Claude/convoy readme installer#12
Merged
Merged
Conversation
…plugin Author stage — less rigid: - fallbackDockerfile now uses every relevant scan signal: scan.port (no more hardcoded 8080), scan.packageManager (npm ci vs pnpm frozen-lockfile vs yarn frozen-lockfile vs bun), scan.buildCommand, scan.startCommand, scan.dataLayer (adds prisma generate when Prisma is detected), scan.framework (Next.js gets a proper 3-stage standalone Dockerfile instead of the generic node template). - Node base image bumped node:20 → node:22 (Node 20 EOL Oct 2026). - Python path picks the right installer for poetry/uv/pip and uses the detected framework for the start command (uvicorn/gunicorn/python). - Each non-node template now uses scan.port. - All of this degrades gracefully when signals are absent — the enricher still overwrites with a fully tailored Dockerfile when a key is set, but the deterministic fallback is now production-quality for common stacks without enrichment. Plugin commands — platformization story: - ship.md now leads with a platform-selection table and explains WHY Convoy picks a platform (evidence-based scoring, not guessing). Added an explicit b2dp/ship-to-vps handoff paragraph — agents and operators coming from those skills now know exactly what to do next. - where.md PLATFORM READINESS section added: checks flyctl, vercel, railway, gcloud, and ssh auth at orientation time so the operator knows which platforms are ready to receive a deploy before they plan. docs/integration.md (new): Documents the full b2dp → ship-to-vps → Convoy chain, the platform scoring table, how to register a post-ship-to-vps repo with Convoy, and the current skill-update status (patches in docs/skill-updates/ need to be applied to teckedd-code2save/ai-build-tools to fix Prisma 7 and Caddy breakages in those external skills). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
closes #4, closes #8 ## demo-app: Meridian concurrency scenario (issue #4) Adds `src/routes/render.ts` — a POST /render route that simulates a PDF renderer with a global serialisation lock. In DEMO_MODE=concurrent each request acquires `renderLock` before processing, serialising all renders through a single queue. With 30 concurrent requests and a 300ms render time, the last request waits 9 seconds — p99 blows past any threshold while error rate stays at 0%. The bug is intentionally legible: a module-level `let renderLock` promise chain with a clear TODO comment. The medic can find it by reading render.ts and identify the lock as the root cause. Also adds --inject-failure=concurrency to the Convoy CLI, which uses scripted fixture logs (render_lock_acquired entries with linearly stacking waited_ms) to trigger the medic without needing a real rehearsal run. Useful for demos without full service setup. ## monorepo: service-relative file paths (issue #8) The author stage was writing all files to the repo root regardless of which lane they belonged to. In a monorepo with apps/api and apps/web, both lanes would author a `Dockerfile` at the root — the second one silently overwrote the first via deduplication. Fix: draftAuthorSection now accepts a servicePath parameter (default '.') and applies it as a prefix to all authored file paths. A lane at `apps/api` produces `apps/api/Dockerfile`, `apps/api/fly.toml`, etc. The manifest path is also prefixed to `apps/api/.convoy/manifest.yaml`. The buildLane caller in planner/index.ts now passes node.path through. Cross-lane deduplication still runs but will no longer encounter path collisions since each lane's files have unique service-scoped paths. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…e_pr closes #5, closes #6 ## Platform scoring evidence (issue #5) PlanPlatformCandidate gains an `adjustments: [{delta, label}]` field. picker.ts now records each individual score delta with a human-readable label via an `adj(delta, label)` helper instead of accumulating reasons as a flat string. The top 3 adjustments by |delta| are the signal. CLI plan output (`renderPlan`) replaces the flat score ranking with a table showing each candidate's score and its top two adjustments: Platform selection fly chosen (scored, score 90) ● fly 90 +25 background worker topology, +15 container-native · railway 70 +20 managed Postgres in one click · vercel 25 -25 background workers not supported Web plan detail page renders the same adjustments inline in the candidate grid cards: coloured deltas (+N green / −N red) with the adjustment label, replacing the flat reason string. No data-fetching change — adjustments come from the plan JSON already in the page. ## Rehearsal evidence on merge_pr card (issue #6) The `merge_pr` approval was missing the `rehearsal` summary that `open_pr` already carries. Added `rehearsal: summarizeRehearsalForApproval(ctx.prior['rehearse'])` to the merge_pr awaitApproval call in AuthorStage.#runReal. MergePrApprovalCard in the web UI now reads `summary['rehearsal']` and renders `<RehearsalEvidence>` (already existed for open_pr) showing healthy/breached status, p99, error rate, duration, and smoke test count. In scripted mode the rehearsal field is absent and no fallback warning is shown (scripted runs have no real rehearsal data). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…n story Issues #9, #10, #11 from the sprint. #9 — plan detail page as handoff document - LaneSection now shows per-lane scan risks (block/warn/info with colour badges) and required secrets checklist (staged ✓ / already-set / missing with amber pulse). Secrets hint links to `convoy stage-secrets` for missing keys. - web/lib/plans.ts: add `adjustments` to PlanPlatformCandidate (removes the unsafe cast the plan page was doing), add `risks`/`healthPath`/`port` to the lane.scan type so the new fields are type-safe. #10 — demo script (docs/demo-script.md) - 3-min pitch narrative (spoken, no demo) explaining the core problem - 10-minute live walkthrough: Act 1 Plan → Act 2 Concurrency breach (--inject-failure=concurrency) → Act 3 Clean run → Act 4 Secrets gate → Act 5 The handoff - Objection-handling section for CI/CD, Terraform, trust, and platform choice questions #11 — README Meridian story - New "The Meridian scenario" section before "See it in action" - Tells the complete story: 3-service fintech, principal engineer leaving, concurrency bug caught in rehearsal, medic diagnosis, fix-and-resume, secrets gate, canary, 8-min total with zero tribal knowledge required - Replaces "here's what the product does" list with a concrete narrative Co-Authored-By: Claude Opus 4.7 (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
2 participants
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.
No description provided.