hephaestus/docs/changelog.d/v1-prototype.feature.md
Erich Blume 652d2e89e6
Some checks failed
Build / validate (pull_request) Failing after 10m32s
heph: dev/installed isolation — mise run dev task + install how-to
- mise-tasks/dev runs the working-tree hephd on isolated .dev/ paths (gitignored)
  so in-repo development never touches the installed daemon's data; point a dev
  nvim at it via HEPH_SOCKET/HEPH_DB.
- docs/how-to/install-heph.md: install heph/hephd from the forge (build from
  source), the lazy.nvim `dir` setup for the subdir plugin, and dev isolation.
- gitignore .dev/ and the transient .claude scheduled-task state.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-02 09:39:59 -07:00

8 KiB

Begin the v1 prototype (Phase 1, tech-spec §11.1), built in TDD slices:

  • Cargo workspace + heph-core crate; migration-run SQLite schema (§4.5); clock-injected Store trait + LocalStore node create/get; single local-user bootstrap.
  • Markdown extraction (§5): [[wiki-links]] and GFM - [ ] checkbox context-items derived purely and idempotently from a body, skipping code blocks.
  • Committed tasks (§4.3, §6): task.create auto-creates the canonical context doc + canonical-context link; attention/do-date/late-on/state/recurrence columns; set-state/set-attention. Links CRUD (outgoing/backlinks). A body update reconciles wiki links (diff-based, resolved by alias/title, idempotent).
  • "What is next?" ranking (§7): pure, clock-injected, two-stage engine — candidacy filter (do-date as a boolean gate only) then a reorderable list of named dimensions (past-late-on → overdue-amount → attention band → FIFO). late_on is the sole urgency signal; blue hidden; red always shown. Proptest-checked total order. Store::next surfaces it over SQLite.
  • Recurrence — roll-forward in place (§4.4): completing a recurring task resets its checklist to all-unchecked, logs the occurrence, and advances the do-date to the next RRULE instance after now (skipping misses) — completion never carries forward (proptest-checked). Per-task append-only logs (log-of) with log.append/log.tail; skip advances without logging.
  • hephd daemon, local mode (§3, §6): exclusive file lock (handoff-ready), line-delimited JSON-RPC over a unix socket exposing the node/task/next/links/log methods, with DB work on tokio's blocking pool. Synchronous client for surfaces/CLI. Model types are serde-serializable.
  • heph CLI (§1) — a thin client of the daemon: next, task, doc, get, export. Export materializes the store to a <kind>/<id>.md tree with YAML frontmatter + body (§5), one-way, tombstones excluded.
  • Sync engine, local-only (§12): real hybrid logical clock + persistent device origin; an append-only op-log per mutation; an idempotent, order-independent merge/apply engine — last-writer-wins task scalars (discards surfaced in a conflicts queue), OR-set links, monotonic tombstones. Two-replica convergence proven.
  • Body text CRDT (§5, §12, slice 8d): node bodies now merge through the yrs text CRDT (body_crdt) instead of last-writer-wins — whole-buffer writes are diffed into the doc and the yrs delta rides the op, so concurrent edits to different regions both survive and never enqueue a conflict.
  • Network sync over HTTP (§6.1, §12, slice 9a): hephd --mode server exposes a sync hub (POST /sync/push, GET /sync/pull?after=<hlc>, axum) over the same store; hephd --mode local --hub-url <url> becomes a spoke that background-syncs its op-log with that hub (and on demand via the sync.now/sync.status RPC). Exchange is incremental by HLC cursor (sync_state) and idempotent. The merge engine is heph-core's, unchanged. Unauthenticated/single-owner for now (auth lands with OIDC). conflicts.list/conflicts.resolve are now reachable over the daemon socket.
  • Client mode (§3.1, slice 9b): hephd --mode client --server-url <url> runs with no local replica, proxying every store call to a server's POST /rpc endpoint (the full daemon API over HTTP). The daemon is now backend-agnostic (local/server front a LocalStore, client a RemoteStore), so surfaces see the same unix-socket API in every mode.
