README.md
**/README.md
📄 CodeRabbit inference engine (CONTRIBUTING.md)
Update the relevant crate or package README when that package surface changes.
Update relevant package or crate
README.mdfiles when examples or binding guidance changes.
Files:
integrations/coding-agents/codex/README.md
**/{README.md,docs/index.md}
📄 CodeRabbit inference engine (.agents/skills/contribute-docs/SKILL.md)
Update
README.mdordocs/index.mdwhen documentation entry points change.
Files:
integrations/coding-agents/codex/README.md
**/{test,tests}/**/*
📄 CodeRabbit inference engine (CONTRIBUTING.md)
When adding functionality, include tests in the appropriate test files for each affected language binding.
Files:
crates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rs
{crates/**/tests/**,python/tests/**,go/nemo_relay/**/*_test.go}
⚙️ CodeRabbit configuration file
{crates/**/tests/**,python/tests/**,go/nemo_relay/**/*_test.go}: Tests should cover the behavior promised by the changed API surface, including error paths and cross-request isolation where relevant.
Prefer assertions on lifecycle events, scope stacks, middleware ordering, and binding parity over shallow smoke tests.
Files:
crates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rs
🧠 Learnings (3)
📚 Learning: 2026-08-03T19:55:03.931Z
Learnt from: afourniernv
Repo: NVIDIA/NeMo-Relay PR: 558
File: crates/pii-redaction/src/rampart/mod.rs:265-274
Timestamp: 2026-08-03T19:55:03.931Z
Learning: In NeMo Relay first-party plugin registration helpers, treat the documented duplicate-registration `PluginError::RegistrationFailed` result from `register_plugin` as success when registration is intended to be idempotent. Do not locally reclassify this as `PluginError::Conflict`; changing the classification requires a core-wide review of the public API and FFI behavior.
Applied to files:
crates/cli/src/events/mod.rscrates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rscrates/cli/tests/coverage/shared/events_tests.rscrates/cli/tests/coverage/shared/server_tests.rscrates/cli/src/server/mod.rs
📚 Learning: 2026-07-28T03:31:05.964Z
Learnt from: willkill07
Repo: NVIDIA/NeMo-Relay PR: 564
File: crates/core/src/api/runtime/subscriber_dispatcher.rs:297-314
Timestamp: 2026-07-28T03:31:05.964Z
Learning: In this codebase’s runtime API, do not implement incremental native LLM stream forwarding via the native ABI v3 asynchronous middleware protocol (it can only settle a single JSON value via a one-shot completion handle and cannot forward stream chunks incrementally). If a latency-sensitive plugin needs streaming behavior, review for use of synchronous native stream intercepts or worker plugins instead of trying to chunk-deliver or incrementally forward over the ABI v3 async path.
Applied to files:
crates/core/src/api/runtime/subscriber_dispatcher.rs
📚 Learning: 2026-07-28T20:07:29.880Z
Learnt from: willkill07
Repo: NVIDIA/NeMo-Relay PR: 571
File: crates/core/src/api/runtime/state.rs:996-1020
Timestamp: 2026-07-28T20:07:29.880Z
Learning: In NeMo Relay (RELAY-509), sanitizer callback failures must be treated as intentional fail-open behavior. When an event/tool (request/response) or LLM (request/response) sanitizer callback fails, the sanitizer chain should retain and publish the last valid event/payload snapshot (rather than dropping/invalidating the data) and log the failure including callback context (e.g., which sanitizer/callback failed and relevant identifiers). Apply this consistently across all sanitizer chains mentioned in the RELAY-509 documentation/migration guide.
