omnidesk-mcp 0.1.0__tar.gz

omnidesk_mcp-0.1.0/.gitignore

@@ -0,0 +1,19 @@
+!desktop/src-tauri/resources/.gitignore
+!desktop/src-tauri/resources/README.txt
+*.bak_*
+*.log
+*.pyc
+.DS_Store
+.env
+.venv/
+/data/
+__pycache__/
+apps/site/dist/
+apps/web/dist/
+desktop/.cache/
+desktop/src-tauri/gen/schemas/
+desktop/src-tauri/resources/*
+desktop/src-tauri/target/
+dist/
+node_modules/
+tools/private_key.b64

omnidesk_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,50 @@
+Metadata-Version: 2.5
+Name: omnidesk-mcp
+Version: 0.1.0
+Summary: OmniDesk MCP server — let AI agents convert files across 325 formats
+Project-URL: Homepage, https://omnidesk.win
+Project-URL: Documentation, https://omnidesk.win/en/api/
+Author-email: Projectteamforyou <projectteamforyou@gmail.com>
+License: MIT
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.27
+Requires-Dist: mcp>=2.0
+Description-Content-Type: text/markdown
+
+# omnidesk-mcp
+
+An MCP server that lets Claude, Cursor and other agents convert files through
+[OmniDesk](https://omnidesk.win) — 325 formats, routed through intermediate
+formats when no direct converter exists.
+
+## Setup
+
+You need an OmniDesk API key (Team plan). Create one at
+**app.omnidesk.win → Control panel → API keys**.
+
+Add to your MCP client config (Claude Desktop: `claude_desktop_config.json`):
+
+```json
+{
+  "mcpServers": {
+    "omnidesk": {
+      "command": "uvx",
+      "args": ["omnidesk-mcp"],
+      "env": { "OMNIDESK_API_KEY": "od_..." }
+    }
+  }
+}
+```
+
+## Tools
+
+| tool | what it does |
+|---|---|
+| `convert_file` | convert a local file to another format; returns the result path |
+| `list_formats` | the 325 supported formats by category |
+| `find_route` | how a conversion would be routed, and alternatives |
+| `file_action` | one of 33 file actions: extract tables, redact, OCR, trim, … |
+| `list_actions` | the actions and their options |
+
+Files are uploaded to your company drive, converted there, and the result is
+downloaded back to a path you choose. Nothing is kept on the client beyond that.

omnidesk_mcp-0.1.0/README.md

@@ -0,0 +1,37 @@
+# omnidesk-mcp
+
+An MCP server that lets Claude, Cursor and other agents convert files through
+[OmniDesk](https://omnidesk.win) — 325 formats, routed through intermediate
+formats when no direct converter exists.
+
+## Setup
+
+You need an OmniDesk API key (Team plan). Create one at
+**app.omnidesk.win → Control panel → API keys**.
+
+Add to your MCP client config (Claude Desktop: `claude_desktop_config.json`):
+
+```json
+{
+  "mcpServers": {
+    "omnidesk": {
+      "command": "uvx",
+      "args": ["omnidesk-mcp"],
+      "env": { "OMNIDESK_API_KEY": "od_..." }
+    }
+  }
+}
+```
+
+## Tools
+
+| tool | what it does |
+|---|---|
+| `convert_file` | convert a local file to another format; returns the result path |
+| `list_formats` | the 325 supported formats by category |
+| `find_route` | how a conversion would be routed, and alternatives |
+| `file_action` | one of 33 file actions: extract tables, redact, OCR, trim, … |
+| `list_actions` | the actions and their options |
+
+Files are uploaded to your company drive, converted there, and the result is
+downloaded back to a path you choose. Nothing is kept on the client beyond that.

omnidesk_mcp-0.1.0/omnidesk_mcp/__init__.py

@@ -0,0 +1 @@
+"""OmniDesk MCP server."""