  • Hub authentication (§13, slice 10a): the sync hub now verifies an OIDC bearer token on /sync/* and /rpc — RS256-pinned JWT validation with exact issuer/audience, expiry, and a required subject; JWKS discovered and cached, refetched on key rotation (jsonwebtoken). Enabled with hephd --mode server --oidc-issuer <url> --oidc-audience <client-id> (open when unset, for local dev). A single-tenant owner gate binds the hub to the first authenticated identity and rejects any other. Verification sits behind a TokenVerifier trait, so it's tested entirely offline (stub middleware + an adversarial battery against an in-process mock IdP).
  • Client authentication (§13, slice 10b): heph auth login --hub-url <url> --issuer <url> --client-id <id> runs the OAuth 2.0 device-code flow and caches the token in the OS keyring; spokes and client mode attach it to hub requests, refreshing on expiry (--oidc-issuer/--oidc-client-id). Offline-tested against a mock OAuth server and a full spoke-to-authenticated-hub loop. (Auth/proxy HTTP uses the runtime-free ureq, since reqwest::blocking is unsafe inside the async daemon.)
  • CI runs the Rust suite (fmt/clippy/test) via the project build hook.
  • heph.nvim slice 11a (§8) — the primary surface begins: a Neovim plugin that is a thin client of the local hephd over its unix socket. A vim.uv JSON-RPC client (blocking call via vim.wait, id-demuxed, partial-line buffered, JSON null→Lua nil); buffer-backed nodes (heph://node/<id> with BufReadCmdnode.get / BufWriteCmdnode.update, whole-buffer body round-tripping exactly through the CRDT); [[wiki-link]] follow on <CR> via a new exact node.resolve {title} RPC (alias-then-title, the same mapping that materializes wiki links — unresolved links allowed); the daily journal (:Heph today); and the :Heph command surface. Headless e2e (§9) drives the plugin against a real daemon over a temp socket with a self-contained busted-style runner (no external plugins, no network): journal round-trip, follow-link, and link-two-docs/backlink.
  • heph.nvim slice 11b (§8) — task views: list is enriched to return titled rows (the same shape as next, with the canonical-context id) so the Organizational survey needs no per-row node.get. The plugin gains the Tactical :Heph next and Organizational :Heph list views (<CR> opens a task's canonical-context doc), task capture, set-attention, done/drop, skip, and per-task log append — each resolving "the current task" from the buffer (a task node, or a context doc via its canonical-context backlink). A vim.ui.select picker (Telescope auto-upgrade when installed) backs :Heph search/capture/attention. Headless e2e adds the capture→next→context→checklist→done workflow and the recurring fresh-checklist workflow (completing a recurring task rolls it forward and the next occurrence presents an all-unchecked checklist).
  • heph.nvim slice 11c (§8) — promotion + CI: task.promote mints a committed task from a - [ ] context-item line (addressed by its 1-based index) and rewrites that line into a [[link]] to the new task; :Heph promote does this for the line under the cursor. Wiki-link resolution now excludes a task's canonical-context doc, so [[Task Title]] resolves to the task itself (not its identically-titled context doc). The headless e2e suite runs in CI via a Dagger function that bakes a pinned, arch-detected Neovim onto a Rust image and runs the same self-contained suite developers run natively with mise run test-nvim; the runner fails on a zero-spec discovery so a misconfigured path can't pass silently.
  • heph.nvim managed daemon — plug-and-play by default: require("heph").setup({}) spawns and supervises a local hephd against the default paths when none is running, kills only the daemon it spawned on exit, and self-heals (respawns + reconnects if the daemon dies mid-session). A daemon you started yourself (a server/client architecture, or a service) is always respected — the plugin only spawns when nothing is serving the socket; with autostart = false it connects only and warns if unreachable. $HEPH_SOCKET / $HEPH_DB isolate a development Neovim onto a separate daemon + DB.
  • Dev/installed isolation tooling: a mise run dev task runs the working-tree hephd on isolated .dev/ paths, and a how-to (install-heph) covers installing heph/hephd from the forge (build-from-source), the lazy.nvim plugin setup, and pointing a dev Neovim at the dev daemon via $HEPH_SOCKET/$HEPH_DB so it never touches the installed store.