arch-review: provider seams and feature delivery cost #52

Open
opened 2026-10-04 16:09:18 +00:00 by kayg · 5 comments
Owner

Owner-requested read-only architecture and delivery-cost review. Worktree /mnt/hdd/infraternal/wt-lp/arch-review; detached at assigned dev base d1eefa9a1b. Product code will remain unchanged. Report: /mnt/hdd/infraternal/jobs/arch-review/report.md. Scope: provider seams, complexity/churn, gate and build costs, ranked recommendations, first-release product needs.

Owner-requested read-only architecture and delivery-cost review. Worktree /mnt/hdd/infraternal/wt-lp/arch-review; detached at assigned dev base d1eefa9a1bbd68b1ff75fdfb394aebb094d64cd7. Product code will remain unchanged. Report: /mnt/hdd/infraternal/jobs/arch-review/report.md. Scope: provider seams, complexity/churn, gate and build costs, ranked recommendations, first-release product needs.
Author
Owner

Started: assigned detached worktree /mnt/hdd/infraternal/wt-lp/arch-review at d1eefa9a1b. No product edits or pushes. Code-review-graph was absent and was rebuilt for orientation. Recent git log --stat and churn evidence will accompany the report. Build timings use this worktree own target directory with two Cargo jobs; observed VM load1 was 18.41, so these are loaded-dev-VM diagnostics, not reportable product performance.

Started: assigned detached worktree /mnt/hdd/infraternal/wt-lp/arch-review at d1eefa9a1bbd68b1ff75fdfb394aebb094d64cd7. No product edits or pushes. Code-review-graph was absent and was rebuilt for orientation. Recent git log --stat and churn evidence will accompany the report. Build timings use this worktree own target directory with two Cargo jobs; observed VM load1 was 18.41, so these are loaded-dev-VM diagnostics, not reportable product performance.
Author
Owner

Architecture findings at d1eefa9:

  • New factories exist, but default startup calls LegacyProvider::new through the Engine facade (crates/agent/src/lib.rs:1319-1337). FactoryRegistry::new and connect_configured are invoked only by tests in the examined tree. Public routes, Scope, CLI/MCP and OPEN translation remain Docker/Podman-specific (server/resources.rs:218; proto/permissions.rs:11; host_route.rs:574; host-resource-client/wire.rs:154).
  • Recent reachable non-merge history (2026-09-20 through 2026-10-04) contains 1,135 commits. browser-check.mjs appears in 282; 208 touch only that file. contracts.md appears in 150. Installation-core file grouping has 24 commits across 33 files; feature grouping is path-based and overlaps. Raw git log --stat and exact membership are saved with the report.
  • Historical issue #31 records fast at 844.674 s with load5=14.47 on eight CPUs: clippy 204.978 s, tests 344.892 s, OpenAPI/client drift 170.754 s. This is existing issue evidence, not a gate run by this review. Merge includes fast and additionally release and real-service smokes; therefore that run alone uses 93.9% of its 900 s target.
  • Current own-target diagnostic: cold provider-api build PASS 51.221 s, warm PASS 0.327 s, jobs=2 at load1 19.8-22.5. Server cold build remains in progress. These are build-cost measurements, not reportable product performance.
Architecture findings at d1eefa9: - New factories exist, but default startup calls LegacyProvider::new through the Engine facade (crates/agent/src/lib.rs:1319-1337). FactoryRegistry::new and connect_configured are invoked only by tests in the examined tree. Public routes, Scope, CLI/MCP and OPEN translation remain Docker/Podman-specific (server/resources.rs:218; proto/permissions.rs:11; host_route.rs:574; host-resource-client/wire.rs:154). - Recent reachable non-merge history (2026-09-20 through 2026-10-04) contains 1,135 commits. browser-check.mjs appears in 282; 208 touch only that file. contracts.md appears in 150. Installation-core file grouping has 24 commits across 33 files; feature grouping is path-based and overlaps. Raw git log --stat and exact membership are saved with the report. - Historical issue #31 records fast at 844.674 s with load5=14.47 on eight CPUs: clippy 204.978 s, tests 344.892 s, OpenAPI/client drift 170.754 s. This is existing issue evidence, not a gate run by this review. Merge includes fast and additionally release and real-service smokes; therefore that run alone uses 93.9% of its 900 s target. - Current own-target diagnostic: cold provider-api build PASS 51.221 s, warm PASS 0.327 s, jobs=2 at load1 19.8-22.5. Server cold build remains in progress. These are build-cost measurements, not reportable product performance.
Author
Owner

Architecture review complete at unchanged detached head d1eefa9a1bbd68b1ff75fdfb394aebb094d64cd7 (assigned dev base). Report: /mnt/hdd/infraternal/jobs/arch-review/report.md.

No product source edits, commits, branch changes or push. Worktree status is clean. The supply-chain gate rewrote THIRD_PARTY_NOTICES.md; its diff was saved externally and only that gate-produced edit was restored. Incremental probes checked unchanged source bytes and restored timestamps.