omnidesk_mcp-0.1.0/omnidesk_mcp/server.py

@@ -0,0 +1,263 @@
+"""OmniDesk MCP 서버.
+
+AI 에이전트가 파일을 다룰 때 "이걸 PDF 로 바꿔" 라고 하면 실제로 바꿔 주는 도구.
+Claude Desktop·Cursor·기타 MCP 클라이언트가 stdio 로 이 프로그램을 띄우고,
+여기서는 OmniDesk API 를 부른다. 로컬 파일을 올려 변환하고 결과를 지정한 경로에
+받아 온다.
+
+키는 환경변수 OMNIDESK_API_KEY 로 받는다. 클라이언트 설정 파일에 적어 두면
+된다. 키가 없으면 도구 목록은 보이되 부르면 그 이유를 말한다 — 설정이 잘못됐을
+때 조용히 실패하는 것이 가장 나쁘다.
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import os
+import sys
+import time
+from pathlib import Path
+from typing import Any
+
+import httpx
+from mcp.server.mcpserver import MCPServer
+
+BASE = os.environ.get("OMNIDESK_BASE", "https://app.omnidesk.win").rstrip("/")
+KEY = os.environ.get("OMNIDESK_API_KEY", "").strip()
+
+server = MCPServer("omnidesk",
+                   instructions=("Convert files across 325 formats and run 33 file actions through OmniDesk. "
+                                 "Files are uploaded to the user's company drive and the result is written "
+                                 "back to a local path."))
+
+
+def _client() -> httpx.AsyncClient:
+    if not KEY:
+        raise RuntimeError(
+            "OMNIDESK_API_KEY is not set. Create a key at app.omnidesk.win → Control panel → "
+            "API keys (Team plan), then put it in the MCP server's env.")
+    return httpx.AsyncClient(base_url=BASE, headers={"X-API-Key": KEY}, timeout=120)
+
+
+async def _wait_job(c: httpx.AsyncClient, job_id: str, timeout: float = 600) -> dict:
+    """변환이 끝날 때까지. 큰 파일은 몇 분 걸릴 수 있다."""
+    t0 = time.time()
+    delay = 1.0
+    while time.time() - t0 < timeout:
+        r = await c.get(f"/api/jobs/{job_id}")
+        r.raise_for_status()
+        j = r.json()
+        if j["status"] in ("done", "failed"):
+            return j
+        await asyncio.sleep(delay)
+        delay = min(delay * 1.4, 6.0)
+    raise TimeoutError(f"job {job_id} did not finish in {timeout:.0f}s")
+
+
+async def _download(c: httpx.AsyncClient, node_id: str, out_path: Path) -> Path:
+    r = await c.get(f"/api/drive/{node_id}/download")
+    r.raise_for_status()
+    url = r.json()["url"]
+    async with httpx.AsyncClient(timeout=300, follow_redirects=True) as raw:
+        d = await raw.get(url)
+        d.raise_for_status()
+        out_path.parent.mkdir(parents=True, exist_ok=True)
+        out_path.write_bytes(d.content)
+    return out_path
+
+
+def _quality_lines(job: dict) -> str:
+    q = (job.get("quality") or {}).get("changes") or []
+    if not q:
+        return ""
+    return "\n".join(("⚠ " if x["level"] == "warn" else "· ") + x["text"] for x in q)
+
+
+# ───────────────────────────────────────────── tools
+def _err(e: Exception) -> str:
+    if isinstance(e, httpx.HTTPStatusError):
+        code = e.response.status_code
+        if code == 401:
+            return "OmniDesk returned 401: API key rejected. Check OMNIDESK_API_KEY."
+        if code == 402:
+            return "OmniDesk returned 402: this account's plan does not include API access (Team plan required)."
+        if code == 429:
+            return "OmniDesk returned 429: monthly conversion quota exhausted."
+        return f"OmniDesk returned {code}: {e.response.text[:400]}"
+    return f"error: {e}"
+
+
+@server.tool()
+async def convert_file(path: str, to: str, output_path: str | None = None,
+                       route: list[str] | None = None) -> str:
+    """Convert a local file to another format through OmniDesk.
+
+    325 formats; when no direct converter exists the file is routed through
+    intermediate formats. Returns the output path plus a ledger of what changed
+    (lost transparency, dropped audio track, substituted fonts, ...).
+
+    Args:
+        path: Local path of the file to convert.
+        to: Target format, e.g. pdf, jpg, mp3.
+        output_path: Where to write the result. Defaults to the source path with the new extension.
+        route: Optional engine list from find_route to force a specific path.
+    """
+    try:
+        return await _convert({"path": path, "to": to, "output_path": output_path, "route": route})
+    except Exception as e:
+        return _err(e)
+
+
+@server.tool()
+async def file_action(path: str, action: str, options: dict | None = None,
+                      output_path: str | None = None) -> str:
+    """Run one of OmniDesk's 33 file actions on a local file.
+
+    Extract tables from a PDF to xlsx, redact phone numbers in an image, OCR a
+    scanned PDF, extract audio or subtitles from a video, trim, compress, add
+    signature fields, and more. Use list_actions to see codes and options.
+
+    Args:
+        path: Local path of the file.
+        action: Action code, e.g. tables, redact, pdf_ocr, extract_audio.
+        options: Action options as key/value.
+        output_path: Where to write the result.
+    """
+    try:
+        return await _action({"path": path, "action": action, "options": options or {}, "output_path": output_path})
+    except Exception as e:
+        return _err(e)
+
+
+@server.tool()
+async def find_route(src: str, dst: str) -> str:
+    """Show how OmniDesk would convert one format to another, including alternative routes and the engines involved.
+
+    Args:
+        src: Source format, e.g. docx.
+        dst: Target format, e.g. jpg.
+    """
+    try:
+        return await _route({"src": src, "dst": dst})
