EVEMISSTechnology

Repository 指南 · 架構 · v2

microsoft / markitdown · 架構

這是一個 Python repository,在 packages/ 下有四個套件:包含格式轉換器的核心 markitdown 函式庫,以及 MCP、OCR 和 sample-plugin 套件。靜態呼叫度數指出 _markitdown.py 是核心樞紐。列出的執行路徑只涵蓋 MCP 進入點,而且很快就停在未解析的外部呼叫。

原始 repository
microsoft/markitdown
授權
MIT · 開源授權
分析的版本 · 最後驗證
ed178952a6672abdfd35b5f9e76a80b6cefc086e ·

本頁中文由負責的 AI 編輯依英文正式版本 v2 翻譯;程式碼名稱、路徑、行號與數字都經確定性檢查,與英文版一致。陳述與依據表保留英文原文,因為那是獨立驗證者核對過的紀錄。看英文原文

microsoft/markitdown 架構:它是如何組織的

本頁用語說明
Entrypoint record(進入點記錄)
分析器標記為「程式可能從這裡開始執行」的檔案。附有 __main__ guard 摘錄時,表示檔案裡有 if __name__ == "__main__": 區塊,摘錄顯示它呼叫了什麼;沒有摘錄時,只是依檔名判斷。
Bounded static execution path(有界靜態執行路徑)
不執行程式、直接從原始碼重建出的呼叫鏈,走到固定步數就停。它說明在紙面上能追到多遠,不代表實際執行時會發生什麼。
Unresolved boundary(未解析邊界)
靜態路徑停下的地方:下一個呼叫進入外部函式庫,或不執行就無法確定。這是本次分析的邊界,不是 repository 的缺陷。
Static relations(靜態關係)
在原始碼中找到的呼叫與 import。「external or unresolved」表示指向分析範圍以外。
Module role(模組角色)
分析器依呼叫進出次數推論的檔案標籤(例如 core、entry/orchestration、leaf)。它描述的是在呼叫圖中的位置,不是作者的設計意圖。
Test-like files(類測試檔案)
名稱或位置看起來像測試的檔案。本次分析只計數,不執行。
Observed · inferred · author-claimed · unresolved
每句話的依據:直接讀自分析的檔案;由分析器從檔案推論;repository 作者自述(metadata、README);或本次分析無法確定。
Verified(已驗證)
本頁有兩種用法。在分析器的記錄裡(「verified provenance」「verified entrypoint」),它表示這筆記錄是直接從分析的檔案讀到的,也就是本指南所說的 observed;以進入點來說,就是檔案裡確實有 __main__ guard 的文字。分析器在 2026-10-04 修正後,只有 cli.py、main.py 這類像進入點的檔名、沒有 guard 的檔案,會記為 inferred(推測);在那之前分析的指南仍把這類檔案標成 verified,它依然只是對檔案用途的推測。它不代表程式被執行或測試過。「最後驗證」是這份指南最後一次通過本實驗室檢查的日期,檢查對象是所標示版本的分析記錄,不是對 repository 本身的審查。

整體樣貌

分析器清點了 197 個檔案:181 個在 packages/ 下,10 個在根目錄,5 個在 .github,1 個在 .devcontainer。其中 112 個被它算作類似測試的檔案。分析資料中沒有子系統或模組記錄,所以以下各部分是從 manifest 和檔案路徑讀出的。

  • packages/markitdown 是核心函式庫。它的 src/markitdown 目錄樹包含 _markitdown.py、一個 converters/ 套件和 converter_utils/ 輔助模組。
  • packages/markitdown-mcp 有自己的 __main__.py。
  • packages/markitdown-ocr 包含 _plugin.py 和 _ocr_service.py。
  • packages/markitdown-sample-plugin 有一個帶有 __main__ 守衛的測試檔案。

這四個目錄各有自己的 pyproject.toml。Repository metadata 把這個專案描述為一個將檔案和 office 文件轉換成 Markdown 的 Python 工具。

進入點與控制流程

分析器列出分布在 22 個檔案中的 24 筆 __main__ 守衛記錄。其中只有兩個是套件模組:packages/markitdown/src/markitdown/__main__.py 和 packages/markitdown-mcp/src/markitdown_mcp/__main__.py。其餘都位於 tests/ 下,包括兩個 fixture 產生器。這兩個套件的守衛都只是呼叫 main()。