Applied to files:
crates/core/src/api/runtime/subscriber_dispatcher.rscrates/core/src/api/runtime.rs
🔇 Additional comments (6)
crates/cli/src/events/mod.rs (1)
84-89: LGTM!crates/core/src/api/runtime.rs (1)
35-36: LGTM!crates/cli/src/server/mod.rs (1)
44-49: LGTM!Also applies to: 1188-1188, 1201-1201, 1214-1226
crates/cli/tests/coverage/shared/server_tests.rs (1)
21-24: LGTM!Also applies to: 130-136
docs/nemo-relay-cli/codex.mdx (1)
412-415: LGTM!integrations/coding-agents/codex/README.md (1)
47-51: LGTM!
| if await_delivery { | ||
| // Agent and turn scope-end subscribers write the final trajectory snapshot. Use the | ||
| // queued-only barrier so an active managed publication in another session cannot hold the | ||
| // hook response open. | ||
| tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers) | ||
| .await | ||
| .map_err(|error| { | ||
| CliError::Launch(format!("subscriber delivery barrier task failed: {error}")) | ||
| })??; | ||
| } |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
Bound the barrier wait; an unbounded flush blocks the hook response and the blocking pool.
flush_queued_subscribers waits for every queued subscriber callback. Those callbacks include user-registered subscribers, event sanitizers, and scope-end guardrails. A slow or hung callback holds this await open with no upper bound, and the hook response never returns. The invoking agent process waits on that response, so the coding agent stalls instead of finishing its turn. Your own test proves the mechanism: the codex-atif-boundary-sanitizer in crates/cli/tests/coverage/shared/server_tests.rs keeps the Stop response pending until a channel is released, and only the test's tokio::time::timeout bounds it.
The second effect is pool pressure. Each finalizing hook occupies one Tokio blocking thread for the whole barrier. Concurrent Codex, Claude Code, and Hermes sessions share this gateway, so slow callbacks can saturate the blocking pool and stall unrelated spawn_blocking work.
Wrap the barrier in a timeout. On expiry, log and return the hook response instead of hanging, since the pre-change behavior was fully asynchronous delivery and is the correct degraded mode.
🛡️ Proposed fix to bound the wait
if await_delivery {
// Agent and turn scope-end subscribers write the final trajectory snapshot. Use the
// queued-only barrier so an active managed publication in another session cannot hold the
// hook response open.
- tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers)
- .await
- .map_err(|error| {
- CliError::Launch(format!("subscriber delivery barrier task failed: {error}"))
- })??;
+ let barrier =
+ tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers);
+ match tokio::time::timeout(TRAJECTORY_BARRIER_TIMEOUT, barrier).await {
+ Ok(joined) => joined.map_err(|error| {
+ CliError::Launch(format!("subscriber delivery barrier task failed: {error}"))
+ })??,
+ Err(_) => tracing::warn!(
+ timeout_secs = TRAJECTORY_BARRIER_TIMEOUT.as_secs(),
+ "subscriber delivery barrier timed out; returning hook response with delivery still queued"
+ ),
+ }
}Define the constant near the other server tunables:
/// Upper bound on the trajectory-boundary subscriber barrier. A slow subscriber
/// callback must not hold a hook response open indefinitely.
const TRAJECTORY_BARRIER_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| if await_delivery { | |
| // Agent and turn scope-end subscribers write the final trajectory snapshot. Use the | |
| // queued-only barrier so an active managed publication in another session cannot hold the | |
| // hook response open. | |
| tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers) | |
| .await | |
| .map_err(|error| { | |
| CliError::Launch(format!("subscriber delivery barrier task failed: {error}")) | |
| })??; | |
| } | |
| if await_delivery { | |
| // Agent and turn scope-end subscribers write the final trajectory snapshot. Use the | |
| // queued-only barrier so an active managed publication in another session cannot hold the | |
| // hook response open. | |
| let barrier = | |
| tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers); | |
| match tokio::time::timeout(TRAJECTORY_BARRIER_TIMEOUT, barrier).await { | |
| Ok(joined) => joined.map_err(|error| { | |
| CliError::Launch(format!("subscriber delivery barrier task failed: {error}")) | |
| })??, | |
| Err(_) => tracing::warn!( | |
| timeout_secs = TRAJECTORY_BARRIER_TIMEOUT.as_secs(), | |
| "subscriber delivery barrier timed out; returning hook response with delivery still queued" | |
| ), | |
| } | |
| } |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/src/server/mod.rs` around lines 1227 - 1236, Bound the await
around flush_queued_subscribers in the await_delivery branch using a timeout
such as TRAJECTORY_BARRIER_TIMEOUT, defined with the other server tunables.
Preserve existing task and subscriber errors, but when the timeout expires log
the condition and return the hook response without propagating a barrier
failure, restoring asynchronous degraded behavior and preventing blocking-pool
starvation.
| tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers) | ||
| .await | ||
| .map_err(|error| { | ||
| CliError::Launch(format!("subscriber delivery barrier task failed: {error}")) | ||
| })??; |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Description: Inspect CliError variants, its From impls, and its IntoResponse mapping.
set -euo pipefail
fd -t f 'error.rs' crates/cli/src | while IFS= read -r file; do
echo "===== $file ====="
cat -n "$file"
done
echo "===== From impls / IntoResponse for CliError ====="
rg -nP -C 6 'impl\s+(From<[^>]+>\s+for\s+CliError|IntoResponse\s+for\s+CliError)' crates/cli/src
echo "===== return type of flush_queued_subscribers and its Result alias ====="
ast-grep run --pattern 'pub fn flush_queued_subscribers() -> $RET { $$$ }' --lang rust crates/core/src
rg -nP -C 3 'pub (type|use).*\bResult\b' crates/core/src/api/runtime/subscriber_dispatcher.rs crates/core/src/error.rs 2>/dev/null || trueRepository: NVIDIA/NeMo-Relay
Length of output: 9292
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "===== server call site ====="
sed -n '1195,1250p' crates/cli/src/server/mod.rs
echo "===== CliError construction and hook handlers ====="
rg -n -P -C 4 'CliError::(Launch|Flow|HookDelivery)|flush_queued_subscribers|IntoResponse' crates/cli/src crates/core/src
echo "===== direct status probe from source semantics ====="
python3 - <<'PY'
from pathlib import Path
error = Path("crates/cli/src/error.rs").read_text()
assert "Flow(#[from] nemo_relay::error::FlowError)" in error
assert "Self::Flow(_)" in error
assert "StatusCode::INTERNAL_SERVER_ERROR" in error
server = Path("crates/cli/src/server/mod.rs").read_text()
needle = "tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers)"
assert needle in server
assert "CliError::Launch(format!(\"subscriber delivery barrier task failed: {error}\"))" in server
print("flush_queued_subscribers errors convert through CliError::Flow or the explicit JoinError mapping; both map to HTTP 500.")
print("The explicit JoinError mapping uses CliError::Launch, whose log kind is launch and whose HTTP status is also 500.")
PYRepository: NVIDIA/NeMo-Relay
Length of output: 50373
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "===== log_kind consumers ====="
rg -n -P -C 5 'log_kind\(\)|hook.*error|error_kind' crates/cli/src/server crates/cli/src/hooks crates/cli/src/tests 2>/dev/null | head -n 240
echo "===== hook route registration ====="
rg -n -P -C 6 'claude_hook|hermes_hook|apply_hook_outcome|Router::|route\(' crates/cli/src/server/mod.rs | head -n 260
echo "===== focused error tests ====="
sed -n '1,260p' crates/cli/tests/coverage/shared/error_tests.rs 2>/dev/null || trueRepository: NVIDIA/NeMo-Relay
Length of output: 16204
Use a request-specific error variant for the subscriber delivery barrier.
The inner ? converts FlowError through CliError::Flow, and both error paths return HTTP 500. CliError::Launch still mislabels this request failure as a launcher error in the response and logs. Preserve the message text with a request-specific variant.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/src/server/mod.rs` around lines 1231 - 1235, Change the error
mapping around the subscriber delivery barrier spawned by spawn_blocking to use
the request-specific CliError variant instead of CliError::Launch, while
preserving the existing message text and inner FlowError propagation.
| #[test] | ||
| fn trajectory_boundaries_require_subscriber_delivery() { | ||
| let event = || SessionEvent { | ||
| session_id: "session-1".into(), | ||
| agent_kind: AgentKind::Codex, | ||
| event_name: "test".into(), | ||
| payload: json!({}), | ||
| metadata: json!({}), | ||
| }; | ||
|
|
||
| assert!(NormalizedEvent::TurnEnded(event()).finalizes_trajectory()); | ||
| assert!(NormalizedEvent::AgentEnded(event()).finalizes_trajectory()); | ||
| assert!(!NormalizedEvent::HookMark(event()).finalizes_trajectory()); | ||
| } |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win
Assert the terminal-but-not-finalizing variants.
is_terminal returns true for SubagentEnded, LlmEnded, and ToolEnded, but finalizes_trajectory must return false for them. That distinction is the intentional part of the contract, and no assertion protects it. A future change that routes the hook barrier through is_terminal would pass this test.
