feat(search): Elasticsearch searcher (name+content) and oo search (#37)

Reviewed-on: #24

.env.example

@@ -34,3 +34,10 @@ ONLYOFFICE_PROJECT_ID=33
 # MINIO_BUCKET=office
 # MINIO_ACCESS_KEY=
 # MINIO_SECRET_KEY=
+
+# oo search — direct Elasticsearch access for name + content search. ES lives
+# inside the OnlyOffice VM on localhost:9200; expose it with an SSH tunnel
+# (see docs/elasticsearch.md). ONLYOFFICE_ES_INDEX defaults to files_file.
+# ONLYOFFICE_ES_URL=http://127.0.0.1:9200
+# ONLYOFFICE_ES_INDEX=files_file
+# ONLYOFFICE_TENANT=

AGENTS.md

@@ -9,16 +9,17 @@ Canonical Go client for OnlyOffice Workspace (Projects + Calendar + CRM) and the
   - `request.go` — `Request`, `Query`, `Time`, `Token`, `MetaResponse`, `Permissions`.
   - `auth.go` — `Authenticate`, `AuthenticateContext`, `InvalidateToken`, `Auth`, token lifecycle.
   - `http.go` — transport + DRY response decoders (`ResponseArray`/`ResponseObject`/`postFormObject`/`putFormObject`/`deleteObject`).
-  - `projects.go`, `tasks.go`, `users.go`, `calendar.go`, `crm.go`, `files.go`, `mails.go`, `invoices.go` — typed / untyped domain methods. **`files.go`** — CRM opportunity upload plus **project/task Documents**. **`mails.go`** — OnlyOffice Workspace Mail. **`invoices.go`** — CRM invoices, PDF regen/cleanup, status. Association rules: [`docs/crm-associations.md`](docs/crm-associations.md).
+  - `projects.go`, `tasks.go`, `users.go`, `calendar.go`, `crm.go`, `files.go`, `files_webdav.go`, `files_stem.go`, `retry.go`, `mails.go`, `invoices.go` — typed / untyped domain methods. **`files.go`** — CRM opportunity upload plus **project/task Documents** (`UpdateFile`, `UploadToFolderReplacing`). **`files_webdav.go`** — Documents module by id (`ListDavFolder`, `MoveDavItems`/`CopyDavItems` with per-operation error surfacing, `ListFileOps`). **`retry.go`** — `DoRetry`: deterministic linear backoff (no jitter) on 429/502/503/504; every bulk tool routes API calls through it. **`mails.go`** — OnlyOffice Workspace Mail. **`invoices.go`** — CRM invoices, PDF regen/cleanup, status. Association rules: [`docs/crm-associations.md`](docs/crm-associations.md).
   - Pure stdlib + `google/go-querystring`; no UI, no dotenv.
 - **CLI — `cmd/oo/` as `package main`.** Cobra wrapper that loads `.env` via `godotenv` at startup. **Subject-based command tree** mirroring [`tea`](https://gitea.com/gitea/tea):
   - `main.go` — entry point (docstring lists the command tree).
   - `common.go` — `rootCmd`, `newOO`, `printTable`/`printObject`, `--output table|json` flag.
-  - `calendar.go`, `projects.go`, `projects_files.go`, `tasks.go`, `tasks_files.go`, `users.go`, `contacts.go`, `opportunities.go`, `cases.go`, `crm_tasks.go`, `mails.go`, `invoices.go` — one file per subject (or per subject facet), each registers in `init()`.
+  - `calendar.go`, `projects.go`, `projects_files.go`, `tasks.go`, `tasks_files.go`, `users.go`, `contacts.go`, `opportunities.go`, `cases.go`, `crm.go`, `crm_tasks.go`, `catalog.go`, `docs.go`, `dav.go`, `mails.go`, `invoices.go` — one file per subject (or per subject facet), each registers in `init()`. `dav.go` exposes the Documents module by id (`oo dav ls|move|copy|mkdir|rename-file|rename-folder|download|fileops`).
   - CLI-only deps (`spf13/cobra`, `joho/godotenv`) stay out of the library.
 - **TUI — `cmd/office/` as `package main`.** Bubble Tea three-pane browser (module tree, selectable list, markdown preview). Reuses `cmd/internal/bootstrap` for env/auth and the root `onlyoffice` library for all API calls. UI logic in `cmd/office/ui/`; preview/formatting in `cmd/office/preview/`; list loaders in `cmd/office/fetch/`.
   - **List table (`DataTable`)** — `cmd/office/ui/table*.go`. Column layout policies live in `cmd/office/model/table_layout.go` (`TableFlexLayoutFor`); cell rendering uses the bubbles/table inline pattern in `table_render.go` (`renderTableCell`, `padANSIWidth`). See `.cursor/skills/office-tui-table/SKILL.md` before changing center-pane tables.
 - **Shared bootstrap — `cmd/internal/bootstrap/`.** `LoadEnv()` + `NewClient(ctx)` extracted from `oo`; both binaries import it.
+- **Bulk Documents tools — `cmd/ooscan/`, `cmd/pdfamount/`, `cmd/kontoblatt/`, `cmd/kontolink/`.** Single-purpose binaries (folder index, PDF amounts, Kontoblatt summary/linking). Pace requests, route API calls through `DoRetry`; usage in README.
 - **Personal ops tooling** (disk inventory, dossier→CRM sync, SearXNG) lives in private [`eSlider/oo-workspace`](https://git.produktor.io/eSlider/oo-workspace) (`oow`), not in this public tree.
 
 ## Rules
@@ -27,7 +28,7 @@ Canonical Go client for OnlyOffice Workspace (Projects + Calendar + CRM) and the
 - New endpoints go into the library first; CLI commands are thin wrappers.
 - Prefer `ResponseObject` / `postFormObject` / `putFormObject` / `deleteObject` over hand-rolled `json.Unmarshal(responseField(...))` blocks — they exist for DRY, use them.
 - Domain split is by file, **not** by subpackage. Don't introduce `internal/` or `pkg/*` subpackages inside the library — it flattens the `*Client` call surface for a reason.
-- CLI commands follow **subject → verb** structure (`oo <subject> <verb>`), never `oo <verb>-<subject>`. Add new commands to the existing subject file if one fits; create a new `cmd/oo/<subject>.go` for a genuinely new domain.
+- CLI commands follow **subject → verb** structure (`oo <subject> <verb>`), never `oo <verb>-<subject>`. Add new commands to the existing subject file if one fits; create a new `cmd/oo/<subject>.go` for a genuinely new domain. The subject→verb tree in `cmd/oo/main.go` and the README table are documentation — update them with the code.
 - **Documents for agents:** prefer Markdown in git; OnlyOffice UI is weak for `.md`/`.txt`. Use `oo docs put-md` (md→docx) and `oo docs put-txt` (txt→docx, preserves line breaks). All upload paths default to **upsert** by `stem|ext` (`--replace`, default true); `--no-replace` fails on conflict; `--allow-duplicate` opts into raw OO append. `oo projects files dedupe PROJECT_ID` reports/removes duplicate stem|ext copies (`--apply`, `--cross`; includes project root folder).
 - Every table output goes through `printTable(headers, rows)`; every single-object through `printObject(v)`. Do not `fmt.Println` rows ad-hoc or the `--output json` flag breaks for that command.
 - No secrets in the repo; use `.env` (gitignored). Commit `.env.example` only.

README.md

@@ -503,6 +503,29 @@ type Task struct {
 |---|---|
 | `GetUsers()` | List all users with profiles |
 
+### Documents Files
+
+| Method | Description |
+|---|---|
+| `ListDavFolder(ctx, id)` | List a Documents folder (`@root` for virtual sections) |
+| `ListDavSections(ctx)` | Virtual sections (Documents, Projects, …) |
+| `CreateDavFolder(ctx, parentID, title)` | Create a subfolder |
+| `RenameDavFolder(ctx, id, title)` / `RenameDavFile(ctx, id, title)` | Rename folder / file |
+| `DownloadFile(ctx, id, dst)` / `DownloadDavFile(ctx, id, w)` | Download file bytes |
+| `UploadDavFile(ctx, folderID, fileName, src)` | Upload from a reader |
+| `UploadToFolder(ctx, folderID, localPath)` | Upload a local file into a folder |
+| `UploadToFolderReplacing(ctx, folderID, localPath)` | Upsert by `stem\|ext`; returns replaced ids |
+| `UpdateFile(ctx, fileID, localPath)` | New version of an existing file (same id, no copy) |
+| `MoveDavItems(ctx, folderIDs, fileIDs, dest)` | Move (`resolveType=Skip`); per-operation errors surfaced, not silent nil |
+| `CopyDavItems(ctx, folderIDs, fileIDs, dest)` | Copy (`conflictResolveType=Skip`); errors surfaced |
+| `MoveFiles(ctx, destFolderID, fileIDs)` | Move with `resolveType=Skip` + `holdResult`; errors surfaced |
+| `ListFileOps(ctx)` | Active file operations (move/copy status polling) |
+| `FolderFiles(ctx, folderID)` | Flat file list of a folder (stem helpers) |
+| `DeleteFilesByStem(ctx, folderID, stem)` | Remove `stem\|ext` copies |
+| `DoRetry(ctx, policy, fn)` | Deterministic linear backoff (N·Base, no jitter) on 429/502/503/504 |
+| `DefaultRetryPolicy()` | 5 attempts, 1s·2s·3s·4s waits, 30s cap |
+| `Transient(err)` | True for retriable OnlyOffice answers |
+
 ### Helper Types
 
 | Type | Description |
@@ -622,6 +645,8 @@ oo docs convert ./note.docx               # → note.md
 oo docs ocr ./scan.jpg --md ./scan.md     # searchable PDF + markdown
 oo docs hocr ./scan.jpg --lang spa --md ./scan.hocr.md --yaml ./scan.yml
 oo docs put-md 7 ./OO-HONDA-7-INDEX.md --folder 490
+oo docs put-txt 7 ./notes.txt --folder 490
+oo docs put-xlsx 7 ./table.xlsx --folder 490
 oo docs as-md 2815 --to ./parte.md        # download OO file as MD (OCR if needed)
 oo docs as-md 307 --hocr --lang spa       # OO download via go-hocr structure
 oo projects files put-md 7 ./note.md      # alias
@@ -631,21 +656,84 @@ oo tasks files upload 208 ./notes.pdf
 oo tasks files detach 208 12345
 ```
 
+### Documents module (`oo dav`)
+
+Direct access to the Documents module by folder/file id — the same calls that
+back `oo-webdav` and the project/task file commands. `move` sends
+`resolveType=Skip` + `holdResult=true`: without those params the legacy
+`fileops/move` endpoint answers 200 without moving anything, and the library
+surfaces such per-operation errors instead of a silent nil
+(`MoveDavItems` / `CopyDavItems` / `MoveFiles`).
