Skip to content

docs(rfc): RFC-010 verification-comment current-state fixups - #215

Merged
aaltshuler merged 1 commit into
mainfrom
docs/rfc-010-verification-fixups
Jun 13, 2026
Merged

docs(rfc): RFC-010 verification-comment current-state fixups#215
aaltshuler merged 1 commit into
mainfrom
docs/rfc-010-verification-fixups

Conversation

@aaltshuler

@aaltshuler aaltshuler commented Jun 13, 2026

Copy link
Copy Markdown
Contributor

Applies the Verification comments (Codex) on RFC-010 — current-state accuracy fixups found against the now-merged 3a/3b/3c code. Kept verbatim with per-point Resolution (accepted) notes.

  • graphs list marked remote-only today in the current-state table (embedded arm bails; rides GraphClient only to share the resolver).
  • init noted as positional-URI-only today (no --target) — adding --target is part of the proposal, entangled with the initcluster apply signpost, not current state.
  • Validated-fact Add OpenAPI documentation endpoint and schema #1 rephrased to the post-collapse reality (GraphClient::resolve*; only the two factories call apply_server_flag) — drops the stale "16 call sites" count.
  • Authority rule gains a flag-shape caveat: --graph is already a global flag requiring --server, so the cluster-managed resolver + its flag shape are deferred to a later slice; the illustrative --cluster <dir> --graph <id> spelling is marked not-final.

Docs-only. This lands before the RFC-010 Slice 1 implementation so the RFC's current-state is accurate when the code starts.

🤖 Generated with Claude Code

Greptile Summary

This docs-only PR applies four current-state accuracy fixups to RFC-010 against the now-merged RFC-009 Phases 3a–3c code, each paired with a "Resolution (accepted)" note appended as a Verification comments appendix section.

  • graphs list / init table corrections: The current-state plane table now marks graphs list as remote-only today (with an explanatory note on its GraphClient routing) and adds an inline caveat that init accepts only a required positional URI today — neither --target nor config aliases.
  • Validated-fact rewrite: The stale "16 call sites through apply_server_flag" description is replaced with the post-collapse reality (GraphClient::resolve*; only the two factories call apply_server_flag).
  • Authority rule flag-shape deferral: The proposed --cluster <dir> --graph <id> addressing for cluster-managed maintenance is now explicitly marked "for illustration" and the whole cluster-managed resolver is deferred to a later slice, because --graph is already a global flag requiring --server with a different meaning that clap cannot disambiguate.

Confidence Score: 4/5

Safe to merge — docs-only change with no code or behavior impact.

All four fixups are accurate and well-reasoned. The one issue is a cross-section inconsistency: the authority rule now marks --cluster <dir> --graph <id> as not-final and deferred, but the target ergonomics code block still presents that same flag shape as a clean example with no matching qualification. An implementer reading the ergonomics section first would see what looks like a settled flag contract.

The target ergonomics section of docs/dev/rfc-010-cli-planes-restructure.md (the --cluster ./cluster --graph knowledge code block, ~line 147) should carry a caveat consistent with the flag-shape deferral note added to the authority rule.

Important Files Changed

Filename Overview
docs/dev/rfc-010-cli-planes-restructure.md Docs-only RFC fixup: updates the current-state table to mark graphs list as remote-only, calls out init's positional-URI-only addressing today, replaces the stale "16 call sites" count with the post-collapse GraphClient::resolve* description, and adds a flag-shape deferral caveat to the Authority rule. One cross-section inconsistency: the new authority-rule caveat marks --cluster <dir> --graph <id> as not-final/deferred, but the target ergonomics code block still presents that same flag shape as a clean end-state example without qualification.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[CLI invocation] --> B{Which plane?}

    B -->|Data verb| C[GraphClient resolver]
    C --> D{--server/--graph or\n--target/positional URI?}
    D -->|--server prod --graph id| E[GraphClient::resolve*\napply_server_flag in factory]
    D -->|--target / URI| F[Embedded engine\nor HTTP per config]
    E --> G[Remote HTTP execution]
    F --> H[Embedded or Remote\nper target resolution]

    B -->|Storage/Maintenance verb| I[resolve_uri /\nresolve_local_uri]
    I --> J{Addressing form?}
    J -->|positional URI| K[Direct storage\nfile:// or s3://]
    J -->|--target| L{Resolves to?}
    L -->|direct storage| K
    L -->|remote http://| M[Loud error:\nstorage-plane needs\ndirect storage]
    J -->|init today| N[Required positional URI only\nno --target today]
    K --> O[Embedded engine\non storage]

    B -->|Control verb| P[cluster directory\nreader]
    P --> Q[--config dir\nfile:// or s3://]

    style M fill:#f88,stroke:#c00
    style N fill:#ffd,stroke:#aa0
Loading

Comments Outside Diff (1)

  1. docs/dev/rfc-010-cli-planes-restructure.md, line 143-148 (link)

    P2 Target-ergonomics example inconsistent with new flag-shape caveat

    The PR adds a flag-shape caveat in the Authority rule section (lines 213–220) that explicitly marks --cluster <dir> --graph <id> as "for illustration" and defers the cluster-managed resolver to a later slice. The target ergonomics section below (this code block plus the error-message example on ~line 140) still presents omnigraph optimize --cluster ./cluster --graph knowledge and --cluster <dir> --graph <id> as committed end-state shapes without any matching qualification. An implementer of Slice 1 who reads the ergonomics section first will see what looks like a settled flag contract, then find the deferral signal only much later in the authority rule. Given AGENTS.md rule 6 ("don't lie — replace wrong text rather than leaving it silently incorrect"), a brief inline note on this block mirroring the authority-rule caveat would make the RFC self-consistent.

    Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

    Fix in Claude Code

Fix All in Claude Code

Reviews (1): Last reviewed commit: "docs(rfc): RFC-010 — apply verification-..." | Re-trigger Greptile

Folds in the Codex verification review (kept verbatim with per-point
Resolution notes):

- `graphs list` is marked remote-only today in the current-state table
  (the embedded arm bails; it rides GraphClient only to share the resolver).
- `init` is noted as positional-URI-only today (no `--target`); adding
  `--target` to init is part of the proposal, entangled with the
  init→cluster apply signpost, not current state.
- Validated-fact #1 now describes the post-collapse reality
  (`GraphClient::resolve*`; only the two factories call `apply_server_flag`),
  dropping the stale "16 call sites" count.
- The Authority rule carries a flag-shape caveat: `--graph` is already a
  global flag requiring `--server`, so the cluster-managed resolver and its
  flag shape are deferred to a later slice; the illustrative
  `--cluster <dir> --graph <id>` spelling is marked not-final.

Docs-only; no code change.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@aaltshuler
aaltshuler requested a review from ragnorc as a code owner June 13, 2026 18:56

@greptile-apps greptile-apps Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

aaltshuler has reached the 50-review limit for trial accounts. To continue receiving code reviews, upgrade your plan.

@aaltshuler
aaltshuler merged commit 2ddb88f into main Jun 13, 2026
8 checks passed
@aaltshuler
aaltshuler deleted the docs/rfc-010-verification-fixups branch June 13, 2026 19:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant