swaitsandClaude Fable 5.1
af553a3706
The planning session read plan/served.md, every design document and the five branches the owner paused on 2026-10-27 against main (44ee4083), with dry-run merges touching no branch. plan/2026-10-28_DRIFT_REPORT.md lists every divergence and judges every recorded departure (all kept; three owed back; one needs a decision: P6 made ledger-store depend on ledger-import). The plan: section 0 "What changed on 2026-10-28"; 8.0.1 and the unit Status lines name S4.4, S5, S6, S7 and P6 as built but unmerged; 8.0.2 records the destination (D-56: the installed app, P11a starts now) and the resume order with the known conflicts (the shared documents.rs test file, the types.rs appends, P6 retaking schema 18, the law numbers 34 to 37); 8.0.3 lists the names the EM approved during the wave; P11a gains step 0 (the macOS and Windows keyring stores, the Windows data root and pipes, a cross-platform CI build: today the agent path runs on Linux alone); Q-5 is due and Q-8 asks whether D-47 to D-54 stand; the Landlock sentences in section 2 and server.toml are struck. STATUS.md and HANDOFF.md say where things stand and how to resume; roadmap.md and gui.md are rewritten to today. The docs lose their stale facts: 122 tools, 2,700 tests, the parity floor at 21 with the P4b notes the slices never wrote, the ten crates and their dependency graph (jiff added to the core's allowed list), the cutover of 2026-10-20 recorded in capability-parity.md and parallel-run.md, the P4b S1 law and its two Kani harnesses added to verification.md (law 34), the proof counts (51 and 9), and README, the site and install.sh honest about macOS (the viewer runs; connecting an agent waits for P11a's keyring store, and the installer now says so instead of failing). just check on main at44ee4083: exit 0 (167 suites, 2,871 passed, 0 failed, 7 ignored, the server-only build clean). No code changed. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HS8WZziDnyxunC23MUxrLp
102 lines
5.0 KiB
Markdown
102 lines
5.0 KiB
Markdown
# Roadmap
|
|
|
|
Where wanderbooks has been, where it is, and where it is going. The
|
|
detailed plan for the remaining work is `plan/served.md` (section 8.0:
|
|
status, order, hand-off); `plan/STATUS.md` says where things stand this
|
|
week. Revised 2026-10-28.
|
|
|
|
## Done
|
|
|
|
### Phase 1 — Semantics core (2026-10-10 to 2026-10-13)
|
|
|
|
The engine: exact money, five account types, balanced entries, the
|
|
immutable posted log with reversals, trial balance, P&L, balance sheet,
|
|
the cash flow statement, period close, fixed assets, schedules, tax
|
|
categories, 1099 tracking, parties, A/R and A/P, payroll (accounting
|
|
only), projects; golden files, property tests, mutation testing, fuzzing;
|
|
the formal methods bar (Kani on the kernel, Lean on the laws,
|
|
`just prove`).
|
|
|
|
### Phase 2 — AI-first interface (2026-10-10 to 2026-10-15)
|
|
|
|
MCP is the primary interface, not a cloned GUI (the owner, 2026-10-10).
|
|
`ledger-store` (hash-chained journal, SQLite projection, blob store),
|
|
`ledger-mcp` (now 122 tools on `main`, 2026-10-28), `ledger-import`
|
|
(QuickBooks CSV in; IIF, XLSX, PDF out), `ledger-package` (the CPA
|
|
package), the read-only viewer `ledger-gui` (first version 2026-10-12;
|
|
`plan/gui.md`). Capability parity 12 of 12 against the GSheet+GDrive
|
|
workflow; the Sheet was frozen as of 2026-10-11 and the books cut over
|
|
on 2026-10-20 (`docs/capability-parity.md`, `docs/parallel-run.md`, both
|
|
historical now).