+
+```bash
+oo dav ls 659
+oo dav ls @root                       # virtual sections (Documents, Projects, …)
+oo dav mkdir 659 "2026 inbox"
+oo dav move 659 22881 22882           # DEST_FOLDER_ID FILE_ID…
+oo dav move 659 22881 --folders 670   # move folders along with files
+oo dav copy 659 22881
+oo dav rename-file 22881 invoice-v2.pdf
+oo dav rename-folder 671 o2-archive
+oo dav download 22881 --to ./copy.pdf # default path: ./<server title>
+oo dav fileops                        # active move/copy operations (status polling)
+```
+
+### Search (`oo search`)
+
+Full-text search over the Documents index. The REST endpoint
+`/api/2.0/files/@search/{query}` only searches file names in the database, so
+`oo search` talks to the OnlyOffice **Elasticsearch** directly (index
+`files_file`). Name search is default; `--content` also matches extracted
+document text (`document.attachment.content`, Office formats only).
+See [`docs/elasticsearch.md`](docs/elasticsearch.md) for the tunnel setup.
+
+```bash
+oo search "Rechnung"                          # names only
+oo search "Mahngebühr" --content              # names + document text
+oo search "Rechnung" --folder 649 --limit 50
+oo search "Rechnung" --json                   # shorthand for -o json
+```
+
+Requires `ONLYOFFICE_ES_URL` (plus optional `ONLYOFFICE_ES_INDEX`,
+`ONLYOFFICE_TENANT`).
+
+### Bulk tools (`cmd/`)
+
+Small single-purpose binaries for bulk Documents work. All of them pace
+requests and retry transient OnlyOffice answers (429/502/503/504) with a
+deterministic linear backoff — no jitter, same waits on every run
+(see `DoRetry` below). Build with `go build ./cmd/<tool>`.
+
+```bash
+ooscan 659                             # recursive index → TSV: file_id, folder_id, path, title
+ooscan 659 666 > oo-index.tsv          # several roots into one index
+pdfamount 671                          # "Zu zahlender Betrag" per PDF → TSV: file_id, title, amount
+kontoblatt 3906 ./kontoblatt.xlsx      # summary (Gegenkonto/Monat) uploaded next to source
+kontolink IN.xlsx oo-index.tsv OUT.xlsx [FILE_ID] [AMOUNTS_TSV]
+# kontolink writes DocEditor links into the Link column: Beleg → supplier+month
+# → amount+date (5th arg = pdfamount output); with FILE_ID it updates the
+# source file in place, else uploads an "(links)" copy next to it.
+```
+
 | Subject | Verbs |
 |---|---|
 | `calendar` | `list`, `events`, `add`, `delete` |
-| `projects` | `list`, `get`, `milestones`, `create`, `update`, `delete`, **`files`** (`list`, `upload`, `download`, `rename`, `delete`) |
+| `projects` | `list`, `get`, `milestones`, `milestone-create`, `create`, `update`, `delete`, `contacts` (`add`, `remove`), `link-authors`, `link-git`, **`files`** (`list`, `upload`, `download`, `rename`, `delete`, `dedupe`, `as-md`, `put-md`, `put-txt`, `put-xlsx`) |
 | `tasks` | `list`, `get`, `create`, `update`, `delete`, `subtask add`, **`files`** (`list`, `upload`, `detach`) |
 | `users` | `list`, `self` (alias: `oo whoami`) |
-| `contacts` | `list`, `get`, `delete`, `info-add`, `merge`, `dedupe-info` |
+| `contacts` | `list`, `get`, `delete`, `info-add`, `merge`, `dedupe-info`, `tags`, `tag-add`, `tag-create`, `tag-remove` |
 | `persons` | `list`, `create`, `delete`, `dedupe` |
 | `companies` | `list`, `create`, `delete`, `dedupe`, `dedupe-persons` |
-| `opportunities` | `list`, `get`, `create`, `delete`, `stages`, `member-add`, `dedupe`, `dedupe-members`, `fix-titles` |
+| `opportunities` | `list`, `get`, `create`, `update`, `delete`, `stages`, `member-add`, `dedupe`, `dedupe-members`, `fix-titles` |
 | `invoices` | `list`, `get`, `create`, `update`, `pdf`, `pdf-cleanup`, `status`, `delete`, `items …` |
 | `crm` | `cleanup` |
-| `mails` | `accounts`, `folders`, `list`, `get`, `draft`, `attach`, `draft-invoice`, `delete` |
+| `mails` | `accounts`, `folders`, `list`, `get`, `download-attachment`, `draft`, `attach`, `draft-invoice`, `send`, `delete` |
 | `cases` | `list`, `create`, `delete`, `member-add` |
-| `crm-tasks` | `list`, `create`, `delete`, `categories` |
+| `crm-tasks` | `list`, `create`, `delete`, `categories`, `reassign-self` |
+| `docs` | `tools`, `convert`, `optimize`, `ocr`, `hocr`, `as-md`, `put-md`, `put-txt`, `put-xlsx` |
+| `catalog` | `match`, `merge`, `apply`, `scan-contacts`, `scan-projects`, `scan-thunderbird` |
+| `dav` | `ls`, `move`, `copy`, `mkdir`, `rename-file`, `rename-folder`, `download`, `fileops` |
+| `search` | `QUERY` (`--content`, `--folder ID`, `--limit N`, `--json`) |
 
 The CLI reads only `.env` from the current working directory (godotenv is a
 CLI-only concern — the library itself never loads dotfiles).

cmd/kontoblatt/main.go

@@ -0,0 +1,220 @@
+// Command kontoblatt builds a summary ("сводная таблица") of a Kontoblatt XLSX
+// (Datum, Gegenkonto, Buchungstext, Beleg, Soll, Haben, Bemerkung) and uploads
+// it back to the same OnlyOffice folder as the source file.
+//
+// Usage: kontoblatt <FILE_ID> <LOCAL_XLSX>
+package main
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+
+	onlyoffice "github.com/eslider/go-onlyoffice"
+	"github.com/xuri/excelize/v2"
+)
+
+type agg struct {
+	count   int
+	soll    float64
+	haben   float64
+	reFehlt int
+}
+
+type rec struct {
+	date, month, konto, text string
+	soll, haben              float64
+	reFehlt                  bool
+}
+
+var dateRe = regexp.MustCompile(`^\d{2}\.\d{2}\.\d{4}$`)
+
+func parseAmount(s string) float64 {
+	s = strings.TrimSpace(s)
+	s = strings.ReplaceAll(s, "€", "")
+	s = strings.ReplaceAll(s, " ", "")
+	s = strings.ReplaceAll(s, ",", "") // German thousands separator
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return 0
+	}
+	v, err := strconv.ParseFloat(s, 64)
+	if err != nil {
+		return 0
+	}
+	return v
+}
+
+func cell(row []string, i int) string {
+	if i < len(row) {
+		return strings.TrimSpace(row[i])
+	}
+	return ""
+}
+
+func main() {
+	if len(os.Args) < 3 {
+		fmt.Fprintln(os.Stderr, "usage: kontoblatt <FILE_ID> <LOCAL_XLSX>")
+		os.Exit(2)
+	}
+	fileID, path := os.Args[1], os.Args[2]
+	ctx := context.Background()
+
+	f, err := excelize.OpenFile(path)
+	if err != nil {
+		panic(err)
+	}
+	defer f.Close()
+
+	var recs []rec
+	for _, sh := range f.GetSheetList() {
+		rows, err := f.GetRows(sh)
+		if err != nil {
+			continue
+		}
+		for _, r := range rows {
+			d := cell(r, 0)
+			if !dateRe.MatchString(d) {
+				continue
+			}
+			text := cell(r, 2)
+			recs = append(recs, rec{
+				date:    d,
+				month:   d[3:10],
+				konto:   cell(r, 1),
+				text:    text,
+				soll:    parseAmount(cell(r, 4)),
+				haben:   parseAmount(cell(r, 5)),
+				reFehlt: strings.Contains(strings.ToUpper(text), "FEHLT"),
+			})
+		}
+	}
+
+	byKonto := map[string]*agg{}
+	byMonth := map[string]*agg{}
+	getK := func(k string) *agg {
+		if byKonto[k] == nil {
+			byKonto[k] = &agg{}
+		}
+		return byKonto[k]
+	}
+	getM := func(k string) *agg {
+		if byMonth[k] == nil {
+			byMonth[k] = &agg{}
+		}
+		return byMonth[k]
+	}
+	var tot agg
+	for _, r := range recs {
+		k := getK(r.konto)
+		k.count++
+		k.soll += r.soll
+		k.haben += r.haben
+		if r.reFehlt {
+			k.reFehlt++
+		}
+		m := getM(r.month)
+		m.count++
+		m.soll += r.soll
+		m.haben += r.haben
+		if r.reFehlt {
+			m.reFehlt++
+		}
+		tot.count++
+		tot.soll += r.soll
+		tot.haben += r.haben
+		if r.reFehlt {
+			tot.reFehlt++
+		}
+	}
+
+	out := excelize.NewFile()
+	defer out.Close()
+	writeSheet(out, "Nach Gegenkonto", "Gegenkonto", byKonto, tot)
+	writeSheet(out, "Nach Monat", "Monat", byMonth, tot)
+	outPath := "/tmp/opencode/kontoblatt-zusammenfassung.xlsx"
+	if err := out.SaveAs(outPath); err != nil {
+		panic(err)
+	}
+
+	// upload next to the source file
+	creds := onlyoffice.GetEnvironmentCredentials()
+	c := onlyoffice.NewClient(creds)
+	var src *onlyoffice.FileEntry
+	if derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+		var err error
+		src, err = c.GetFile(ctx, fileID)
+		return err
+	}); derr != nil {
+		panic(derr)
+	}
+	folder := ""
+	if src.FolderID != nil {
+		folder = src.FolderID.String()
+	}
+	title := ""
+	if src.Title != nil {
+		title = *src.Title
+	}
+	fmt.Printf("source: id=%s title=%q folder=%s\n", fileID, title, folder)
+
+	name := "Kontoblatt-1591-2025-Zusammenfassung.xlsx"
+	tmp := "/tmp/opencode/" + name
+	data, _ := os.ReadFile(outPath)
+	if err := os.WriteFile(tmp, data, 0o600); err != nil {
+		panic(err)
+	}
+	var entry *onlyoffice.FileEntry
+	if derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+		var err error
+		entry, _, err = c.UploadToFolderReplacing(ctx, folder, tmp)
+		return err
+	}); derr != nil {
+		panic(derr)
+	}
+	fmt.Printf("uploaded: %s -> folder %s (id %v)\n", name, folder, entry.ID)
+
+	// print the summary
+	printAgg("Nach Gegenkonto", byKonto, tot)
+	printAgg("Nach Monat", byMonth, tot)
+}
+
+func writeSheet(f *excelize.File, sheet, key string, m map[string]*agg, tot agg) {
+	f.NewSheet(sheet)
+	rows := [][]any{{key, "Anzahl", "Soll", "Haben", "Saldo", `davon "fehlt"`}}
+	keys := make([]string, 0, len(m))
+	for k := range m {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	for _, k := range keys {
+		a := m[k]
+		rows = append(rows, []any{k, a.count, a.soll, a.haben, a.soll - a.haben, a.reFehlt})
+	}
+	rows = append(rows, []any{"GESAMT", tot.count, tot.soll, tot.haben, tot.soll - tot.haben, tot.reFehlt})
+	for i, row := range rows {
+		for j, v := range row {
+			cellRef, _ := excelize.CoordinatesToCellName(j+1, i+1)
+			_ = f.SetCellValue(sheet, cellRef, v)
+		}
+	}
+}
+
+func printAgg(title string, m map[string]*agg, tot agg) {
+	fmt.Printf("\n== %s ==\n", title)
+	keys := make([]string, 0, len(m))
+	for k := range m {
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	fmt.Printf("%-12s %6s %12s %12s %12s %7s\n", "key", "count", "soll", "haben", "saldo", "fehlt")
+	for _, k := range keys {
+		a := m[k]
+		fmt.Printf("%-12s %6d %12.2f %12.2f %12.2f %7d\n", k, a.count, a.soll, a.haben, a.soll-a.haben, a.reFehlt)
+	}
+	fmt.Printf("%-12s %6d %12.2f %12.2f %12.2f %7d\n", "GESAMT", tot.count, tot.soll, tot.haben, tot.soll-tot.haben, tot.reFehlt)
+}

cmd/kontolink/main.go

@@ -0,0 +1,478 @@
+// Command kontolink fills the "Link" column of a Kontoblatt ("ungeklärte
+// Posten") XLSX by matching each row to an OnlyOffice document.
