feat(ai): AutoPublishCrew — in-process SEO generation crew (AI-042)#337
Merged
Conversation
Admin-triggered crew path that runs the AI-041 specialists (researcher→ drafter→critic→editor) over ILlmService to generate Edition SEO prose (Description + SeoRelevanceText). Per-field 4-stage CrewPlan, persists each as crew.autopublish agent_run, fail-closed review gate, writes the Edition only when BOTH fields pass clean. Edition stays Draft (no auto-publish). Legacy bash+Claude-CLI poller untouched and default. New in-process path rather than swapping the poller: legacy gen is out-of-process (Claude CLI, Max sub, $0) while the crew is metered nano — keep both, A/B in AI-046 before any cutover. Data-loss guards (per QA): - gate flags NeedsReview on empty/whitespace OR below-MinLength editor output — never blanks a real field - honors SeoSource.Manual: a hand-written Description/Relevance is never overwritten (returns manualProtected), matching legacy backfill contract CostCapUsd 0.02/field; autopublish.crew rate limit (4/min). 22 unit tests (fake ILlmService + recording IAgentRunWriter), no DB/network. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
AI-042 —
AutoPublishCrewintegration (Phase 7)Admin-triggered, in-process crew that runs the AI-041 specialists (researcher→drafter→critic→editor) over
ILlmServiceto generate an Edition's SEO prose (Description+SeoRelevanceText). Per-field 4-stageCrewPlan, persists each run as acrew.autopublishagent_run, applies a fail-closed review gate, and writes the Edition only when both fields pass clean. Edition stays Draft — no auto-publish.Why a new path, not a swap
Legacy AutoPublish gen is out-of-process (systemd bash poller → Claude CLI, Max sub, effectively $0). The crew is in-process metered nano. So "replace the single call" isn't a one-liner — this adds a parallel, observable path; the bash poller stays intact and default. A/B (AI-046) decides any cutover.
Endpoint
POST /admin/autopublish/editions/{id}/crew-generate(admin auth,autopublish.crewrate limit 4/min). Loads source material mirroringseo-generate.sh(title, authors, language, first-chapter excerpt), runs the crew twice, gates, writes back. Returns bothrunIds + per-field critique summaries +needsReview/manualProtectedso AI-045's transcript UI can fetch later.Fail-closed gate + data-loss guards (hardened per adversarial QA)
NeedsReview= true unless statuscompletedAND critic parsed AND noblockerissue.MinLengtheditor text → flagged, never blanks a real field.Description/Relevanceon aSeoSource.ManualEdition is never overwritten (manualProtected: true) — matches the legacy backfill contract.Cost
CostCapUsd = 0.02/field (4 nano calls). ~$0.04/edition worst case.Tests — 22 unit (full suite 377 green)
Fake
ILlmServicerouted perFeatureTag+ recordingIAgentRunWriter, no DB/network: clean→write, blocker→flag, ParseFailed→flag, empty/whitespace/below-MinLength→flag, cost-cap halt→flag, persistence (crew.autopublish, editionId, nested sub-agent steps), manual-protection battery. StudyBuddy tool set-equality still green (noIToolleaked).Verify
dotnet test tests/TextStack.UnitTests→ 377 passdotnet test tests/TextStack.AiEvals→ 24 pass / 5 skip (no-key floor)dotnet format --verify-no-changes→ cleanLegacy
seo-publish-poll.sh/seo-generate.shunmodified. No DB migration.🤖 Generated with Claude Code