Findings: production still connects a closed Docker/Podman legacy provider list rather than configured FactoryRegistry; generic public resources, permissions, OPEN and clients remain coupled. Remote and embedded native operations already share the runtime, so the recommendation is to finish the migration rather than invent another plugin system. Contracts have 2,869 lines / 26,219 words, with stale status claims in contracts and PROVIDERS.md. Browser harness changed in 282 of 1,135 recent non-merge commits (208 harness-only commits); installation core changed in 24 commits across 33 files, including 14 fix subjects. These are churn counts, not person-hours or measured review-round counts.

Current fast PASS: 864.038 seconds, jobs=2, 8 CPUs, load5 13.09 to 10.74. Clippy/test/drift total 791.011 seconds (91.6%); fast consumes 96% of the 15-minute merge budget before additional merge work. A timestamp-invalidated server library build passed in 3.208 seconds and provider-api in 1.365 seconds. A post-fast server build took 96.007 seconds with dependency/profile rebuilds and sccache fallback. Missing-web cold server attempt failed after 658.361 seconds; reported explicitly as failure, not a successful cold build. Artifacts preserve all timings and logs. No semantic edit, final executable link comparison, merge/nightly/real-service/browser run or perf-VM A/B is claimed.

Priorities and cost estimates: serialize gates (0.5–1d); acceptance matrix plus bounded review (0.5d); two deterministic flake repairs (≤1d pilot); contracts navigation/test index (0.5–1d); owner scope decision, Docker/Podman/SSH fleet first (≤0.5d); production factory migration (3–5d); generic resource vertical slice (5–10d); session-authority/lifecycle ownership (4–7d); lightweight schema generation/typed stream and API bindings (2–4d); feature-off dependency forwarding (1–2d after compatibility migration). Security simplifications are proposals with explicit contract/proof requirements, preserving signed dispatch, mTLS and host-key verification. Native NetBird/Tailscale SSH authentication needs real compatibility work; current SSH client calls public-key auth only.

Exact gate output:

WARN load5=13.09 exceeds online CPU count=8; timing-sensitive checks may be delayed
PASS fast: migrations are append-only (0.270s)
PASS fast: workspace package dependencies (0.451s)
PASS fast: gate target-directory and launch diagnostics tests (0.333s)
PASS fast: JavaScript dependencies (5.794s)
PASS fast: advisory review policy tests (0.189s)
PASS fast: supply-chain checks (7.426s)
PASS fast: build web app (10.745s)
PASS fast: cargo fmt --check (2.505s)
PASS fast: cargo clippy (209.587s)
PASS fast: cargo test (479.623s)
PASS fast: svelte-check (13.552s)
PASS fast: UI package svelte-check (5.700s)
PASS fast: stream package typecheck (5.814s)
PASS fast: Vitest (6.821s)
PASS fast: UI package Vitest (5.918s)
PASS fast: stream package Vitest (1.673s)
PASS fast: OpenAPI and TypeScript client drift (101.801s)
PASS fast: terminal TypeScript (3.284s)
PASS fast: terminal Vitest (2.340s)
Architecture review complete at unchanged detached head `d1eefa9a1bbd68b1ff75fdfb394aebb094d64cd7` (assigned dev base). Report: `/mnt/hdd/infraternal/jobs/arch-review/report.md`. No product source edits, commits, branch changes or push. Worktree status is clean. The supply-chain gate rewrote THIRD_PARTY_NOTICES.md; its diff was saved externally and only that gate-produced edit was restored. Incremental probes checked unchanged source bytes and restored timestamps. Findings: production still connects a closed Docker/Podman legacy provider list rather than configured FactoryRegistry; generic public resources, permissions, OPEN and clients remain coupled. Remote and embedded native operations already share the runtime, so the recommendation is to finish the migration rather than invent another plugin system. Contracts have 2,869 lines / 26,219 words, with stale status claims in contracts and PROVIDERS.md. Browser harness changed in 282 of 1,135 recent non-merge commits (208 harness-only commits); installation core changed in 24 commits across 33 files, including 14 fix subjects. These are churn counts, not person-hours or measured review-round counts. Current fast PASS: 864.038 seconds, jobs=2, 8 CPUs, load5 13.09 to 10.74. Clippy/test/drift total 791.011 seconds (91.6%); fast consumes 96% of the 15-minute merge budget before additional merge work. A timestamp-invalidated server library build passed in 3.208 seconds and provider-api in 1.365 seconds. A post-fast server build took 96.007 seconds with dependency/profile rebuilds and sccache fallback. Missing-web cold server attempt failed after 658.361 seconds; reported explicitly as failure, not a successful cold build. Artifacts preserve all timings and logs. No semantic edit, final executable link comparison, merge/nightly/real-service/browser run or perf-VM A/B is claimed. Priorities and cost estimates: serialize gates (0.5–1d); acceptance matrix plus bounded review (0.5d); two deterministic flake repairs (≤1d pilot); contracts navigation/test index (0.5–1d); owner scope decision, Docker/Podman/SSH fleet first (≤0.5d); production factory migration (3–5d); generic resource vertical slice (5–10d); session-authority/lifecycle ownership (4–7d); lightweight schema generation/typed stream and API bindings (2–4d); feature-off dependency forwarding (1–2d after compatibility migration). Security simplifications are proposals with explicit contract/proof requirements, preserving signed dispatch, mTLS and host-key verification. Native NetBird/Tailscale SSH authentication needs real compatibility work; current SSH client calls public-key auth only. Exact gate output: ```text WARN load5=13.09 exceeds online CPU count=8; timing-sensitive checks may be delayed PASS fast: migrations are append-only (0.270s) PASS fast: workspace package dependencies (0.451s) PASS fast: gate target-directory and launch diagnostics tests (0.333s) PASS fast: JavaScript dependencies (5.794s) PASS fast: advisory review policy tests (0.189s) PASS fast: supply-chain checks (7.426s) PASS fast: build web app (10.745s) PASS fast: cargo fmt --check (2.505s) PASS fast: cargo clippy (209.587s) PASS fast: cargo test (479.623s) PASS fast: svelte-check (13.552s) PASS fast: UI package svelte-check (5.700s) PASS fast: stream package typecheck (5.814s) PASS fast: Vitest (6.821s) PASS fast: UI package Vitest (5.918s) PASS fast: stream package Vitest (1.673s) PASS fast: OpenAPI and TypeScript client drift (101.801s) PASS fast: terminal TypeScript (3.284s) PASS fast: terminal Vitest (2.340s) ```
Author
Owner

