aboutsummaryrefslogtreecommitdiffstats
path: root/completions
AgeCommit message (Collapse)AuthorFilesLines
13 hoursfeat: completion suggests --type values per commandv0.3.4Danilo M.1-0/+14
Tab-completing --type now offers the valid values instead of nothing: account list -> asset/expense/revenue/liability, account create -> asset/expense/revenue, tx add/edit -> withdrawal/deposit/transfer. Keyed per command off a static FLAG_VALUES map in gen_completion.py (types are a fixed Firefly enum, so no API call in the completion). Partial input filters and unrelated flags still complete normally. Completion-only: no CLI/JSON/exit change, ships under v0.3.4. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
14 hoursfeat: tx add --skip-dupes for idempotent imports (v0.3.4)Danilo M.1-1/+1
Re-running an import (a retried or double-read batch) created phantom duplicate transactions and drifted balances, with no signal. --skip-dupes searches for an existing tx with the same amount + date + source + destination before writing; on a match it emits {"skipped": "duplicate", "matched_id": "<id>"} and exits 0 without writing. Off by default (one extra search per add only when set). dry-run wins over skip-dupes: a preview never triggers the search. Match is amount+date+ accounts, not description, so genuinely distinct same-value same-day transfers between the same accounts look like duplicates; documented. Also document a Firefly quirk in SKILL.md: a missing/deleted transaction id returns 401 (not 404) on tx get/edit/delete, so a 401 after delete confirms the record is gone rather than signalling an auth failure. Verified live on the test instance: create -> identical --skip-dupes skips with matched_id and no write, --dry-run --skip-dupes previews without searching, no-match --skip-dupes writes; records cleaned up after. PATCH: new optional flag, contract unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
14 hoursfeat: tx add --dry-run to validate before writing (v0.3.3)Danilo M.1-1/+1
An agent importing rows in a loop could fail mid-batch: a --to account that didn't exist yet errored after earlier rows had already been written, leaving a half-applied batch to clean up by hand. --dry-run resolves --from/--to and infers the type but sends nothing, printing {"dry_run": true, "would_send": {...}}. A missing account stays a hard error (exit 1) so the agent can dry-run every row first, create the accounts the errors name, then run the batch for real. No batch input mode: the agent already owns the loop. Verified live read-only: valid accounts emit would_send and write nothing (search confirms []), missing account exits 1 with the candidate list. PATCH: new optional flag, contract unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
14 hoursfeat: account balance --at for historical balance (v0.3.2)Danilo M.1-0/+1
Reconciliation needed an account's balance as of a past date; the command only reported the current balance, forcing hand-written sum scripts. --at YYYY-MM-DD re-fetches the account with Firefly's ?date= param, which recomputes current_balance as of that date (end of day, no client math). Without --at the current-balance path is unchanged (no extra API call). Verified live read-only against real data: Mediolanum current 183.23 vs --at 2026-06-10 164.38, boundary probe gives distinct daily balances. PATCH: new optional flag, contract unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
14 hoursfeat: tx list --all + truncation signal (v0.3.1)Danilo M.1-1/+1
tx list silently returned only --limit rows (default 20) with no signal when more matched, making a correct ledger look wrong during reconciliation. - --all auto-paginates every page via meta.pagination.total_pages - without --all, print "showing N of M (use --all for all)" to stderr when the result is truncated (stdout JSON stays clean) PATCH: new optional flag + stderr note, contract unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
31 hoursfeat: add tx edit and tx delete (v0.3.0)v0.3.0Danilo M.1-1/+3
Implements ISSUES.md #1, the missing other half of an import tool: correcting and removing mis-imported transactions without the web UI. - tx edit <id>: PATCH a single-split journal; only the fields passed are sent (--amount/--date/--desc/--from/--to/--category/--tags/--type). Errors if no field is given. Accounts resolve to ids; category/tags pass raw. - tx delete <id>: requires --yes (no interactive prompt, agent-first); prints {"deleted": "<id>"} on success. Also fix gen_completion.py: running it as a script put scripts/ on sys.path[0] and imported the installed (stale) firefly_cli, so the generated completion drifted (missing account balance/get, tx get/search). Prepend the repo root so the documented regen command uses this tree. The regenerated completion now reflects the full command surface. MINOR bump per the contract-keyed scheme: new commands, no breaking change. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
33 hourshelp, completion: descriptive help text and bash completionv0.2.1Danilo M.1-0/+64
Add group/leaf descriptions to argparse help and richer command help strings. Add generated bash completion (completions/firefly.bash) plus its generator (scripts/gen_completion.py), wired into the command checklist in CLAUDE.md and documented in the README. Bump to 0.2.1. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>