Add negative cases for those variants. SubagentEnded needs a SubagentEvent and LlmEnded/ToolEnded need their own payload types, so use the existing constructors in this test module.
Based on learnings from the path instructions: "Tests should cover the behavior promised by the changed API surface".
♻️ Suggested additional assertions
assert!(NormalizedEvent::TurnEnded(event()).finalizes_trajectory());
assert!(NormalizedEvent::AgentEnded(event()).finalizes_trajectory());
assert!(!NormalizedEvent::HookMark(event()).finalizes_trajectory());
+ // Terminal scopes that must not trigger the hook-response barrier.
+ assert!(!NormalizedEvent::Compaction(event()).finalizes_trajectory());
+ assert!(!NormalizedEvent::AgentStarted(event()).finalizes_trajectory());
}🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/tests/coverage/shared/events_tests.rs` around lines 17 - 30,
Extend trajectory_boundaries_require_subscriber_delivery with negative
assertions that NormalizedEvent::SubagentEnded, LlmEnded, and ToolEnded return
false from finalizes_trajectory. Construct each variant using the existing
test-module constructors and required payload types, while preserving the
current positive and HookMark assertions.
Source: Path instructions
| tokio::time::timeout(std::time::Duration::from_secs(1), started_rx) | ||
| .await | ||
| .expect("terminal sanitizer should observe the turn end") | ||
| .unwrap(); |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Raise the sanitizer-readiness timeout.
This 1-second bound must cover an HTTP round trip to the gateway, hook classification, apply_events, and the first scope-end dispatch after plugin activation. On a loaded CI runner that is tight, and expiry fails the test even when the code is correct.
The later 100 ms window at Line 1169 is sound, because release_tx fires only at Line 1175, so an early return there is a genuine defect. This readiness wait has no such property. Raise it to a few seconds.
🔧 Proposed change
- tokio::time::timeout(std::time::Duration::from_secs(1), started_rx)
+ tokio::time::timeout(std::time::Duration::from_secs(10), started_rx)
.await
.expect("terminal sanitizer should observe the turn end")
.unwrap();📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| tokio::time::timeout(std::time::Duration::from_secs(1), started_rx) | |
| .await | |
| .expect("terminal sanitizer should observe the turn end") | |
| .unwrap(); | |
| tokio::time::timeout(std::time::Duration::from_secs(10), started_rx) | |
| .await | |
| .expect("terminal sanitizer should observe the turn end") | |
| .unwrap(); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/tests/coverage/shared/server_tests.rs` around lines 1164 - 1167,
Increase the timeout used by the readiness wait around started_rx in the
sanitizer test from one second to a few seconds, while leaving the later 100 ms
release window unchanged because it still detects a genuine defect.
| let returned_early = early_response.is_some(); | ||
| let health = client.get(format!("{url}/healthz")).send().await.unwrap(); | ||
| assert_eq!(health.status(), StatusCode::OK); | ||
| release_tx.send(()).unwrap(); |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Do not unwrap the release send; it masks the failure this test exists to report.
If the Stop response returned early, the sanitizer future may already be dropped, and release_tx.send returns Err. This unwrap then panics here, before the run reaches the descriptive assertion at Line 1193 that names the actual defect. The failure output would point at a channel send instead of the response-ordering violation.
Ignore the send result.
🔧 Proposed change
- release_tx.send(()).unwrap();
+ // The receiver is gone if the response already returned; the ordering assertion below reports that.