|
|
|
|
### Phase 3 — Oracle harness (partly done)
|
|
|
|
Done: the GnuCash differential (`just oracle`: 200 random books, 15,800
|
|
exact comparisons, adjudicated from principles; `docs/oracle.md` Layer
|
|
5), the mutation, fuzz and proof layers. Open: fixture books diffed as
|
|
`snapshot()` output automatically, and the machine-readable parity
|
|
catalog with CI enforcing the non-regression floor (the floor is 21
|
|
`done` rows, enforced by review until then).
|
|
|
|
### "Served" P0 to P3, Projects, P4b S1 to S4.3 (2026-10-16 to 2026-10-27)
|
|
|
|
The books became a server that owns its data, reached only through the
|
|
MCP from anywhere, every setting in the journal, one executable, backups
|
|
that restore, the format gate, the live books moved and re-verified
|
|
(`plan/served.md` 8.1, `plan/STATUS.md`). Engine correctness (the 24
|
|
audit defects), projects, receivables and payables first class
|
|
(documents read their figures from tagged standing lines), the invoice
|
|
PDF, terms, numbers and line accounts in the core and the journal.
|
|
|
|
## In flight (2026-10-28)
|
|
|
|
Built on paused branches, to be rebased and landed in order: P4b S4.4
|
|
(the create tools preview and confirm; the party contact block, D-55),
|
|
S5 (sent tracking), S6 (paying on terms: multi-document payments,
|
|
discounts, the processor fee, bills due, aging exports), S7 (bad debts),
|
|
P6 (the evidence pipeline: extraction, dedup, annotations). Then S4.5
|
|
(the PDF, viewer and handbook of `documents_v2`) and the owner's one
|
|
`documents_v2` upgrade of the live books (Q-5).
|
|
|
|
## Next
|
|
|
|
The order is `plan/served.md` 8.0.2. Two tracks run beside each other:
|
|
|
|
1. **The installed app: the destination (D-56, owner 2026-10-28).**
|
|
P11a: the platform groundwork first (the macOS and Windows keyring
|
|
stores, the Windows data root and pipes, a macOS and Windows CI
|
|
build; today the agent path works on Linux only), then the
|
|
cargo-packager trial, `wanderbooks update` and `rollback`, the Linux
|
|
packages (AUR, apt, dnf, AppImage), signing, the `.dmg` and the
|
|
per-user Windows installer, updates staged and applied at restart
|
|
(D-19, D-22, D-24 to D-26). Then P11b: the agent dock (Claude Code or
|
|
Codex in a terminal along the bottom of the viewer) and books homes
|
|
(D-27). "One app, one executable, separate processes."
|
|
2. **The agent surface and the re-audit.** P4 (draft states, one
|
|
evidence gate, the preview digest, the digest tool), P5 (the result
|
|
envelope, prompts, `books.start` v2, journal history, agent evals),
|
|
P8 (audit and attestations), P10 (the owner's full re-audit of the
|
|
first company's books through the MCP alone). P13 (bookspec, the
|
|
formal accounting specification: steps 4 to 6, the spec text, the
|
|
spec's own checker, the GnuCash adapter) runs beside. P12 (the whole
|
|
max-testing pass again: mutation, fuzzing, full property runs, proofs)
|
|
comes last, over everything.
|
|
|
|
## And then beyond
|
|
|
|
Self-host for the owner is built (P9a: a system unit, the move with the
|
|
head check, a static bearer over a tunnel). What follows the installed
|
|
app, in order of the plan's decisions: claude.ai and ChatGPT reaching a
|
|
self-hosted server through an external identity provider (P9b, after
|
|
Q-6); then the paid, multi-tenant hosted service `plan/served.md` section
|
|
11 describes and does not build (per-tenant processes, client-side vault
|
|
unwrap before any paid launch, a browser viewer, email-in, invitations).
|
|
The design is kept compatible with it at every step.
|
|
|
|
## Explicitly not on the roadmap
|
|
|
|
Bank feeds, payment processing, payroll tax tables and filing, tax
|
|
filing. Those are integrations and regulated domains, not cloning
|
|
targets. The engine gets clean seams for them; someone else's API
|
|
provides them. Budgets were declined as bloat (R-FUTURE-6).
|