- services/atomizer: standalone Go module implementing §5.1 — Anthropic
OpenAI-compatible chat completions by default, native /v1/messages when
LLM_API_STYLE=anthropic, strict-JSON prompting, defensive fence-stripping
parse with one retry, deterministic equal-split fallback without an API
key, exact coefficient-sum normalization, bearer auth
- services/work-performer: Node 20 http server implementing §5.2 — single
concurrency, /work/{jobId}/TASK.md preparation, attachment downloads,
optional shallow git clone, claude CLI execution with JSON output,
simulated success when the CLI is unavailable (offline demo), artifact
endpoint, idempotent HMAC-signed callbacks with retry, best-effort cancel
- compose profile 'mocks': separate builds/ports/tokens, healthchecks,
${HOME}/.claude(.json) mounted read-only into the performer
- contract tests (go test -tags=contract) for both services; Makefile
test-contract target; verified live incl. a real Claude Code job run
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- aggregations over the immutable bountyAwards ledger: totals, weekly
buckets ($dateTrunc), per-developer and per-customer groupings
- timeline-derived: approval rate, time logged, assigned→approved lead
time, imported→published atomization lead time, open board depth
- developer + consultant dashboards (admin = global consultant view),
date-range filters, CSV export of the ledger
- leaderboard (top 10 by bounty) honoring the new leaderboardOptOut
profile setting
- hand-rolled SVG line and bar charts (~150 lines, no chart library)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- task domain model + exhaustive §4.4 status machine tests (single source of truth)
- per-customer polling workers reconciled every 30s from the customers
collection; panic-safe runs; manual sync-now trigger; status provider
for the admin panel
- §5.3 upsert keyed on (system, key, customerId): refreshes content while
status=imported, records upstream_changed timeline + notification after
- attachments cached into GridFS via connector-authorized downloads
- orphaned tickets flagged with timeline + consultant notification, never
deleted; idempotent across polls
- consultant identity from users.extra.ticketingIdentities per system
- notifications store (insert/list/unread-count/mark-read)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- customer CRUD wizard backend: per-system credential shapes validated via
connector construction, AES-256-GCM at rest, write-only over the API
- ticketing connectors (jira/azure_devops/youtrack/demo) with test-connection
- user management: roles, disable (revokes sessions), force password reset,
delete; self-demotion/disable/delete guards
- admin-editable runtime settings ({_id:app}) incl. atomizer URL override
- audit log written on every privileged mutation + filterable list API
- service status panel: mongo/atomizer/work-performer health + latency with
late-bound breaker, sync and jobs status providers
- tabbed admin UI (customers wizard dialog, users table, settings, status, audit)
- compose: mongo nofile ulimit 64000 (1024 default crashed WiredTiger)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- embedded Go templates + vanilla ES-module JS + hand-written CSS, no build step
- exact §10 beige light / dark design tokens, square edges (radius 2px),
system font stack, visible focus rings
- theme toggle persisted to localStorage and the user profile
- login/register/change-password pages wired to the auth API
- profile page: avatar upload (image-sniffed, old file cleanup), bio,
contacts, arbitrary extra key/value fields, optimistic-concurrency 409
- role-based top navigation with placeholders for later-phase areas
- GET /files/{id} with scope-based access (session) or signed token (§5.1)
- security headers incl. CSP without unsafe-inline scripts
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
etalon