+ let _ = release_tx.send(());📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| release_tx.send(()).unwrap(); | |
| // The receiver is gone if the response already returned; the ordering assertion below reports that. | |
| let _ = release_tx.send(()); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/tests/coverage/shared/server_tests.rs` at line 1175, Replace the
unwrap on release_tx.send in the surrounding test flow with an ignored send
result, allowing execution to reach the descriptive response-ordering assertion
when the receiver has already been dropped.
| .iter() | ||
| .find(|trajectory| atif_matches_session(trajectory, "codex-atif-boundary")) | ||
| .expect("the Stop response must not precede its ATIF snapshot"); | ||
| assert_eq!(trajectory["schema_version"], "ATIF-v1.7"); |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win
Drop the exact ATIF schema-version assertion.
This test verifies response-to-snapshot ordering. Pinning "ATIF-v1.7" couples it to ATIF schema evolution, so a schema bump fails a test that has nothing to do with the schema. The sibling test at Lines 1058-1062 asserts trajectory content without pinning a version.
Assert the property you care about, or remove the line.
♻️ Suggested change
- assert_eq!(trajectory["schema_version"], "ATIF-v1.7");
+ assert!(
+ trajectory["schema_version"].is_string(),
+ "expected an ATIF schema version, got {trajectory:?}"
+ );📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| assert_eq!(trajectory["schema_version"], "ATIF-v1.7"); | |
| assert!( | |
| trajectory["schema_version"].is_string(), | |
| "expected an ATIF schema version, got {trajectory:?}" | |
| ); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/cli/tests/coverage/shared/server_tests.rs` at line 1201, Remove the
exact "ATIF-v1.7" assertion from the response-to-snapshot ordering test, keeping
the test focused on ordering behavior and aligned with the sibling test’s
content-based assertions.
| /// Lifecycle boundaries use this narrower barrier when they need queued | ||
| /// observer work to be visible without waiting for unrelated managed | ||
| /// publications that are still in flight. | ||
| #[doc(hidden)] | ||
| pub fn flush_queued_subscribers() -> Result<()> { |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win
Document the callback prohibition on the newly public barrier.
docs/about-nemo-relay/concepts/subscribers.mdx states that a subscriber flush must not run inside a subscriber, event-sanitizer, guardrail, or intercept callback, because it does not establish a barrier there and can create a wait cycle. This function is now pub, so callers outside the crate can reach it, and #[doc(hidden)] keeps it out of rendered docs. The source comment is the only place a caller will read the constraint.
Add the constraint to the doc comment.
♻️ Suggested doc addition
/// Lifecycle boundaries use this narrower barrier when they need queued
/// observer work to be visible without waiting for unrelated managed
/// publications that are still in flight.
+///
+/// Do not call this from a subscriber, event-sanitizer, guardrail, or
+/// intercept callback. It does not establish a barrier there and can create a
+/// wait cycle.
#[doc(hidden)]
pub fn flush_queued_subscribers() -> Result<()> {📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| /// Lifecycle boundaries use this narrower barrier when they need queued | |
| /// observer work to be visible without waiting for unrelated managed | |
| /// publications that are still in flight. | |
| #[doc(hidden)] | |
| pub fn flush_queued_subscribers() -> Result<()> { | |
| /// Lifecycle boundaries use this narrower barrier when they need queued | |
| /// observer work to be visible without waiting for unrelated managed | |
| /// publications that are still in flight. | |
| /// | |
| /// Do not call this from a subscriber, event-sanitizer, guardrail, or | |
| /// intercept callback. It does not establish a barrier there and can create a | |
| /// wait cycle. | |
| #[doc(hidden)] | |
| pub fn flush_queued_subscribers() -> Result<()> { |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@crates/core/src/api/runtime/subscriber_dispatcher.rs` around lines 1265 -
1269, Update the doc comment for flush_queued_subscribers to explicitly state
that it must not be called from subscriber, event-sanitizer, guardrail, or
intercept callbacks, because it does not establish a barrier there and may
create a wait cycle.
| The coding-agent gateway adds a targeted barrier at trajectory boundaries. | ||
| Before the gateway acknowledges a hook that closes an agent or turn scope, it | ||
| waits for subscriber callbacks queued through that boundary. When a local ATIF | ||
| write succeeds, the trajectory file is therefore visible when the hook response | ||
| returns. Other hook events and native runtime calls remain asynchronous. | ||
|
|
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Move this paragraph; it separates the table lead-in from the table.