+    except Exception as e:
+        return _err(e)
+
+
+@server.tool()
+async def list_formats() -> str:
+    """List the supported formats, grouped by category."""
+    try:
+        return await _formats()
+    except Exception as e:
+        return _err(e)
+
+
+@server.tool()
+async def list_actions() -> str:
+    """List the available file actions with their accepted inputs and options."""
+    try:
+        return await _actions()
+    except Exception as e:
+        return _err(e)
+
+
+async def _convert(a: dict) -> str:
+    src = Path(a["path"]).expanduser()
+    if not src.is_file():
+        return f"no such file: {src}"
+    to = str(a["to"]).lower().lstrip(".")
+    out = Path(a.get("output_path") or src.with_suffix("." + to)).expanduser()
+    opts = {}
+    if a.get("route"):
+        opts["route"] = list(a["route"])
+    async with _client() as c:
+        with src.open("rb") as fh:
+            r = await c.post("/api/jobs/quick",
+                             files={"file": (src.name, fh, "application/octet-stream")},
+                             data={"dst_fmt": to, "options": json.dumps(opts)})
+        r.raise_for_status()
+        job = (r.json().get("jobs") or [r.json()["job"]])[0]
+        job = await _wait_job(c, job["id"])
+        if job["status"] != "done":
+            return f"conversion failed: {job.get('error') or 'unknown error'}"
+        await _download(c, job["result_node_id"], out)
+    lines = [f"converted {src.name} → {out}",
+             f"route: {(job.get('route') or {}).get('chain', '')}"]
+    ql = _quality_lines(job)
+    if ql:
+        lines.append("what changed:\n" + ql)
+    return "\n".join(lines)
+
+
+async def _action(a: dict) -> str:
+    src = Path(a["path"]).expanduser()
+    if not src.is_file():
+        return f"no such file: {src}"
+    action = str(a["action"])
+    opts = dict(a.get("options") or {})
+    async with _client() as c:
+        with src.open("rb") as fh:
+            up = await c.post("/api/drive/upload",
+                              files={"file": (src.name, fh, "application/octet-stream")},
+                              data={"scope": "company"})
+        up.raise_for_status()
+        node_id = up.json()["id"]
+        r = await c.post("/api/jobs/action",
+                         json={"action": action, "node_ids": [node_id], "options": opts, "scope": "company"})
+        r.raise_for_status()
+        job = await _wait_job(c, r.json()["created"][0]["id"])
+        if job["status"] != "done":
+            return f"action failed: {job.get('error') or 'unknown error'}"
+        ext = job.get("dst_fmt") or src.suffix.lstrip(".")
+        out = Path(a.get("output_path") or src.with_name(f"{src.stem}.{action}.{ext}")).expanduser()
+        await _download(c, job["result_node_id"], out)
+    return f"{action} on {src.name} → {out}"
+
+
+async def _route(a: dict) -> str:
+    async with _client() as c:
+        r = await c.get("/api/catalog/routes", params={"src": a["src"], "dst": a["dst"]})
+        if r.status_code == 404:
+            return f"no route from {a['src']} to {a['dst']}"
+        r.raise_for_status()
+        items = r.json()["items"]
+    lines = [f"{a['src']} → {a['dst']}: {len(items)} route(s)"]
+    for i, it in enumerate(items):
+        tag = " (default)" if i == 0 else ""
+        lines.append(f"  {it['hops']} step(s): {it['chain']}{tag}  engines={it['engines']}")
+    return "\n".join(lines)
+
+
+async def _formats() -> str:
+    async with _client() as c:
+        r = await c.get("/api/catalog/formats")
+        r.raise_for_status()
+        d = r.json()
+    cats = d.get("cats") or d.get("categories") or {}
+    lines = []
+    for cat, items in cats.items():
+        ids = [x["id"] if isinstance(x, dict) else x for x in items]
+        lines.append(f"{cat} ({len(ids)}): {' '.join(ids)}")
+    return "\n".join(lines) or json.dumps(d)[:2000]
+
+
+async def _actions() -> str:
+    async with _client() as c:
+        r = await c.get("/api/catalog/actions", params={"lang": "en"})
+        r.raise_for_status()
+        acts = r.json().get("actions", [])
+    lines = []
+    for x in acts:
+        opts = ", ".join(f"{o['key']}" + (f"={o.get('default')}" if o.get("default") not in (None, "") else "")
+                         for o in x.get("options", []))
+        lines.append(f"{x['code']}: {x.get('en') or x.get('name', '')} · inputs={len(x.get('inputs', []))}"
+                     + (f" · options: {opts}" if opts else ""))
+    return "\n".join(lines)
+
+
+def main():
+    server.run()   # stdio
+
+
+if __name__ == "__main__":
+    main()

omnidesk_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,23 @@
+[project]
+name = "omnidesk-mcp"
+version = "0.1.0"
+description = "OmniDesk MCP server — let AI agents convert files across 325 formats"
+readme = "README.md"
+requires-python = ">=3.10"
+license = { text = "MIT" }
+authors = [{ name = "Projectteamforyou", email = "projectteamforyou@gmail.com" }]
+dependencies = ["mcp>=2.0", "httpx>=0.27"]
+
+[project.scripts]
+omnidesk-mcp = "omnidesk_mcp.server:main"
+
+[project.urls]
+Homepage = "https://omnidesk.win"
+Documentation = "https://omnidesk.win/en/api/"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.build.targets.wheel]
+packages = ["omnidesk_mcp"]