+//
+// Strategy (deterministic, conservative — no LLM):
+//  1. Beleg token (letters/digits from the "Beleg" column) appears in the file
+//     title; among candidates prefer (a) the row's month, (b) real invoices over
+//     copies/dupes, and require the result to be unique;
+//  2. else supplier + row month + "rechnung", again unique.
+//
+// A file is linked at most once (rows already carrying a link are kept and their
+// file counts as used). Ambiguous rows are left UNLINKED for manual review.
+//
+// Usage: kontolink <IN_XLSX> <INDEX_TSV> <OUT_XLSX>
+package main
+
+import (
+	"context"
+
+	"fmt"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+
+	onlyoffice "github.com/eslider/go-onlyoffice"
+	"github.com/xuri/excelize/v2"
+)
+
+var (
+	dateRe = regexp.MustCompile(`^\d{2}\.\d{2}\.\d{4}$`)
+	nonAln = regexp.MustCompile(`[^0-9a-z]+`)
+	fileID = regexp.MustCompile(`fileid=(\d+)`)
+)
+
+func parseDay(s string) (time.Time, bool) {
+	t, err := time.Parse("02.01.2006", strings.TrimSpace(s))
+	return t, err == nil
+}
+
+func titleDay(title string) (time.Time, bool) {
+	if len(title) >= 10 {
+		if t, err := time.Parse("2006-01-02", title[:10]); err == nil {
+			return t, true
+		}
+	}
+	return time.Time{}, false
+}
+
+// nearest picks the candidate whose title date is closest to rd. Ties and
+// undated candidates (when >1) are rejected.
+func nearest(cands []entry, rd time.Time) (entry, bool) {
+	if len(cands) == 1 {
+		return cands[0], true
+	}
+	best, bestD, tie := -1, 0.0, false
+	for i, e := range cands {
+		td, ok := titleDay(e.title)
+		if !ok {
+			continue
+		}
+		d := td.Sub(rd).Hours() / 24
+		if d < 0 {
+			d = -d
+		}
+		if best < 0 || d < bestD {
+			best, bestD, tie = i, d, false
+		} else if d == bestD {
+			tie = true
+		}
+	}
+	if best < 0 || tie {
+		return entry{}, false
+	}
+	return cands[best], true
+}
+
+const linkPrefix = "https://office.pro-dukt.de/Products/Files/DocEditor.aspx?fileid="
+
+type entry struct {
+	id, path, title, norm string
+}
+
+func norm(s string) string { return nonAln.ReplaceAllString(strings.ToLower(s), "") }
+
+func main() {
+	if len(os.Args) < 4 {
+		fmt.Fprintln(os.Stderr, "usage: kontolink <IN_XLSX> <INDEX_TSV> <OUT_XLSX>")
+		os.Exit(2)
+	}
+	in, idxPath, out := os.Args[1], os.Args[2], os.Args[3]
+
+	idxRaw, err := os.ReadFile(idxPath)
+	if err != nil {
+		panic(err)
+	}
+	var entries []entry
+	for _, line := range strings.Split(string(idxRaw), "\n") {
+		parts := strings.Split(line, "\t")
+		if len(parts) < 4 || parts[0] == "" {
+			continue
+		}
+		entries = append(entries, entry{id: parts[0], path: parts[2], title: parts[3], norm: norm(parts[3])})
+	}
+
+	f, err := excelize.OpenFile(in)
+	if err != nil {
+		panic(err)
+	}
+	defer f.Close()
+	sheet := f.GetSheetList()[0]
+	rows, err := f.GetRows(sheet)
+	if err != nil {
+		panic(err)
+	}
+
+	// optional 5th arg: amounts TSV "file_id\ttitle\tamount" (see cmd/pdfamount)
+	var amts []amtEntry
+	if len(os.Args) >= 6 && os.Args[5] != "" {
+		amts = loadAmounts(os.Args[5])
+	}
+
+	used := map[string]bool{}
+	for _, r := range rows {
+		if m := fileID.FindStringSubmatch(cell(r, 7)); m != nil {
+			used[m[1]] = true
+		}
+	}
+
+	var linked, byBeleg, bySupplier, byAmount, unmatched, ambiguous int
+	for i, r := range rows {
+		if i == 0 || !dateRe.MatchString(cell(r, 0)) || strings.TrimSpace(cell(r, 7)) != "" {
+			continue
+		}
+		beleg := norm(cell(r, 3))
+		supplier := supplierNorm(cell(r, 2))
+		month := monthYear(cell(r, 0))
+		rd, _ := parseDay(cell(r, 0))
+
+		e, kind, ok := pick(entries, used, beleg, supplier, month, rd)
+		if !ok {
+			if ae, aok := amountPick(amts, used, supplier, rowAmount(r), rd); aok {
+				e, kind, ok = entry{id: ae.id, title: ae.title}, "amount", true
+			}
+		}
+		if !ok {
+			if beleg != "" {
+				ambiguous++
+			} else {
+				unmatched++
+			}
+			continue
+		}
+		ref, _ := excelize.CoordinatesToCellName(8, i+1)
+		if err := f.SetCellValue(sheet, ref, linkPrefix+e.id); err != nil {
+			panic(err)
+		}
+		used[e.id] = true
+		linked++
+		switch kind {
+		case "beleg":
+			byBeleg++
+		case "supplier":
+			bySupplier++
+		case "amount":
+			byAmount++
+		}
+		fmt.Printf("row %3d  %-30s -> %s  [%s]\n", i+1, cell(r, 2), e.title, kind)
+	}
+
+	if err := f.SaveAs(out); err != nil {
+		panic(err)
+	}
+	fmt.Printf("\nlinked=%d (beleg=%d, supplier=%d, amount=%d), ambiguous=%d, no-candidate=%d\n",
+		linked, byBeleg, bySupplier, byAmount, ambiguous, unmatched)
+
+	// Optional 4th arg: source OnlyOffice file id. Try to update it in place;
+	// if it is locked (OnlyOffice 500), upload a "(links)" copy next to it.
+	if len(os.Args) >= 5 && os.Args[4] != "" {
+		c := onlyoffice.NewClient(onlyoffice.GetEnvironmentCredentials())
+		ctx, cancel := context.WithTimeout(context.Background(), 120*time.Second)
+		defer cancel()
+		var src *onlyoffice.FileEntry
+		if derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+			var err error
+			src, err = c.GetFile(ctx, os.Args[4])
+			return err
+		}); derr != nil {
+			panic(derr)
+		}
+		folder, title := "", ""
+		if src.FolderID != nil {
+			folder = src.FolderID.String()
+		}
+		if src.Title != nil {
+			title = *src.Title
+		}
+		uderr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+			_, err := c.UpdateFile(ctx, os.Args[4], out)
+			return err
+		})
+		if uderr == nil {
+			fmt.Printf("updated file %s in place\n", os.Args[4])
+			return
+		}
+		fmt.Printf("in-place update failed (locked?); uploading a copy to folder %s\n", folder)
+		ext := filepath.Ext(title)
+		name := strings.TrimSuffix(title, ext) + " (links)" + ext
+		tmp := filepath.Join(os.TempDir(), name)
+		data, _ := os.ReadFile(out)
+		if err := os.WriteFile(tmp, data, 0o600); err != nil {
+			panic(err)
+		}
+		if derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+			_, _, err := c.UploadToFolderReplacing(ctx, folder, tmp)
+			return err
+		}); derr != nil {
+			panic(derr)
+		}
+		fmt.Printf("uploaded copy: %s -> folder %s\n", name, folder)
+	}
+}
+
+func cell(r []string, i int) string {
+	if i < len(r) {
+		return strings.TrimSpace(r[i])
+	}
+	return ""
+}
+
+func supplierNorm(s string) string {
+	s = strings.ToUpper(s)
+	if i := strings.Index(s, ","); i >= 0 {
+		s = s[:i]
+	}
+	for _, w := range []string{"RE FEHLT", "GS FEHLT", "WOFR", "WOFÜR"} {
+		s = strings.ReplaceAll(s, w, "")
+	}
+	return norm(s)
+}
+
+func monthYear(date string) string {
+	if len(date) == 10 {
+		return date[6:10] + "-" + date[3:5]
+	}
+	return ""
+}
+
+// pick returns an unused candidate. Beleg match wins; supplier+month is a
+// fallback. When several candidates qualify, the one closest in time to the row
+// date wins; a tie is rejected (ambiguous) rather than guessed.
+func pick(entries []entry, used map[string]bool, beleg, supplier, month string, rd time.Time) (entry, string, bool) {
+	free := func(e entry) bool { return !used[e.id] }
+
+	if len(beleg) >= 5 {
+		var inMonth []entry
+		for _, e := range entries {
+			if free(e) && belegMatches(e.norm, beleg) &&
+				(month == "" || strings.Contains(e.title, month)) {
+				inMonth = append(inMonth, e)
+			}
+		}
+		if supplier != "" {
+			var s []entry
+			for _, e := range inMonth {
+				if strings.Contains(e.norm, supplier) {
+					s = append(s, e)
+				}
+			}
+			if len(s) > 0 {
+				inMonth = s
+			}
+		}
+		inMonth = topRank(inMonth)
+		if e, ok := nearest(inMonth, rd); ok {
+			return e, "beleg", true
+		}
+		// A Beleg is present but no file carries it: do NOT fall back to a
+		// supplier guess (that links the wrong invoice).