列出的十條執行路徑全部從 MCP 套件的 __main__.py 開始。沒有一條被截斷,也沒有一條含有循環,而且每一條都以終止原因 unresolved_boundary 結束。其中八條路徑先經過 main(),然後停在 argparse.ArgumentParser、parser.add_argument、parser.parse_args、parser.error 或 sys.exit。另外兩條在模組層級停在 logging.getLogger 和 MCPServer。

重建結果計有 155 條有界路徑,但分析資料中只有十條。這十條中沒有一條從核心套件的 __main__.py 開始,所以核心命令列流程沒有呈現出來。分析器的摘要選擇 MCP 的 __main__.py 作為主要起點;這是分析器的判斷,不是執行時期的觀察。

核心模組與推論出的角色

以下角色是分析器根據已解析的跨模組呼叫次數和進入點歸屬推論出來的。它們不是來自解讀程式碼的意圖。

  • _markitdown.py 是最明顯的樞紐,有 33 個外向呼叫和 4 個內向呼叫;它被標記為 service/core 候選。
  • 被大量呼叫的葉節點模組:_base_converter.py(21 個內向呼叫)、converters/__init__.py(20)、_exceptions.py(19)和 _stream_info.py(14)。
  • converter_utils/docx/math/omml.py 發出 25 個外向呼叫,latex_dict.py 接收 25 個。相同的數字暗示(但不能證明)兩者是一組呼叫端/查表的配對。
  • xlsx、pptx、docx、pdf、audio、epub 和 Outlook msg 的格式轉換器都沒有任何已解析的內向呼叫,被標記為 orchestration 候選。它們很可能是透過分析器無法解析的 dispatch 被呼叫。
  • _html_converter.py(6 個內向、4 個外向)、_markdownify.py(6 個內向)和 converter_utils/_image.py(3 個內向)是較小的共用節點。
  • markitdown/__main__.py 是 entry/orchestration 候選。在 OCR 套件中,_plugin.py 有 5 個外向呼叫,_ocr_service.py 有 4 個內向呼叫。

靜態路徑停在哪裡

每一條列出的路徑都停在外部或未解析的目標:標準函式庫的 logging、argparse 的 parser 物件、sys.exit 或 MCPServer。沒有任何列出的路徑進入核心 markitdown 套件。

在整個 repository 中,6,816 個靜態關係裡有 5,812 個是外部或未解析的,只有 1,004 個在本地解析。

MCPServer 未解析。markitdown-mcp 套件把 mcp 宣告為執行時期依賴,合理推測可能是由它提供,但分析資料沒有顯示這個 import。

外掛的載入方式看不到。OCR 套件有一個 _plugin.py 模組,但分析資料沒有確定核心套件在執行時期如何找到並載入它。動態 import 不在分析器的範圍內。

各部分之間的依賴

Manifest 記錄顯示哪些套件建立在核心套件之上:

  • markitdown-mcp、markitdown-ocr 和 markitdown-sample-plugin 各自宣告對 markitdown 的執行時期依賴。前兩者把它限制在 >=0.1.9,<0.2.0。
  • 核心套件的執行時期清單很短:beautifulsoup4、requests、certifi、markdownify、magika、charset-normalizer 和 defusedxml。
  • 在核心 manifest 中,格式函式庫是選用的:python-pptx、mammoth、pandas、openpyxl、xlrd、lxml、pdfminer.six、pdfplumber、olefile、pydub、SpeechRecognition、youtube-transcript-api 和 Azure 套件。其中有幾個名稱出現在不只一筆記錄中。
  • markitdown-ocr 在執行時期需要 pdfminer.six、pdfplumber、PyMuPDF、mammoth、python-docx、python-pptx、pandas、openpyxl 和 Pillow,openai 則是選用。
  • markitdown-mcp 加入 mcp,sample plugin 加入 striprtf。

在靜態層面,分析器擷取了 925 個 import 關係和 5,891 個呼叫關係。依賴記錄只來自 requirements 檔案、package.json 和 pyproject.toml。setup.py、setup.cfg、Pipfile 和 lock 檔案等檔案不會被解析。

靜態分析無法呈現的部分

  • **執行時期的串接:**分析器不會解析 reflection、dependency injection、動態 import、monkey-patching、產生的程式碼、框架串接或動態 dispatch。
  • **設定與指令:**安裝、執行和建置的推論並不完整,也沒有任何安裝指令經過驗證。這裡不提供任何指令。
  • **涵蓋範圍的限制:**超過 400 kB 的檔案、minified bundle 和 vendored 目錄會被清點,但不會被解析。關係擷取在每個檔案和每個 repository 都有上限。
  • **把測試當作行為:**分析器計有 112 個類似測試的檔案,這暗示至少有一部分預期行為被寫在測試中。靜態分析只把它們當作程式碼讀取,不會回報它們是否通過。