Starting issue filing for the architecture review. Worktree: /mnt/hdd/infraternal/wt-lp/issues-arch, detached at dev head 0b4fc37d55. Source report: /mnt/hdd/infraternal/jobs/arch-review/report.md. I will verify actionable claims against this checkout, search for duplicates, and create or update the matching issues. Product code remains read-only.

Starting issue filing for the architecture review. Worktree: /mnt/hdd/infraternal/wt-lp/issues-arch, detached at dev head 0b4fc37d555efacf9d5ca731451cfa0738ed4fb3. Source report: /mnt/hdd/infraternal/jobs/arch-review/report.md. I will verify actionable claims against this checkout, search for duplicates, and create or update the matching issues. Product code remains read-only.
Author
Owner

Finished issues-arch at detached HEAD 0b4fc37d555efacf9d5ca731451cfa0738ed4fb3. The product worktree is clean; no source edits, commits or pushes.

Filed 23 actionable issues, #53–#75. Their bodies include current code evidence, the recommendation, payoff/cost, priority, the owner-decision relationship, the report path and the required source line. Added the report’s timing/cleanup/browser evidence to existing issues #31, #42 and #44, and its local-build-profile recommendation to performance-audit issue #51.

Decision checks:

  • The report’s request to decide between completing bootstrap and a manual-enrollment-first cut was already settled by DECISIONS §30 and the owner’s 2026-10-04 decision. #72 tracks completing the approved paths.
  • The proposed contract requirements-to-test index conflicts with the owner’s code/tests-as-spec decision. #55 instead tracks invariant-focused contract cleanup and stale-reference fixes.
  • The report does not recommend replacing the SQL revocation fence; no issue was filed to move it in memory.
  • The custom signed tunnel and already-shared embedded/remote runtime are retained. Future provider work stays conditional on the deferred PLAN steps.

Gate output: scripts/verify was not run for this read-only issue-filing job; there is no product diff to verify.

Source: arch-review 2026-10-04; /mnt/hdd/infraternal/jobs/arch-review/report.md.

Finished issues-arch at detached HEAD `0b4fc37d555efacf9d5ca731451cfa0738ed4fb3`. The product worktree is clean; no source edits, commits or pushes. Filed 23 actionable issues, #53–#75. Their bodies include current code evidence, the recommendation, payoff/cost, priority, the owner-decision relationship, the report path and the required source line. Added the report’s timing/cleanup/browser evidence to existing issues #31, #42 and #44, and its local-build-profile recommendation to performance-audit issue #51. Decision checks: - The report’s request to decide between completing bootstrap and a manual-enrollment-first cut was already settled by DECISIONS §30 and the owner’s 2026-10-04 decision. #72 tracks completing the approved paths. - The proposed contract requirements-to-test index conflicts with the owner’s code/tests-as-spec decision. #55 instead tracks invariant-focused contract cleanup and stale-reference fixes. - The report does not recommend replacing the SQL revocation fence; no issue was filed to move it in memory. - The custom signed tunnel and already-shared embedded/remote runtime are retained. Future provider work stays conditional on the deferred PLAN steps. Gate output: `scripts/verify` was not run for this read-only issue-filing job; there is no product diff to verify. Source: arch-review 2026-10-04; /mnt/hdd/infraternal/jobs/arch-review/report.md.
Commenting is not possible because the repository is archived.
No labels
No milestone
No project
No assignees
1 participant
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/infraternal-v0#52
No description provided.