+		return entry{}, "", false
+	}
+
+	if supplier != "" && month != "" {
+		var c []entry
+		for _, e := range entries {
+			if free(e) && strings.Contains(e.norm, supplier) &&
+				strings.Contains(e.title, month) && strings.Contains(e.norm, "rechnung") {
+				c = append(c, e)
+			}
+		}
+		c = topRank(c)
+		if e, ok := nearest(c, rd); ok {
+			return e, "supplier", true
+		}
+	}
+	return entry{}, "", false
+}
+
+// topRank keeps only the highest-ranked candidates (real invoice over copy /
+// dupe / op), so a tie with a duplicate does not mask the real file.
+func topRank(cands []entry) []entry {
+	if len(cands) < 2 {
+		return cands
+	}
+	best := 0
+	for _, e := range cands {
+		if rank(e) > best {
+			best = rank(e)
+		}
+	}
+	out := cands[:0]
+	for _, e := range cands {
+		if rank(e) == best {
+			out = append(out, e)
+		}
+	}
+	return out
+}
+
+func rank(e entry) int {
+	s := 0
+	if strings.Contains(e.path, "/2025") || strings.Contains(e.path, "/2024") {
+		s += 4
+	}
+	if strings.Contains(e.norm, "rechnung") {
+		s += 2
+	}
+	if strings.Contains(e.norm, "dupe") || strings.Contains(e.norm, "copy") ||
+		strings.Contains(e.norm, "op") {
+		s--
+	}
+	return s
+}
+
+// belegMatches reports whether a Beleg identifies the file: the whole normalized
+// Beleg appears, or (for long numeric Belege, e.g. "24/641393110") an 8-digit
+// window of its longest digit run appears.
+func belegMatches(titleNorm, beleg string) bool {
+	if strings.Contains(titleNorm, beleg) {
+		return true
+	}
+	run := longestDigitRun(beleg)
+	for i := 0; i+8 <= len(run); i++ {
+		if strings.Contains(titleNorm, run[i:i+8]) {
+			return true
+		}
+	}
+	return false
+}
+
+func longestDigitRun(s string) string {
+	var best, cur strings.Builder
+	for _, r := range s {
+		if r >= '0' && r <= '9' {
+			cur.WriteRune(r)
+			if cur.Len() > best.Len() {
+				best.Reset()
+				best.WriteString(cur.String())
+			}
+		} else {
+			cur.Reset()
+		}
+	}
+	return best.String()
+}
+
+type amtEntry struct {
+	id      string
+	title   string
+	norm    string
+	amount  float64
+	date    time.Time
+	hasDate bool
+}
+
+func loadAmounts(path string) []amtEntry {
+	raw, err := os.ReadFile(path)
+	if err != nil {
+		return nil
+	}
+	var out []amtEntry
+	for _, line := range strings.Split(string(raw), "\n") {
+		p := strings.Split(line, "\t")
+		if len(p) < 3 {
+			continue
+		}
+		v, err := strconv.ParseFloat(strings.TrimSpace(p[2]), 64)
+		if err != nil {
+			continue
+		}
+		e := amtEntry{id: p[0], title: p[1], norm: norm(p[1]), amount: v}
+		if len(p[1]) >= 10 {
+			if t, err := time.Parse("2006-01-02", p[1][:10]); err == nil {
+				e.date, e.hasDate = t, true
+			}
+		}
+		out = append(out, e)
+	}
+	return out
+}
+
+func rowAmount(r []string) float64 {
+	if v := parseAmount(cell(r, 4)); v != 0 {
+		return v
+	}
+	return parseAmount(cell(r, 5))
+}
+
+func parseAmount(s string) float64 {
+	s = strings.ReplaceAll(s, "€", "")
+	s = strings.ReplaceAll(s, " ", "")
+	s = strings.ReplaceAll(s, ",", ".")
+	if s == "" {
+		return 0
+	}
+	v, err := strconv.ParseFloat(s, 64)
+	if err != nil {
+		return 0
+	}
+	return v
+}
+
+// amountPick matches a row to an O2 invoice by amount + nearest date. Scoped to
+// Telefonica/O2 rows and O2 files, so it cannot cross-link other suppliers.
+func amountPick(amts []amtEntry, used map[string]bool, supplier string, amt float64, rd time.Time) (amtEntry, bool) {
+	if amt <= 0 || len(amts) == 0 {
+		return amtEntry{}, false
+	}
+	if !strings.Contains(supplier, "telefonica") && !strings.Contains(supplier, "o2") {
+		return amtEntry{}, false
+	}
+	var cands []amtEntry
+	for _, a := range amts {
+		if used[a.id] || !strings.Contains(a.norm, "o2") {
+			continue
+		}
+		d := a.amount - amt
+		if d < 0 {
+			d = -d
+		}
+		if d > 0.005 {
+			continue
+		}
+		if a.hasDate && !rd.IsZero() {
+			days := a.date.Sub(rd).Hours() / 24
+			if days < 0 {
+				days = -days
+			}
+			if days > 75 {
+				continue
+			}
+		}
+		cands = append(cands, a)
+	}
+	if len(cands) == 1 {
+		return cands[0], true
+	}
+	best, bestD, tie := -1, 0.0, false
+	for i, a := range cands {
+		if !a.hasDate {
+			continue
+		}
+		d := a.date.Sub(rd).Hours() / 24
+		if d < 0 {
+			d = -d
+		}
+		if best < 0 || d < bestD {
+			best, bestD, tie = i, d, false
+		} else if d == bestD {
+			tie = true
+		}
+	}
+	if best < 0 || tie {
+		return amtEntry{}, false
+	}
+	return cands[best], true
+}

cmd/oo/main.go

@@ -3,20 +3,22 @@
 // Command tree is subject-based (mirrors the library split and the `tea` CLI):
 //
 //	oo calendar      list | events | add | delete
-//	oo projects      list | get | milestones | create | update | delete | files (list|upload|download|rename|delete|as-md|put-md)
+//	oo projects      list | get | milestones | milestone-create | create | update | delete | contacts (add|remove) | link-authors | link-git | files (list|upload|download|rename|delete|dedupe|as-md|put-md|put-txt|put-xlsx)
 //	oo tasks         list | get | create | update | delete | subtask add | files (list|upload|detach)
 //	oo users         list | self            (alias: oo whoami)
-//	oo contacts      list | get | delete | info-add | merge | dedupe-info
+//	oo contacts      list | get | delete | info-add | merge | dedupe-info | tags | tag-add | tag-create | tag-remove
 //	oo persons       list | create | delete | dedupe
 //	oo companies     list | create | delete | dedupe | dedupe-persons
-//	oo opportunities list | get | create | delete | stages | member-add | dedupe | dedupe-members | fix-titles
+//	oo opportunities list | get | create | update | delete | stages | member-add | dedupe | dedupe-members | fix-titles
 //	oo cases         list | create | delete | member-add
-//	oo crm-tasks     list | create | delete | categories
+//	oo crm-tasks     list | create | delete | categories | reassign-self
 //	oo crm           cleanup
-//	oo mails         accounts | folders | list | get | download-attachment | draft | attach | draft-invoice | delete
+//	oo mails         accounts | folders | list | get | download-attachment | draft | attach | draft-invoice | send | delete
 //	oo invoices      list | get | create | update | pdf | pdf-cleanup | status | delete | items …
-//	oo docs          tools | convert | ocr | as-md | put-md
+//	oo docs          tools | convert | optimize | ocr | hocr | as-md | put-md | put-txt | put-xlsx
+//	oo catalog       match | merge | apply | scan-contacts | scan-projects | scan-thunderbird
 //	oo dav           ls | move | copy | mkdir | rename-file | rename-folder | download | fileops
+//	oo search        QUERY [--content] [--folder ID] [--limit N] [--json]
 //
 // CRM association rules: docs/crm-associations.md
 //

cmd/oo/search.go

@@ -0,0 +1,72 @@
+package main
+
+import (
+	onlyoffice "github.com/eslider/go-onlyoffice"
+	"github.com/spf13/cobra"
+)
+
+func init() {
+	rootCmd.AddCommand(searchCmd())
+}
+
+// searchCmd queries the OnlyOffice Elasticsearch index directly. The REST
+// /api/2.0/files/@search endpoint only searches file names in the database;
+// content search needs ES (see docs/elasticsearch.md).
+func searchCmd() *cobra.Command {
+	var (
+		content bool
+		folder  string
+		limit   int
+		asJSON  bool
+	)
+	cmd := &cobra.Command{
+		Use:   "search QUERY",
+		Short: "Full-text search over documents by name, optionally by content (Elasticsearch)",
+		Long: "Search the OnlyOffice Documents index.\n\n" +
+			"By default only file names are matched. With --content the query also\n" +
+			"matches extracted document text (document.attachment.content); this covers\n" +
+			"Office formats (docx/xlsx/pptx) and is slower.\n\n" +
+			"Requires ONLYOFFICE_ES_URL (and optionally ONLYOFFICE_ES_INDEX,\n" +
+			"ONLYOFFICE_TENANT). See docs/elasticsearch.md for the tunnel setup.",
+		Args: cobra.ExactArgs(1),
+		RunE: func(cmd *cobra.Command, args []string) error {
+			if asJSON {
+				outputFormat = "json"
+			}
+			es, err := onlyoffice.NewESSearcher(onlyoffice.ESConfigFromEnv())
+			if err != nil {
+				return err
+			}
+			hits, err := es.Search(cmd.Context(), onlyoffice.SearchQuery{
+				Text:      args[0],
+				InContent: content,
+				FolderID:  folder,
+				Limit:     limit,
+			})
+			if err != nil {
+				return err
+			}
+			rows := make([]map[string]any, 0, len(hits))
+			for _, h := range hits {
+				rows = append(rows, map[string]any{
+					"id":        h.ID,
+					"title":     h.Title,
+					"folder":    h.ParentID,
+					"score":     h.Score,
+					"highlight": h.Highlight,
+				})
+			}
+			if outputFormat == "json" {
+				printJSON(rows)
+				return nil
+			}
+			printTable([]string{"id", "title", "folder", "score", "highlight"}, rows)
+			return nil
+		},
+	}
+	cmd.Flags().BoolVar(&content, "content", false, "also match extracted document content")
+	cmd.Flags().StringVar(&folder, "folder", "", "limit to a Documents folder id")
+	cmd.Flags().IntVar(&limit, "limit", 20, "maximum number of results")
+	cmd.Flags().BoolVar(&asJSON, "json", false, "shorthand for --output json")
+	return cmd
+}

cmd/oo/search_test.go

@@ -0,0 +1,46 @@
+package main
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+)
+
+func TestSearchCommandRegisteredWithFlags(t *testing.T) {
+	cmd, _, err := rootCmd.Find([]string{"search"})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if cmd.Name() != "search" {
+		t.Fatalf("search resolved to %q", cmd.Name())
+	}
+	for _, name := range []string{"content", "folder", "limit", "json"} {
+		if cmd.Flags().Lookup(name) == nil {
+			t.Errorf("search: missing --%s flag", name)
+		}
+	}
+	if got := cmd.Flags().Lookup("limit").DefValue; got != "20" {
+		t.Errorf("--limit default = %q, want 20", got)
+	}
+}
+
+func TestSearchWithoutESURLIsClearError(t *testing.T) {
+	clearEnv(t, "ONLYOFFICE_ES_URL", "ONLYOFFICE_ES_INDEX", "ONLYOFFICE_TENANT")
+	errBuf := &bytes.Buffer{}
+	rootCmd.SetErr(errBuf)
+	rootCmd.SetOut(&bytes.Buffer{})
+	rootCmd.SetArgs([]string{"search", "Rechnung"})
+	t.Cleanup(func() {
+		rootCmd.SetArgs(nil)
+		rootCmd.SetOut(nil)
+		rootCmd.SetErr(nil)
+	})
+
+	err := rootCmd.Execute()
+	if err == nil {
+		t.Fatal("expected error without ONLYOFFICE_ES_URL")
+	}
+	if !strings.Contains(err.Error(), "ONLYOFFICE_ES_URL") {
+		t.Fatalf("error %q missing ONLYOFFICE_ES_URL", err.Error())
+	}
+}

cmd/ooscan/main.go

@@ -0,0 +1,50 @@
+// Command ooscan recursively lists OnlyOffice Documents folders into a TSV
+// index: file_id, folder_id, path, title.
