fix(mcp): query_journal defaults to status 'all' so totals equal ledger balances (#1863)

* fix(mcp): query_journal defaults to status 'all' so totals equal ledger balances

Posted-only was the default, but storno bookkeeping keeps both the
reversed original and its posted storno on the account: one-leg sums are
never balances. A customer's agent summed posted-only lines over a
storno-heavy quarter, found phantom VAT residuals on 2614/2641/2645/2647
and asked support to revert correct books.

- default status 'all' (posted + reversed), the same inclusion rule as
  trial balance, GL and SIE export
- explicit 'posted'/'reversed' get status_filter_warning when the
  opposite leg exists in range (entry-level head count, advisory)
- unknown status values now throw instead of matching nothing
- sibling description trims fund the additions within the tools/list
  payload budget

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(mcp): make status_filter_warning claim strictly true, hint status:'posted' for tag preview

Review feedback (PR Agent + skeptics): the entry-level opposite-status
count cannot prove the line-filtered totals are wrong, so the warning
now says one-leg totals CAN differ from balances; verb agrees with a
count of one. tag_journal_lines preview hint tells agents to pass
status:'posted' now that query_journal defaults to 'all'. Description
trims keep the tools/list payload inside the budget guard.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* fix(mcp): state the status_filter_warning count scope in the message

CodeRabbit: the opposite-status count applies entry-level filters only,
so say so in the warning instead of letting an account-scoped caller
read the count as account-scoped. Line-level scoping declined as
documented in DECISIONS.md (it would re-run the full line fetch).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

DECISIONS.md

@@ -1188,5 +1188,6 @@ One line per decision: `[YYYY-MM-DD] <decision>: <why>`. Appended by agents and
 [2026-08-24] Declared currency/voucher_series nullable in three MCP listing schemas on column-nullability alone (no traced null producer): loosening an output schema can only stop false validation failures, never cause one, and legacy rows predate the columns' defaults. Declined (for now) a full Ajv execute-vs-schema round-trip harness in output-schema.test.ts: right long-term answer to this bug class, but a session-sized project of its own; the audit's seven confirmed sites are pinned by a targeted declaration test instead.
 [2026-08-24] Manual matching (PR 6b) ships N:1 only (many outside rows -> one verifikat): bank links are independent per transaction (the engine allows it by design), skattekonto groups are all-or-nothing with the sum settling the verifikat (one guarded UPDATE, partial hit rolled back). 1:M (one row over several verifikat) and residual booking wait for a link table in 6c: the single journal_entry_id pointer on both row kinds cannot express them, and faking it (pointing the row at the residual verifikat) would break the bridge. The worksheet therefore enables Koppla only when the selection nets to zero and says so otherwise.
 [2026-08-24] Skattekonto payment file gets pain.001 through the supplier-payment generator (generateSupplierPain001), not the salary pain001 generator: the payment is a plain BG+OCR giro transfer (no SALA CtgyPurp), and the supplier dialect is the Validex-validated shape for exactly that; the LB path stays the default so nothing changes for banks still on LB.
+[2026-08-24] query_journal status default 'posted' -> 'all' (posted+reversed, the trial-balance inclusion rule) after a customer's agent summed posted-only lines over a storno-heavy Q2, found phantom VAT residuals on 2614/2641/2645/2647 and demanded a revert of correct books: one-leg sums are never balances. Funded the new status_filter_warning field and richer status docs inside the tools/list token budget by trimming sibling descriptions in the same tool rather than bumping the 59.95K ceiling (the payload guard's own guidance); warning fires off an entry-level opposite-status head count, exact for what the sentence claims and cheap, instead of re-running the line fetch.
 [2026-08-24] Detach-duplicate underlag ships as a SECURITY DEFINER RPC (detach_underlag_duplicate) instead of loosening the document triggers: the WORM guards stay intact for every other path, the carve-out is transaction-local (gnubok.allow_delete) and audit-logged first, and detach is refused unless another anchored underlag remains on the verifikat (BFL 5 kap 7 par) AND a remaining sibling has an identical sha256_hash (only byte-identical duplicates detach; skeptic-hardened 2026-08-24, along with an enforced posted-status guard and company_id on the audit row). Pinned docs (transactions.document_id / supplier_invoices.document_id) stay replace-only.
 [2026-08-24] Single-call chat console (general.help, AskConsole → /api/agent/ask) now carries the thread's earlier turns into every model call, via a new optional `history` on the provider-agnostic GenerateTextRequest (real message turns before the prompt in BOTH adapters: Anthropic-family messages array, OpenAI-compatible via AI SDK `messages`; an absent/empty history leaves the request byte-identical to the single-turn call, so hosted extraction and every other caller are untouched). The 08-20 RIP-3 cutover made each turn stateless (conversationId was only the tool actor id), so a follow-up in a resumed thread was answered blind (user report: "frågar vad jag refererar till"). History is loaded server-side from agent_messages (loadChatHistory: text only, hidden + tool rows dropped, alternation repaired, newest 16 rows / 10k chars) rather than sent by the client, so the client cannot forge earlier turns and old streaming threads replay cleanly. Rejected: inlining a transcript into the prompt (works everywhere but weaker turn semantics and blurs data vs instructions) and loading history in AskConsole (client-trusted history). Separately: the docked assistant panel now remembers its open thread per tab in sessionStorage (lib/agent-panel/session-restore) and reopens it after a full reload (the deploy prompt's "Ladda om" wiped it); sessionStorage, not user_preferences, because this is this-tab-this-session state that must not follow the user to other devices or tabs. And DeployReloadPrompt's full-width wrapper gets pointer-events-none: at z-[60] after the panel in DOM order it swallowed clicks on the panel's composer ("går ej att skriva").