Lines 145-148 end with a colon: "Use the following milestones to choose the barrier that matches the output you need to observe:". The table at Line 156 is what that sentence introduces. This new paragraph now sits between them, so the lead-in no longer reads into the table.
Place the gateway paragraph after the table instead, where it qualifies the milestones it depends on. The content itself is accurate and matches apply_hook_outcome in crates/cli/src/server/mod.rs.
As per coding guidelines: "Introduce code blocks, lists, tables, and images with complete sentences" and "do not let a code block complete or interrupt the grammar of surrounding prose".
📝 Proposed reordering
following milestones to choose the barrier that matches the output you need to
observe:
-The coding-agent gateway adds a targeted barrier at trajectory boundaries.
-Before the gateway acknowledges a hook that closes an agent or turn scope, it
-waits for subscriber callbacks queued through that boundary. When a local ATIF
-write succeeds, the trajectory file is therefore visible when the hook response
-returns. Other hook events and native runtime calls remain asynchronous.
-
| Milestone | What It Establishes | What It Does Not Establish |
|---|---|---|Then insert the paragraph after the table's final row, before the "Use the subscriber flush API" paragraph:
The coding-agent gateway adds a targeted barrier at trajectory boundaries.
Before the gateway acknowledges a hook that closes an agent or turn scope, it
waits for subscriber callbacks queued through that boundary. When a local ATIF
write succeeds, the trajectory file is therefore visible when the hook response
returns. Other hook events and native runtime calls remain asynchronous.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@docs/about-nemo-relay/concepts/subscribers.mdx` around lines 150 - 155, Move
the gateway barrier paragraph from before the milestones table to after the
table’s final row and before the “Use the subscriber flush API” paragraph. Keep
the paragraph text unchanged so the table lead-in directly introduces the table.
Source: Coding guidelines
| // Agent and turn scope-end subscribers write the final trajectory snapshot. Use the | ||
| // queued-only barrier so an active managed publication in another session cannot hold the | ||
| // hook response open. | ||
| tokio::task::spawn_blocking(nemo_relay::api::runtime::flush_queued_subscribers) |
There was a problem hiding this comment.
I don't like this. we should only be flushing subscribers for the session being closed.
There may be one instance of Relay running per user, which means that while multiple sessions with Codex and/or Claude could be running, you'd be potentially deadlocking.
2 participants
Overview
Ensure coding-agent terminal hooks do not acknowledge completion until subscriber callbacks queued through that trajectory boundary have run. This fixes the race where a downstream evaluator snapshots Relay output immediately after the agent exits and can observe ATOF while the ATIF subscriber is still writing the trajectory.
Details
AgentEndedandTurnEndedas trajectory-finalizing events and route Codex, Claude Code, and Hermes hook outcomes through one lifecycle-aware handler.Validation:
cargo nextest run --workspace --profile ci --no-fail-fast: 3,852 passed, 0 skipped.just test-python: 639 passed.just test-go: passed.just test-node: 353 passed on the supported Node 24 toolchain.just docs: passed; Fern reported only the expected unauthenticated redirect-check warning.uv run pre-commit run --all-files: passed.For downstream compatibility validation only, the production diff was built as Relay 0.7.0 and exercised with Fabric's schema-v3/Relay-0.7 compatibility work, which contains no ATIF polling or sleep. This PR itself targets
main; it does not request a release-branch hotfix.Where should the reviewer start?
Start with
apply_hook_outcomeincrates/cli/src/server/mod.rs, then reviewcodex_stop_response_waits_for_atif_finalizationincrates/cli/tests/coverage/shared/server_tests.rs. The key design choice is the queued-only barrier: terminal hook responses gain a delivery guarantee without turning normal hooks or unrelated in-flight sessions into a global synchronous path.Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)
Summary by CodeRabbit
New Features
Bug Fixes
Documentation