+//
+// Usage: ooscan <FOLDER_ID> [<FOLDER_ID>...]
+package main
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"time"
+
+	onlyoffice "github.com/eslider/go-onlyoffice"
+)
+
+func main() {
+	ctx := context.Background()
+	c := onlyoffice.NewClient(onlyoffice.GetEnvironmentCredentials())
+	seen := map[string]bool{}
+	for _, root := range os.Args[1:] {
+		walk(ctx, c, root, "", 0, seen)
+	}
+}
+
+func walk(ctx context.Context, c *onlyoffice.Client, folderID, path string, depth int, seen map[string]bool) {
+	if depth > 8 || seen[folderID] {
+		return
+	}
+	seen[folderID] = true
+	// Throttle: OnlyOffice rate-limits (429) and the host must not be flooded.
+	time.Sleep(350 * time.Millisecond)
+	ctx, cancel := context.WithTimeout(ctx, 60*time.Second)
+	defer cancel()
+	var l *onlyoffice.DavListing
+	derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+		var err error
+		l, err = c.ListDavFolder(ctx, folderID)
+		return err
+	})
+	if derr != nil {
+		fmt.Fprintf(os.Stderr, "list %s (%s): %v\n", path, folderID, derr)
+		return
+	}
+	for _, f := range l.Files {
+		fmt.Printf("%s\t%s\t%s\t%s\n", f.ID, folderID, path, f.Title)
+	}
+	for _, sub := range l.Folders {
+		walk(ctx, c, sub.ID, path+"/"+sub.Title, depth+1, seen)
+	}
+}

cmd/pdfamount/main.go

@@ -0,0 +1,124 @@
+// Command pdfamount walks a Documents folder, downloads matching PDFs and
+// extracts the payable amount, printing "file_id\ttitle\tamount".
+//
+// Usage: pdfamount <FOLDER_ID> [TITLE_FILTER_REGEX]
+package main
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+
+	onlyoffice "github.com/eslider/go-onlyoffice"
+)
+
+var amountRe = regexp.MustCompile(`(?i)(zu zahlender betrag|rechnungsbetrag)\s*[:\s]*([0-9][0-9.]*,[0-9]{2})`)
+
+func main() {
+	if len(os.Args) < 2 {
+		fmt.Fprintln(os.Stderr, "usage: pdfamount <FOLDER_ID> [TITLE_FILTER_REGEX]")
+		os.Exit(2)
+	}
+	folder := os.Args[1]
+	filter := regexp.MustCompile(`(?i)rechnung`)
+	if len(os.Args) >= 3 {
+		filter = regexp.MustCompile(os.Args[2])
+	}
+	ctx := context.Background()
+	c := onlyoffice.NewClient(onlyoffice.GetEnvironmentCredentials())
+
+	files := listAll(ctx, c, folder)
+	for _, f := range files {
+		if !filter.MatchString(f.title) {
+			continue
+		}
+		if !strings.HasSuffix(strings.ToLower(f.title), ".pdf") {
+			continue
+		}
+		amount, err := pdfAmount(ctx, c, f.id)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "%s: %v\n", f.title, err)
+			continue
+		}
+		if amount == "" {
+			continue
+		}
+		fmt.Printf("%s\t%s\t%s\n", f.id, f.title, amount)
+	}
+}
+
+type file struct{ id, title string }
+
+func listAll(ctx context.Context, c *onlyoffice.Client, folder string) []file {
+	seen := map[string]bool{}
+	var out []file
+	var walk func(string)
+	walk = func(id string) {
+		if seen[id] {
+			return
+		}
+		seen[id] = true
+		time.Sleep(300 * time.Millisecond)
+		l, err := c.ListDavFolder(ctx, id)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "list %s: %v\n", id, err)
+			return
+		}
+		for _, f := range l.Files {
+			out = append(out, file{f.ID, f.Title})
+		}
+		for _, sub := range l.Folders {
+			walk(sub.ID)
+		}
+	}
+	walk(folder)
+	return out
+}
+
+func pdfAmount(ctx context.Context, c *onlyoffice.Client, id string) (string, error) {
+	time.Sleep(time.Second)
+	tmp, err := os.CreateTemp("", "pdf-*.pdf")
+	if err != nil {
+		return "", err
+	}
+	defer os.Remove(tmp.Name())
+	derr := onlyoffice.DoRetry(ctx, onlyoffice.DefaultRetryPolicy(), func() error {
+		_ = tmp.Truncate(0)
+		_, _ = tmp.Seek(0, 0)
+		_, err := c.DownloadFile(ctx, id, tmp)
+		return err
+	})
+	if derr != nil {
+		tmp.Close()
+		return "", derr
+	}
+	tmp.Close()
+	var buf bytes.Buffer
+	cmd := exec.CommandContext(ctx, "pdftotext", "-layout", tmp.Name(), "-")
+	cmd.Stdout = &buf
+	if err := cmd.Run(); err != nil {
+		return "", err
+	}
+	m := amountRe.FindStringSubmatch(buf.String())
+	if m == nil {
+		return "", nil
+	}
+	return parseDe(m[2]), nil
+}
+
+// parseDe turns "1.234,56" into 1234.56.
+func parseDe(s string) string {
+	s = strings.ReplaceAll(s, ".", "")
+	s = strings.ReplaceAll(s, ",", ".")
+	v, err := strconv.ParseFloat(s, 64)
+	if err != nil {
+		return s
+	}
+	return strconv.FormatFloat(v, 'f', 2, 64)
+}

docs/elasticsearch.md

@@ -0,0 +1,125 @@
+---
+type: reference
+status: current
+related:
+  - README.md
+  - file_es.go
+---
+
+# Elasticsearch — полнотекстовый поиск OnlyOffice
+
+## Что это
+
+Полнотекстовый поиск OnlyOffice Workspace работает на **Elasticsearch**.
+Клиент на сервере — NEST. Индекс — имя таблицы.
+
+Для файлов индекс `files_file`:
+
+| поле | тип | смысл |
+|------|-----|-------|
+| `id` | integer | id файла (тот же, что в REST/Documents) |
+| `title` | text (`whitespacecustom`) | имя файла |
+| `tenantId` | integer | тенант (портал) |
+| `folders` | nested | список папок: `folderId` (строка), `id`, `tenantId` |
+| `document.attachment.content` | text (`document`) | извлеченный текст (ingest-attachment) |
+| `document.attachment.content_type` | text | MIME |
+
+Важно:
+- Живой сервер — **Elasticsearch 7.16.3**, кластер `elasticsearch`.
+- REST `GET /api/2.0/files/@search/{query}` ищет **только по имени в БД**
+  (`fileDao.Search`), ES не задействует. Для поиска по содержимому нужен
+  прямой ES — это и делает `oo search`.
+- `title` analyzer `whitespacecustom` режет по пробелам и lower-case. Полное
+  имя файла — один токен (`Rechnung-4711.pdf`), поэтому поиск по имени ищет
+  слово целиком, а не подстроку.
+- `document.attachment.content` заполняется **только для Office-форматов**
+  (docx / xlsx / pptx). У PDF/txt, залитых через API, контент не извлекается.
+- Индексация асинхронная (TeamLabSvc) — файл появляется в ES не мгновенно.
+
+## Доступ
+
+ES слушает `127.0.0.1:9200` **внутри** VM OnlyOffice. Снаружи порт закрыт,
+SSH в VM открыт на хосте как `127.0.0.1:32` (контейнер `onlyoffice-v2`,
+QEMU). Схема — SSH-туннель.
+
+```bash
+# из корня go-onlyoffice (ключ и хост — как в infra-доках)
+ssh -f -N -o ControlMaster=no -o ControlPath=none \
+    -p 32 -i ~/.ssh/id_ed25519 \
+    -L 9200:127.0.0.1:9200 root@127.0.0.1
+
+curl -s http://127.0.0.1:9200/ | head            # tagline + version
+curl -s 'http://127.0.0.1:9200/_cat/indices?h=index,docs.count'
+```
+
+`-o ControlMaster=no -o ControlPath=none` обязательны: иначе forward уходит
+в persistent master-соединение из `~/.ssh/config` и порт остаётся занят.
+
+Проверить, что туннель жив:
+
+```bash
+curl -s http://127.0.0.1:9200/files_file/_count
+```
+
+## Переменные
+
+| env | default | смысл |
+|-----|---------|-------|
+| `ONLYOFFICE_ES_URL` | — (обязателен) | `scheme://host:port` ES |
+| `ONLYOFFICE_ES_INDEX` | `files_file` | индекс |
+| `ONLYOFFICE_TENANT` | пусто (все) | фильтр `tenantId` |
+
+Имена — в [`.env.example`](../.env.example). Секретов нет: ES без пароля.