閱讀架構的順序

  1. packages/markitdown/src/markitdown/__main__.py,核心進入點候選。
  2. _markitdown.py,度數最高的樞紐。
  3. _base_converter.py 和 _stream_info.py,兩個被大量呼叫的葉節點模組。
  4. packages/markitdown-mcp/src/markitdown_mcp/__main__.py,接著是 OCR 套件的 _plugin.py 和 _ocr_service.py,看看外層套件如何建立在核心之上。

這次分析無法確定的事

這些是撰寫模型對分析記錄自己的說明。lim_3、exec_* 或 claim_… 這類識別碼指的是那次分析裡的記錄;陳述與依據表引用的也是同一批記錄。

  • 分析資料只列出重建結果回報的 155 條有界執行路徑中的 10 條,而且這十條都從 MCP 的 __main__.py 開始;核心 CLI 和測試進入點的流程沒有呈現。
  • 分析器的其中一項教學陳述指出沒有可用的有界靜態執行路徑,這和列出的十筆執行路徑記錄互相矛盾;本文依據的是那些記錄。
  • 分析資料中的子系統和模組清單是空的,所以套件分組是從路徑讀出的,不是來自分析器偵測到的子系統。
  • 分析資料中的靜態核心流程、邊界和已解析關係樣本摘要沒有可引用的 ID,所以像是 main 呼叫 create_starlette_app,或 OCR 的 __init__.py import register_converters 這類細節,都沒有被當作陳述使用。
  • 有多少已解析關係跨越套件邊界(例如從 markitdown-ocr 到 markitdown)尚未確定;只提供了整個 repository 的總數。
  • packages/markitdown/pyproject.toml 中許多選用依賴名稱出現在行號相同的重複記錄中;它們屬於哪些選用群組尚未確定。
陳述與依據 — 36 條陳述,36 條經獨立驗證者確認(英文原文)

Every substantive statement above is a claim bound to grounding IDs of the analyzed revision. Global grounding IDs are namespaced by the analysis run.