+
+## CLI
+
+```bash
+ONLYOFFICE_ES_URL=http://127.0.0.1:9200 oo search "Rechnung"
+ONLYOFFICE_ES_URL=http://127.0.0.1:9200 oo search "Mahngebühr" --content
+oo search "Rechnung" --folder 649 --limit 50 --json
+```
+
+Флаги: `--content` (искать и по тексту), `--folder ID` (папка
+`folders.folderId`), `--limit N` (по умолчанию 20, максимум 200),
+`--json` = `-o json`.
+
+## Библиотека
+
+`file_es.go` — `ESSearcher` (`Name() = "elasticsearch"`), прямой ES REST на
+stdlib `net/http`:
+
+```go
+es, _ := onlyoffice.NewESSearcher(onlyoffice.ESConfigFromEnv())
+hits, _ := es.Search(ctx, onlyoffice.SearchQuery{
+    Text: "Rechnung", InContent: true, Limit: 20,
+})
+```
+
+Запрос: `multi_match` по `title^2` (+ `document.attachment.content` при
+`InContent`), фильтры `tenantId` и `folders.folderId`, `_source`
+id/title/folders, `highlight` для фрагмента. Ответ → `[]SearchHit` (модель из
+эпика #34; пока объявлена в `file_es.go`, переедет в `file_core.go` с F1 #35).
+
+## Тесты
+
+```bash
+# unit — чистые builders/парсеры, без сети
+go test ./ -run ES
+
+# integration — нужен ONLYOFFICE_ES_URL (+ креды REST для залива)
+set -a; . .env; set +a
+ONLYOFFICE_ES_URL=http://127.0.0.1:9200 ONLYOFFICE_TENANT=1 \
+  go test -tags=integration -run TestIntegrationESSearch -v .
+```
+
+Интеграционный тест заливает временный xlsx (в имени и в ячейке — уникальные
+токены), ждёт индексации, проверяет поиск по имени и по содержимому, затем
+удаляет проект.
+
+## Грабли
+
+- `locale`/версия ES: 7.16.3, `_search` совместим с REST 7.x.
+- ES без auth и слушает только localhost — туннель обязателен.
+- Фильтр `tenantId` сузит выдачу; без него видны документы всех тенантов.
+- Поиск по содержимому PDF, залитых через API, не работает (нет
+  `attachment.content`) — только Office-форматы.

file_es.go

@@ -0,0 +1,329 @@
+package onlyoffice
+
+// Elasticsearch backend of the unified file client (epic #34, F3 #37).
+//
+// OnlyOffice full-text search runs on Elasticsearch (index `files_file`, NEST
+// client on the server). The REST endpoint GET /api/2.0/files/@search/{query}
+// only searches file names in the database, so content search needs a direct
+// ES query. The live server is Elasticsearch 7.16.3; the request shape below
+// is plain REST and stays stdlib-only, matching the repo's no-extra-deps rule.
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"regexp"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// Canonical file/search model (epic #34, F1 #35). Declared here because F1 is
+// not merged yet; move to file_core.go and drop these when it lands. Keep the
+// shape identical to the contract in #35.
+type (
+	// Kind distinguishes a file from a folder.
+	Kind int
+	// Entry is a canonical file/folder record.
+	Entry struct {
+		ID       string
+		ParentID string
+		Title    string
+		Kind     Kind
+		Size     int64
+		MIME     string
+		Created  time.Time
+		Modified time.Time
+		Version  int
+		Provider string
+	}
+	// SearchQuery is a backend-agnostic search request.
+	SearchQuery struct {
+		Text       string
+		InContent  bool
+		FolderID   string
+		Extensions []string
+		Limit      int
+	}
+	// SearchHit is a search result entry plus its relevance data.
+	SearchHit struct {
+		Entry
+		Score     float64
+		Highlight string
+		Path      []string
+	}
+	// Searcher searches a document store by name and optionally content.
+	Searcher interface {
+		Search(ctx context.Context, q SearchQuery) ([]SearchHit, error)
+		Name() string
+	}
+)
+
+// Kind values (epic #34).
+const (
+	File Kind = iota
+	Folder
+)
+
+const (
+	defaultESIndex    = "files_file"
+	defaultESLimit    = 20
+	maxESLimit        = 200
+	maxESResponseSize = 8 << 20
+)
+
+// ESConfig configures the direct Elasticsearch searcher.
+type ESConfig struct {
+	URL    string // scheme://host:port of the ES HTTP endpoint
+	Index  string // index name, default files_file
+	Tenant string // tenantId filter, empty means all tenants
+}
+
+// ESConfigFromEnv reads ONLYOFFICE_ES_URL, ONLYOFFICE_ES_INDEX (default
+// files_file) and ONLYOFFICE_TENANT. The library never loads dotfiles — the
+// CLI does that.
+func ESConfigFromEnv() ESConfig {
+	return ESConfig{
+		URL:    strings.TrimRight(strings.TrimSpace(os.Getenv("ONLYOFFICE_ES_URL")), "/"),
+		Index:  firstNonEmpty(os.Getenv("ONLYOFFICE_ES_INDEX"), defaultESIndex),
+		Tenant: strings.TrimSpace(os.Getenv("ONLYOFFICE_TENANT")),
+	}
+}
+
+// ESSearcher queries OnlyOffice's Elasticsearch index directly for file name
+// and document content.
+type ESSearcher struct {
+	cfg  ESConfig
+	http *http.Client
+}
+
+// NewESSearcher returns a searcher for the OnlyOffice Elasticsearch index.
+// The URL is required; an empty index falls back to files_file.
+func NewESSearcher(cfg ESConfig) (*ESSearcher, error) {
+	if strings.TrimSpace(cfg.URL) == "" {
+		return nil, fmt.Errorf("onlyoffice: elasticsearch URL is empty (set ONLYOFFICE_ES_URL)")
+	}
+	cfg.URL = strings.TrimRight(cfg.URL, "/")
+	if cfg.Index == "" {
+		cfg.Index = defaultESIndex
+	}
+	return &ESSearcher{cfg: cfg, http: &http.Client{Timeout: 30 * time.Second}}, nil
+}
+
+// Name implements Searcher.
+func (s *ESSearcher) Name() string { return "elasticsearch" }
+
+// Search runs a multi_match over title (and, when q.InContent is set,
+// document.attachment.content), filtered by tenant and optional folder.
+func (s *ESSearcher) Search(ctx context.Context, q SearchQuery) ([]SearchHit, error) {
+	q.Text = strings.TrimSpace(q.Text)
+	if q.Text == "" {
+		return nil, fmt.Errorf("onlyoffice: empty search query")
+	}
+	body, err := json.Marshal(esSearchRequest(q, s.cfg.Tenant))
+	if err != nil {
+		return nil, fmt.Errorf("onlyoffice: build elasticsearch query: %w", err)
+	}
+	endpoint := s.cfg.URL + "/" + s.cfg.Index + "/_search"
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, endpoint, bytes.NewReader(body))
+	if err != nil {
+		return nil, err
+	}
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Accept", "application/json")
+	resp, err := s.http.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("onlyoffice: elasticsearch search: %w", err)
+	}
+	defer resp.Body.Close()
+	raw, err := io.ReadAll(io.LimitReader(resp.Body, maxESResponseSize))
+	if err != nil {
+		return nil, err
+	}
+	if resp.StatusCode >= 400 {
+		return nil, fmt.Errorf("onlyoffice: elasticsearch search: %d %s", resp.StatusCode, truncate(string(raw), 400))
+	}
+	return parseESSearchResponse(raw)
+}
+
+// esSearchRequest builds the ES query body. Pure, so it is unit-tested.
+func esSearchRequest(q SearchQuery, tenant string) esRequest {
+	limit := q.Limit
+	if limit <= 0 {
+		limit = defaultESLimit
+	}
+	if limit > maxESLimit {
+		limit = maxESLimit
+	}
+	fields := []string{"title^2"}
+	if q.InContent {
+		fields = append(fields, "document.attachment.content")
+	}
+	must := []esClause{{MultiMatch: &esMultiMatch{Query: q.Text, Fields: fields}}}
+
+	var filter []esClause
+	if t := strings.TrimSpace(tenant); t != "" {
+		filter = append(filter, esClause{Term: map[string]any{"tenantId": numericOrString(t)}})
+	}
+	if f := strings.TrimSpace(q.FolderID); f != "" {
+		filter = append(filter, esClause{Term: map[string]any{"folders.folderId": f}})
+	}
+	for _, ext := range normalizeExtensions(q.Extensions) {
+		filter = append(filter, esClause{Wildcard: map[string]any{"title": "*." + ext}})
+	}
+
+	highlightFields := map[string]struct{}{"title": {}}
+	if q.InContent {
+		highlightFields["document.attachment.content"] = struct{}{}
+	}
+	return esRequest{
+		Size:      limit,
+		Source:    []string{"id", "title", "folders"},
+		Query:     esQuery{Bool: esBool{Must: must, Filter: filter}},
+		Highlight: esHighlight{PreTags: []string{"<em>"}, PostTags: []string{"</em>"}, Fields: highlightFields},
+	}
+}
+
+// normalizeExtensions lowercases, trims leading dots and drops empties.
+func normalizeExtensions(exts []string) []string {
+	out := make([]string, 0, len(exts))
+	seen := map[string]bool{}
+	for _, e := range exts {
+		e = strings.ToLower(strings.TrimSpace(strings.TrimPrefix(strings.TrimSpace(e), ".")))
+		if e == "" || seen[e] {
+			continue
+		}
+		seen[e] = true
+		out = append(out, e)
+	}
+	return out
+}
+
+// numericOrString keeps an integer-looking filter value numeric (tenantId is
+// a long) and leaves anything else as a string (folderId is a text token).
+func numericOrString(s string) any {
+	if n, err := strconv.ParseInt(s, 10, 64); err == nil {
+		return n
+	}
+	return s
+}
+
+// esRequest is the subset of the ES query DSL this client emits.
+type esRequest struct {
+	Size      int         `json:"size"`
+	Source    []string    `json:"_source"`
+	Query     esQuery     `json:"query"`
+	Highlight esHighlight `json:"highlight"`
+}
+
+type esQuery struct {
+	Bool esBool `json:"bool"`
+}
+
+type esBool struct {
+	Must   []esClause `json:"must,omitempty"`
+	Filter []esClause `json:"filter,omitempty"`
+}
+
+type esClause struct {
+	MultiMatch *esMultiMatch  `json:"multi_match,omitempty"`
+	Term       map[string]any `json:"term,omitempty"`
+	Wildcard   map[string]any `json:"wildcard,omitempty"`
+}
+
+type esMultiMatch struct {
+	Query  string   `json:"query"`
+	Fields []string `json:"fields"`
+}
+
+type esHighlight struct {
+	PreTags  []string            `json:"pre_tags,omitempty"`
+	PostTags []string            `json:"post_tags,omitempty"`
+	Fields   map[string]struct{} `json:"fields"`
+}
+
+// esResponse is the subset of an ES search response we consume.
+type esResponse struct {
+	Took int `json:"took"`
+	Hits struct {
+		Total struct {
+			Value    int    `json:"value"`
+			Relation string `json:"relation"`
+		} `json:"total"`
+		Hits []esResponseHit `json:"hits"`
+	} `json:"hits"`
+}
+
+type esResponseHit struct {
+	ID     string  `json:"_id"`
+	Score  float64 `json:"_score"`
+	Source struct {
+		ID      int    `json:"id"`
+		Title   string `json:"title"`
+		Folders []struct {
+			FolderID string `json:"folderId"`
+			ID       int    `json:"id"`
+		} `json:"folders"`
+	} `json:"_source"`
+	Highlight map[string][]string `json:"highlight"`
+}
+
+// parseESSearchResponse converts an ES search response into SearchHit values.
+// Pure, so it is unit-tested.
+func parseESSearchResponse(raw []byte) ([]SearchHit, error) {
+	var r esResponse
+	if err := json.Unmarshal(raw, &r); err != nil {
+		return nil, fmt.Errorf("onlyoffice: decode elasticsearch response: %w", err)
+	}
+	hits := make([]SearchHit, 0, len(r.Hits.Hits))
+	for _, h := range r.Hits.Hits {
+		id := strconv.Itoa(h.Source.ID)
+		if h.Source.ID == 0 {
+			id = h.ID
+		}
+		var parent string
+		path := make([]string, 0, len(h.Source.Folders))
+		for i, f := range h.Source.Folders {
+			path = append(path, f.FolderID)
+			if i == 0 {
+				parent = f.FolderID
+			}
+		}
+		hits = append(hits, SearchHit{
+			Entry: Entry{
+				ID:       id,
+				ParentID: parent,
+				Title:    h.Source.Title,
+				Kind:     File,
+				Provider: "elasticsearch",
+			},
+			Score:     h.Score,
+			Highlight: esHighlightText(h.Highlight),
+			Path:      path,
+		})
+	}
+	return hits, nil
+}
+
+var esHighlightTag = regexp.MustCompile(`</?em[^>]*>`)
+
+// esHighlightText flattens a highlight map into one plain-text snippet,
+// preferring the content fragment over the title.
+func esHighlightText(hl map[string][]string) string {
+	for _, key := range []string{"document.attachment.content", "title"} {
+		frags := hl[key]
+		if len(frags) == 0 {
+			continue
+		}
+		clean := make([]string, 0, len(frags))
+		for _, f := range frags {
+			clean = append(clean, esHighlightTag.ReplaceAllString(f, ""))
+		}
+		return strings.Join(clean, " … ")
+	}
+	return ""
+}

file_es_integration_test.go

@@ -0,0 +1,135 @@
+//go:build integration
+
+package onlyoffice
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/xuri/excelize/v2"
+)
+
+// TestIntegrationESSearch uploads a throwaway workbook and verifies that the
+// direct Elasticsearch search finds it by file name and by content.
+//
+// The content index (document.attachment.content) is only populated for Office
+// formats (docx/xlsx/pptx), so the fixture is an xlsx whose cell carries a
+// unique token. Requires ONLYOFFICE_ES_URL (a reachable ES endpoint — in the
+// current setup a tunnel to the ES inside the OnlyOffice VM, see
+// docs/elasticsearch.md) plus the regular REST credentials for the upload.
+// Skips when either is missing.
+func TestIntegrationESSearch(t *testing.T) {
+	esURL := strings.TrimSpace(os.Getenv("ONLYOFFICE_ES_URL"))
+	if esURL == "" {
+		t.Skip("ONLYOFFICE_ES_URL not set — skipping Elasticsearch integration test")
+	}
+	c := liveClient(t)
+	t.Cleanup(func() { cleanupTestProjects(t, c) })
+
+	stamp := time.Now().UTC().Format("20060102-150405")
+	nameToken := "goesname" + stamp
+	contentToken := "goescontent" + stamp
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Minute)
+	defer cancel()
+
+	project, err := c.CreateProject(NewProjectRequest{
+		Title:       testProjectPrefix + "es-" + stamp,
+		Description: "go-onlyoffice elasticsearch integration",
+	})
+	if err != nil {
+		t.Fatalf("CreateProject: %v", err)
+	}
+	if project.ID == nil {
+		t.Fatal("created project without id")
+	}
+	pid := strconv.Itoa(*project.ID)
+
+	title := nameToken + ".xlsx"
+	localPath := filepath.Join(t.TempDir(), title)
+	book := excelize.NewFile()
+	if err := book.SetCellValue("Sheet1", "A1", "OnlyOffice Elasticsearch content fixture "+contentToken); err != nil {
+		t.Fatalf("SetCellValue: %v", err)
+	}
+	if err := book.SaveAs(localPath); err != nil {
+		t.Fatalf("SaveAs: %v", err)
+	}
+
+	entry, err := c.UploadProjectFile(ctx, pid, localPath)
+	if err != nil {
+		t.Fatalf("UploadProjectFile: %v", err)
+	}
+	fileID := strconv.Itoa(int(FileEntryNumericID(entry)))
+	if fileID == "0" {
+		t.Fatalf("upload returned no file id: %+v", entry)
+	}
+
+	es, err := NewESSearcher(ESConfig{
+		URL:    esURL,
+		Index:  os.Getenv("ONLYOFFICE_ES_INDEX"),
+		Tenant: os.Getenv("ONLYOFFICE_TENANT"),
+	})
+	if err != nil {
+		t.Fatalf("NewESSearcher: %v", err)
+	}
+
+	// Indexing is asynchronous on the server; poll until the file shows up.
+	// The server's title analyzer splits on whitespace, so the name query is
+	// the full file name token (including extension), as a user would type it.
+	nameHit := waitForHit(t, ctx, es, SearchQuery{Text: title}, fileID)
+	if nameHit.Title != title {
+		t.Errorf("name hit title = %q, want %q", nameHit.Title, title)
+	}
+	contentHit := waitForHit(t, ctx, es, SearchQuery{Text: contentToken, InContent: true}, fileID)
+	if contentHit.Highlight == "" {
+		t.Error("content hit has no highlight fragment")
+	}
+	if !strings.Contains(contentHit.Title, nameToken) {
+		t.Errorf("content hit title = %q, want the uploaded workbook", contentHit.Title)
+	}
+
+	// The content token is absent from the title, so a name-only search must
+	// not return the file — this proves the content field is really queried.
+	if hits := searchQuiet(t, es, SearchQuery{Text: contentToken}); len(hits) != 0 {
+		t.Errorf("name-only search for content token returned %d hits, want 0", len(hits))
+	}
+}
+
+// waitForHit polls ES until the file with fileID appears and returns that hit.
+func waitForHit(t *testing.T, ctx context.Context, s *ESSearcher, q SearchQuery, fileID string) SearchHit {
+	t.Helper()
+	var lastErr error
+	for {
+		hits, err := s.Search(ctx, q)
+		if err != nil {
+			lastErr = err
+		} else {
+			for _, h := range hits {
+				if h.ID == fileID {
+					return h
+				}
+			}
+		}
+		select {
+		case <-ctx.Done():
+			t.Fatalf("search %q: file %s not indexed in time (last err: %v)", q.Text, fileID, lastErr)
+		case <-time.After(3 * time.Second):
+		}
+	}
+}
+
+func searchQuiet(t *testing.T, s *ESSearcher, q SearchQuery) []SearchHit {
+	t.Helper()
+	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	defer cancel()
+	hits, err := s.Search(ctx, q)
+	if err != nil {
+		t.Fatalf("Search(%q): %v", q.Text, err)
+	}
+	return hits
+}

file_es_test.go

@@ -0,0 +1,188 @@
+package onlyoffice
+
+import (
+	"encoding/json"
+	"reflect"
+	"testing"
+)
+
+func TestESSearchRequestNameOnly(t *testing.T) {
+	got := esSearchRequest(SearchQuery{Text: "Rechnung"}, "1")
+	if got.Size != defaultESLimit {
+		t.Errorf("size = %d, want %d", got.Size, defaultESLimit)
+	}
+	if !reflect.DeepEqual(got.Source, []string{"id", "title", "folders"}) {
+		t.Errorf("_source = %v", got.Source)
+	}
+	if len(got.Query.Bool.Must) != 1 || got.Query.Bool.Must[0].MultiMatch == nil {
+		t.Fatalf("must = %+v, want one multi_match", got.Query.Bool.Must)
+	}
+	mm := got.Query.Bool.Must[0].MultiMatch
+	if mm.Query != "Rechnung" {
+		t.Errorf("query = %q", mm.Query)
+	}
+	if !reflect.DeepEqual(mm.Fields, []string{"title^2"}) {
+		t.Errorf("fields = %v, want title only", mm.Fields)
+	}
+	if _, ok := got.Highlight.Fields["document.attachment.content"]; ok {
+		t.Error("content highlight present without InContent")
+	}
+	if _, ok := got.Highlight.Fields["title"]; !ok {
+		t.Error("title highlight missing")
+	}
+	if len(got.Query.Bool.Filter) != 1 || got.Query.Bool.Filter[0].Term["tenantId"] != int64(1) {
+		t.Errorf("tenant filter = %+v, want numeric tenantId=1", got.Query.Bool.Filter)
+	}
+}
+
+func TestESSearchRequestContentFields(t *testing.T) {
+	got := esSearchRequest(SearchQuery{Text: "Mahnung", InContent: true}, "")
+	mm := got.Query.Bool.Must[0].MultiMatch
+	want := []string{"title^2", "document.attachment.content"}
+	if !reflect.DeepEqual(mm.Fields, want) {
+		t.Errorf("fields = %v, want %v", mm.Fields, want)
+	}
+	if _, ok := got.Highlight.Fields["document.attachment.content"]; !ok {
+		t.Error("content highlight missing with InContent")
+	}
+	if len(got.Query.Bool.Filter) != 0 {
+		t.Errorf("filter = %+v, want none without tenant/folder", got.Query.Bool.Filter)
+	}
+}
+
+func TestESSearchRequestFiltersAndLimit(t *testing.T) {
+	got := esSearchRequest(SearchQuery{
+		Text:       "Storchen",
+		FolderID:   "649",
+		Extensions: []string{".PDF", "pdf", "docx"},
+		Limit:      999,
+	}, "42")
+	if got.Size != maxESLimit {
+		t.Errorf("size = %d, want cap %d", got.Size, maxESLimit)
+	}
+	var tenant, folder, wildcards int
+	for _, f := range got.Query.Bool.Filter {
+		switch {
+		case f.Term != nil && f.Term["tenantId"] != nil:
+			tenant++
+		case f.Term != nil && f.Term["folders.folderId"] != nil:
+			folder++
+			if f.Term["folders.folderId"] != "649" {
+				t.Errorf("folder filter = %+v", f.Term)
+			}
+		case f.Wildcard != nil:
+			wildcards++
+		}
+	}
+	if tenant != 1 || folder != 1 {
+		t.Errorf("term filters tenant=%d folder=%d, want 1 each", tenant, folder)
+	}
+	if wildcards != 2 {
+		t.Errorf("wildcard filters = %d, want deduped PDF+docx", wildcards)
+	}
+}
+
+func TestESSearchRequestRejectsEmptyTextAtSearch(t *testing.T) {
+	s, err := NewESSearcher(ESConfig{URL: "http://localhost:9200"})
+	if err != nil {
+		t.Fatalf("NewESSearcher: %v", err)
+	}
+	if _, err := s.Search(t.Context(), SearchQuery{Text: "  "}); err == nil {
+		t.Error("empty query: want error")
+	}
+}
+
+func TestNewESSearcherRequiresURL(t *testing.T) {
+	if _, err := NewESSearcher(ESConfig{}); err == nil {
+		t.Error("empty URL: want error")
+	}
+	s, err := NewESSearcher(ESConfig{URL: "http://es:9200/"})
+	if err != nil {
+		t.Fatalf("NewESSearcher: %v", err)
+	}
+	if s.cfg.Index != defaultESIndex {
+		t.Errorf("index = %q, want %q", s.cfg.Index, defaultESIndex)
+	}
+	if s.cfg.URL != "http://es:9200" {
+		t.Errorf("url = %q, want trimmed", s.cfg.URL)
+	}
+	if s.Name() != "elasticsearch" {
+		t.Errorf("Name() = %q", s.Name())
+	}
+}
+
+func TestNormalizeExtensions(t *testing.T) {
+	got := normalizeExtensions([]string{" .PDF ", "pdf", "", "xlsx"})
+	want := []string{"pdf", "xlsx"}
+	if !reflect.DeepEqual(got, want) {
+		t.Errorf("normalizeExtensions = %v, want %v", got, want)
+	}
+}
+
+func TestParseESSearchResponse(t *testing.T) {
+	raw := []byte(`{
+	  "took": 12,
+	  "hits": {
+	    "total": {"value": 2, "relation": "eq"},
+	    "hits": [
+	      {
+	        "_id": "2395",
+	        "_score": 7.31,
+	        "_source": {"id": 2395, "title": "Rechnung-4711.pdf",
+	          "folders": [{"folderId": "438", "id": 0}, {"folderId": "11", "id": 0}]},
+	        "highlight": {
+	          "title": ["<em>Rechnung</em>-4711.pdf"],
+	          "document.attachment.content": ["… Zahlung der <em>Rechnung</em> …"]
+	        }
+	      },
+	      {
+	        "_id": "2318",
+	        "_score": 6.02,
+	        "_source": {"id": 2318, "title": "Mahnung.pdf", "folders": []},
+	        "highlight": {"title": ["<em>Mahnung</em>.pdf"]}
+	      }
+	    ]
+	  }
+	}`)
+	hits, err := parseESSearchResponse(raw)
+	if err != nil {
+		t.Fatalf("parseESSearchResponse: %v", err)
+	}
+	if len(hits) != 2 {
+		t.Fatalf("hits = %d, want 2", len(hits))
+	}
+	h0 := hits[0]
+	if h0.ID != "2395" || h0.Title != "Rechnung-4711.pdf" || h0.Kind != File {
+		t.Errorf("hit0 entry = %+v", h0.Entry)
+	}
+	if h0.ParentID != "438" || !reflect.DeepEqual(h0.Path, []string{"438", "11"}) {
+		t.Errorf("hit0 path = %v parent = %q", h0.Path, h0.ParentID)
+	}
+	if h0.Score != 7.31 {
+		t.Errorf("hit0 score = %v", h0.Score)
+	}
+	if h0.Highlight != "… Zahlung der Rechnung …" {
+		t.Errorf("hit0 highlight = %q, want content fragment", h0.Highlight)
+	}
+	if hits[1].Highlight != "Mahnung.pdf" {
+		t.Errorf("hit1 highlight = %q, want title without tags", hits[1].Highlight)
+	}
+	if hits[1].ParentID != "" || len(hits[1].Path) != 0 {
+		t.Errorf("hit1 path = %v", hits[1].Path)
+	}
+}
+
+func TestESSearchRequestJSONShape(t *testing.T) {
+	got := esSearchRequest(SearchQuery{Text: "Rechnung", InContent: true}, "1")
+	b, err := json.Marshal(got)
+	if err != nil {
+		t.Fatalf("marshal: %v", err)
+	}
+	var back map[string]any
+	if err := json.Unmarshal(b, &back); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if _, ok := back["query"].(map[string]any)["bool"]; !ok {
+		t.Errorf("query.bool missing: %s", b)
+	}
+}

files.go

@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"io"
+	"net/http"
 	"net/url"
 	"path"
 	"strconv"
@@ -357,20 +358,31 @@ func (c *Client) UploadToFolder(ctx context.Context, folderID, localPath string)
 
 // UpdateFile uploads a new version of an existing file (same id, name and
 // folder). It does not delete and does not create a second file.
+//
+// The Documents API method is PUT /api/2.0/files/{id}/update; POST is kept as
+// a fallback for older servers. The path is tried with and without .json.
 func (c *Client) UpdateFile(ctx context.Context, fileID, localPath string) (*FileEntry, error) {
 	if fileID == "" || localPath == "" {
 		return nil, fmt.Errorf("file id and local path are required")
 	}
-	uploadPath := fmt.Sprintf("/api/2.0/files/%s/update", url.PathEscape(fileID))
-	raw, err := c.uploadMultipart(ctx, uploadPath, "file", localPath)
-	if err != nil {
-		uploadPath = fmt.Sprintf("/api/2.0/files/%s/update.json", url.PathEscape(fileID))
-		raw, err = c.uploadMultipart(ctx, uploadPath, "file", localPath)
-		if err != nil {
-			return nil, err
-		}
+	base := fmt.Sprintf("/api/2.0/files/%s/update", url.PathEscape(fileID))
+	attempts := []struct {
+		method, path string
+	}{
+		{http.MethodPut, base},
+		{http.MethodPut, base + ".json"},
+		{http.MethodPost, base},
+		{http.MethodPost, base + ".json"},
 	}
-	return decodeResponseFileEntry(raw)
+	var lastErr error
+	for _, a := range attempts {
+		raw, err := c.uploadMultipartMethod(ctx, a.method, a.path, "file", localPath)
+		if err == nil {
+			return decodeResponseFileEntry(raw)
+		}
+		lastErr = err
+	}
+	return nil, lastErr
 }
 
 // FileFolderID returns the parent folder id string for a file entry, if known.

files_webdav.go

@@ -410,12 +410,12 @@ func (f *DavFolder) UnmarshalJSON(b []byte) error {
 // UnmarshalJSON decodes a file row, capturing size and timestamps.
 func (f *DavFile) UnmarshalJSON(b []byte) error {
 	var raw struct {
-		ID        *json.Number `json:"id"`
-		Title     *string      `json:"title"`
-		PureSize  *int64       `json:"pureContentLength"`
-		SizeStr   *string      `json:"contentLength"`
-		Updated   *string      `json:"updated"`
-		ViewURL   *string      `json:"viewUrl"`
+		ID       *json.Number `json:"id"`
+		Title    *string      `json:"title"`
+		PureSize *int64       `json:"pureContentLength"`
+		SizeStr  *string      `json:"contentLength"`
+		Updated  *string      `json:"updated"`
+		ViewURL  *string      `json:"viewUrl"`
 	}
 	if err := json.Unmarshal(b, &raw); err != nil {
 		return err

http.go

@@ -346,6 +346,14 @@ func (c *Client) putJSON(ctx context.Context, path string, body any) (json.RawMe
 
 // uploadMultipart posts a single file to path under the given form field name.
 func (c *Client) uploadMultipart(ctx context.Context, path, fieldName, filePath string) (json.RawMessage, error) {
+	return c.uploadMultipartMethod(ctx, http.MethodPost, path, fieldName, filePath)
+}
+
+// uploadMultipartMethod sends a single-file multipart request with the given
+// HTTP method. The OnlyOffice Documents API needs PUT for /update (a new
+// version) and POST for /upload (a new file); sending POST to /update answers
+// 500 on current servers.
+func (c *Client) uploadMultipartMethod(ctx context.Context, method, path, fieldName, filePath string) (json.RawMessage, error) {
 	auth, err := c.authHeader()
 	if err != nil {
 		return nil, err
@@ -368,7 +376,7 @@ func (c *Client) uploadMultipart(ctx context.Context, path, fieldName, filePath
 	if err := mw.Close(); err != nil {
 		return nil, err
 	}
-	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.baseURL()+path, &buf)
+	req, err := http.NewRequestWithContext(ctx, method, c.baseURL()+path, &buf)
 	if err != nil {
 		return nil, err
 	}

retry.go

@@ -0,0 +1,60 @@
+package onlyoffice
+
+import (
+	"context"
+	"regexp"
+	"time"
+)
+
+// RetryPolicy controls deterministic retries against OnlyOffice: fixed linear
+// backoff without jitter, so repeated runs wait exactly the same schedule.
+// OnlyOffice throttles bulk reads/writes with 429 (and occasional 502/503/504
+// from openresty), so every bulk tool routes API calls through DoRetry.
+type RetryPolicy struct {
+	Attempts int           // total attempts, including the first try
+	Base     time.Duration // wait before retry N is N*Base
+	Max      time.Duration // per-wait cap
+}
+
+// DefaultRetryPolicy retries up to 5 times with 1s, 2s, 3s, 4s waits.
+func DefaultRetryPolicy() RetryPolicy {
+	return RetryPolicy{Attempts: 5, Base: time.Second, Max: 30 * time.Second}
+}
+
+var transientRe = regexp.MustCompile(`:\s*(429|502|503|504)\b`)
+
+// Transient reports whether err looks like a transient OnlyOffice answer
+// (an HTTP 429/502/503/504 surfaced as "...: <code> ...").
+func Transient(err error) bool {
+	if err == nil {
+		return false
+	}
+	return transientRe.MatchString(err.Error())
+}
+
+// DoRetry runs fn until it succeeds, fails non-transiently, or attempts run
+// out. Waits are deterministic: N*Base capped at Max, no jitter.
+func DoRetry(ctx context.Context, p RetryPolicy, fn func() error) error {
+	if p.Attempts < 1 {
+		p.Attempts = 1
+	}
+	var err error
+	for attempt := 1; attempt <= p.Attempts; attempt++ {
+		if ctx.Err() != nil {
+			return ctx.Err()
+		}
+		if err = fn(); err == nil || !Transient(err) || attempt == p.Attempts {
+			return err
+		}
+		wait := time.Duration(attempt) * p.Base
+		if wait > p.Max {
+			wait = p.Max
+		}
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-time.After(wait):
+		}
+	}
+	return err
+}