Claim Epistemic status Verifier Grounding (global IDs)
c1 The analyzer reports 197 files: 181 under packages/, 10 at the root, 5 in .github and 1 in .devcontainer, with 112 test-like files. inferred supported analysis_5d9c9ae9a129bff5:summary_repository
c2 Dependency records come from four pyproject.toml manifests, located in packages/markitdown, packages/markitdown-mcp, packages/markitdown-ocr and packages/markitdown-sample-plugin. observed supported analysis_5d9c9ae9a129bff5:dep_1, analysis_5d9c9ae9a129bff5:dep_3, analysis_5d9c9ae9a129bff5:dep_14, analysis_5d9c9ae9a129bff5:dep_16
c3 The packet carries no subsystem or module records, so any grouping by package is read from manifest and file paths; analyzer-detected subsystems are not established. unresolved supported analysis_5d9c9ae9a129bff5:dep_1, analysis_5d9c9ae9a129bff5:dep_3, analysis_5d9c9ae9a129bff5:dep_14, analysis_5d9c9ae9a129bff5:dep_16
c4 Module-role records place _markitdown.py, a converters/ package and converter_utils/ helpers inside packages/markitdown/src/markitdown. inferred supported analysis_5d9c9ae9a129bff5:role_13, analysis_5d9c9ae9a129bff5:role_22, analysis_5d9c9ae9a129bff5:role_20, analysis_5d9c9ae9a129bff5:role_16
c5 markitdown-mcp has its own __main__.py, markitdown-ocr holds _plugin.py and _ocr_service.py, and markitdown-sample-plugin has a test file with a __main__ guard. inferred supported analysis_5d9c9ae9a129bff5:py_entry_2f852a8015d4, analysis_5d9c9ae9a129bff5:role_6, analysis_5d9c9ae9a129bff5:role_4, analysis_5d9c9ae9a129bff5:py_entry_a63fdf323f3b
c6 The repository metadata describes the project as a Python tool for converting files and office documents to Markdown. author_claimed supported analysis_5d9c9ae9a129bff5:meta_description
c7 There are 24 __main__ guard records in 22 files. Two files are package modules (markitdown/__main__.py and markitdown_mcp/__main__.py); the others are under tests/, including two fixture generators in tests/test_files/. observed supported analysis_5d9c9ae9a129bff5:py_entry_2f852a8015d4, analysis_5d9c9ae9a129bff5:py_entry_1fa8ac31a8f1, analysis_5d9c9ae9a129bff5:py_entry_a63fdf323f3b, analysis_5d9c9ae9a129bff5:py_entry_d19b256901ab, analysis_5d9c9ae9a129bff5:py_entry_f44307e3b0be, analysis_5d9c9ae9a129bff5:py_entry_007b88c947c9, analysis_5d9c9ae9a129bff5:py_entry_48d5677c0b8c, analysis_5d9c9ae9a129bff5:py_entry_6da5faf654d3, analysis_5d9c9ae9a129bff5:py_entry_e1644548438f, analysis_5d9c9ae9a129bff5:py_entry_fd2581a2a85a, analysis_5d9c9ae9a129bff5:py_entry_157267349a98, analysis_5d9c9ae9a129bff5:py_entry_8a3faafff1be, analysis_5d9c9ae9a129bff5:py_entry_c1e9a0696ede, analysis_5d9c9ae9a129bff5:py_entry_4001ed8b5544, analysis_5d9c9ae9a129bff5:py_entry_2eddfb03d1cb, analysis_5d9c9ae9a129bff5:py_entry_791d98438517, analysis_5d9c9ae9a129bff5:py_entry_3c032b18438d, analysis_5d9c9ae9a129bff5:py_entry_32796e5c789a, analysis_5d9c9ae9a129bff5:py_entry_d48046c8aa24, analysis_5d9c9ae9a129bff5:py_entry_f19b18e9138b, analysis_5d9c9ae9a129bff5:py_entry_de1243d2bd26, analysis_5d9c9ae9a129bff5:py_entry_fc521ad126e1, analysis_5d9c9ae9a129bff5:py_entry_1f6c94e0f30f, analysis_5d9c9ae9a129bff5:py_entry_98eaed3ae984
c8 Both package __main__ guards consist of a call to main(). observed supported analysis_5d9c9ae9a129bff5:py_entry_2f852a8015d4, analysis_5d9c9ae9a129bff5:py_entry_d19b256901ab
c9 All ten listed execution paths start at markitdown_mcp/__main__.py. None is truncated, none has a cycle, and each ends with terminal reason unresolved_boundary. inferred supported analysis_5d9c9ae9a129bff5:exec_67998e701fae, analysis_5d9c9ae9a129bff5:exec_b68e1f5a7c8a, analysis_5d9c9ae9a129bff5:exec_11e5ff5e359f, analysis_5d9c9ae9a129bff5:exec_51ee1388edd7, analysis_5d9c9ae9a129bff5:exec_b64608cd49a8, analysis_5d9c9ae9a129bff5:exec_5be36f4ab619, analysis_5d9c9ae9a129bff5:exec_a37ea1d8706f, analysis_5d9c9ae9a129bff5:exec_763ce56ab606, analysis_5d9c9ae9a129bff5:exec_b6c46e3ecba7, analysis_5d9c9ae9a129bff5:exec_a46c678eda14
c10 Eight of the ten paths pass through local main() before stopping at argparse.ArgumentParser, parser.add_argument, parser.parse_args, parser.error or sys.exit. Two stop at module level on logging.getLogger and MCPServer. inferred supported analysis_5d9c9ae9a129bff5:exec_67998e701fae, analysis_5d9c9ae9a129bff5:exec_b68e1f5a7c8a, analysis_5d9c9ae9a129bff5:exec_11e5ff5e359f, analysis_5d9c9ae9a129bff5:exec_51ee1388edd7, analysis_5d9c9ae9a129bff5:exec_b64608cd49a8, analysis_5d9c9ae9a129bff5:exec_5be36f4ab619, analysis_5d9c9ae9a129bff5:exec_a37ea1d8706f, analysis_5d9c9ae9a129bff5:exec_763ce56ab606, analysis_5d9c9ae9a129bff5:exec_b6c46e3ecba7, analysis_5d9c9ae9a129bff5:exec_a46c678eda14
c11 The reconstruction reports 155 bounded paths, but only ten are listed. None of the ten starts at packages/markitdown/src/markitdown/__main__.py, so the core CLI flow is not shown in this packet. inferred supported analysis_5d9c9ae9a129bff5:architecture_reconstruction, analysis_5d9c9ae9a129bff5:exec_67998e701fae, analysis_5d9c9ae9a129bff5:exec_b68e1f5a7c8a, analysis_5d9c9ae9a129bff5:exec_11e5ff5e359f, analysis_5d9c9ae9a129bff5:exec_51ee1388edd7, analysis_5d9c9ae9a129bff5:exec_b64608cd49a8, analysis_5d9c9ae9a129bff5:exec_5be36f4ab619, analysis_5d9c9ae9a129bff5:exec_a37ea1d8706f, analysis_5d9c9ae9a129bff5:exec_763ce56ab606, analysis_5d9c9ae9a129bff5:exec_b6c46e3ecba7, analysis_5d9c9ae9a129bff5:exec_a46c678eda14
c12 The analyzer's summary names markitdown_mcp/__main__.py as the main starting place; this is an analyzer judgement rather than a runtime observation. inferred supported analysis_5d9c9ae9a129bff5:summary_repository
c13 All listed module roles are inferred from static resolved cross-module call degree and entrypoint membership, not from the intent of the code. inferred supported analysis_5d9c9ae9a129bff5:role_13, analysis_5d9c9ae9a129bff5:role_20, analysis_5d9c9ae9a129bff5:role_19, analysis_5d9c9ae9a129bff5:role_11, analysis_5d9c9ae9a129bff5:role_22, analysis_5d9c9ae9a129bff5:role_12, analysis_5d9c9ae9a129bff5:role_14, analysis_5d9c9ae9a129bff5:role_31, analysis_5d9c9ae9a129bff5:role_43, analysis_5d9c9ae9a129bff5:role_39, analysis_5d9c9ae9a129bff5:role_35, analysis_5d9c9ae9a129bff5:role_6, analysis_5d9c9ae9a129bff5:role_4, analysis_5d9c9ae9a129bff5:role_10, analysis_5d9c9ae9a129bff5:role_28, analysis_5d9c9ae9a129bff5:role_16, analysis_5d9c9ae9a129bff5:role_23, analysis_5d9c9ae9a129bff5:role_29, analysis_5d9c9ae9a129bff5:role_36, analysis_5d9c9ae9a129bff5:role_37
c14 _markitdown.py has 33 outgoing and 4 incoming resolved calls and is tagged as a service/core candidate, which makes it the clearest hub among the listed roles. inferred supported analysis_5d9c9ae9a129bff5:role_13
c15 _base_converter.py (21 incoming), converters/__init__.py (20), _exceptions.py (19) and _stream_info.py (14) have no outgoing calls and are tagged as leaf/data-boundary candidates. inferred supported analysis_5d9c9ae9a129bff5:role_11, analysis_5d9c9ae9a129bff5:role_22, analysis_5d9c9ae9a129bff5:role_12, analysis_5d9c9ae9a129bff5:role_14
c16 omml.py makes 25 outgoing calls and latex_dict.py receives 25 incoming calls. The matching counts suggest, but do not establish, that omml.py is the main caller of latex_dict.py. inferred supported analysis_5d9c9ae9a129bff5:role_20, analysis_5d9c9ae9a129bff5:role_19
c17 The xlsx, pptx, docx, pdf, audio, epub and Outlook msg converters have zero resolved incoming calls and are tagged as orchestration candidates. They are probably invoked through dispatch the analyzer does not resolve. inferred supported analysis_5d9c9ae9a129bff5:role_43, analysis_5d9c9ae9a129bff5:role_39, analysis_5d9c9ae9a129bff5:role_28, analysis_5d9c9ae9a129bff5:role_37, analysis_5d9c9ae9a129bff5:role_23, analysis_5d9c9ae9a129bff5:role_29, analysis_5d9c9ae9a129bff5:role_36, analysis_5d9c9ae9a129bff5:lim_1
c18 _html_converter.py (6 incoming, 4 outgoing), _markdownify.py (6 incoming) and converter_utils/_image.py (3 incoming) are smaller shared nodes in the call graph. inferred supported analysis_5d9c9ae9a129bff5:role_31, analysis_5d9c9ae9a129bff5:role_35, analysis_5d9c9ae9a129bff5:role_16
c19 markitdown/__main__.py is an entry/orchestration candidate with 4 outgoing calls. In the OCR package, _plugin.py has 5 outgoing calls and _ocr_service.py has 4 incoming calls. inferred supported analysis_5d9c9ae9a129bff5:role_10, analysis_5d9c9ae9a129bff5:role_6, analysis_5d9c9ae9a129bff5:role_4
c20 Every listed path stops at an external-or-unresolved target (logging, argparse parser objects, sys.exit or MCPServer). No listed path reaches into the core markitdown package. inferred supported analysis_5d9c9ae9a129bff5:exec_67998e701fae, analysis_5d9c9ae9a129bff5:exec_b68e1f5a7c8a, analysis_5d9c9ae9a129bff5:exec_11e5ff5e359f, analysis_5d9c9ae9a129bff5:exec_51ee1388edd7, analysis_5d9c9ae9a129bff5:exec_b64608cd49a8, analysis_5d9c9ae9a129bff5:exec_5be36f4ab619, analysis_5d9c9ae9a129bff5:exec_a37ea1d8706f, analysis_5d9c9ae9a129bff5:exec_763ce56ab606, analysis_5d9c9ae9a129bff5:exec_b6c46e3ecba7, analysis_5d9c9ae9a129bff5:exec_a46c678eda14
c21 Of 6,816 static relations, 5,812 are external or unresolved and 1,004 resolve locally. observed supported analysis_5d9c9ae9a129bff5:relation_counts
c22 MCPServer is an unresolved target. markitdown-mcp declares mcp as a runtime dependency, which plausibly supplies it, but the packet does not show the import. inferred supported analysis_5d9c9ae9a129bff5:exec_b68e1f5a7c8a, analysis_5d9c9ae9a129bff5:dep_1
c23 How the core package discovers and loads the OCR package's _plugin.py at runtime is not established, because the analyzer does not resolve dynamic imports. unresolved supported analysis_5d9c9ae9a129bff5:role_6, analysis_5d9c9ae9a129bff5:lim_1
c24 markitdown-mcp, markitdown-ocr and markitdown-sample-plugin each declare a runtime dependency on markitdown. The first two use the constraint >=0.1.9,<0.2.0. observed supported analysis_5d9c9ae9a129bff5:dep_2, analysis_5d9c9ae9a129bff5:dep_3, analysis_5d9c9ae9a129bff5:dep_14
c25 The core markitdown manifest lists beautifulsoup4, requests, certifi, markdownify, magika, charset-normalizer and defusedxml as runtime dependencies. observed supported analysis_5d9c9ae9a129bff5:dep_16, analysis_5d9c9ae9a129bff5:dep_17, analysis_5d9c9ae9a129bff5:dep_18, analysis_5d9c9ae9a129bff5:dep_19, analysis_5d9c9ae9a129bff5:dep_20, analysis_5d9c9ae9a129bff5:dep_21, analysis_5d9c9ae9a129bff5:dep_22
c26 Format libraries such as python-pptx, mammoth, pandas, openpyxl, xlrd, lxml, pdfminer.six, pdfplumber, olefile, pydub, SpeechRecognition, youtube-transcript-api and Azure packages are optional in the core manifest. Several names recur across records. observed supported analysis_5d9c9ae9a129bff5:dep_23, analysis_5d9c9ae9a129bff5:dep_24, analysis_5d9c9ae9a129bff5:dep_25, analysis_5d9c9ae9a129bff5:dep_26, analysis_5d9c9ae9a129bff5:dep_27, analysis_5d9c9ae9a129bff5:dep_28, analysis_5d9c9ae9a129bff5:dep_29, analysis_5d9c9ae9a129bff5:dep_30, analysis_5d9c9ae9a129bff5:dep_31, analysis_5d9c9ae9a129bff5:dep_32, analysis_5d9c9ae9a129bff5:dep_33, analysis_5d9c9ae9a129bff5:dep_34, analysis_5d9c9ae9a129bff5:dep_35, analysis_5d9c9ae9a129bff5:dep_36, analysis_5d9c9ae9a129bff5:dep_37, analysis_5d9c9ae9a129bff5:dep_38, analysis_5d9c9ae9a129bff5:dep_39, analysis_5d9c9ae9a129bff5:dep_40, analysis_5d9c9ae9a129bff5:dep_41, analysis_5d9c9ae9a129bff5:dep_42, analysis_5d9c9ae9a129bff5:dep_43, analysis_5d9c9ae9a129bff5:dep_44, analysis_5d9c9ae9a129bff5:dep_45, analysis_5d9c9ae9a129bff5:dep_46, analysis_5d9c9ae9a129bff5:dep_47, analysis_5d9c9ae9a129bff5:dep_48, analysis_5d9c9ae9a129bff5:dep_49, analysis_5d9c9ae9a129bff5:dep_50, analysis_5d9c9ae9a129bff5:dep_51, analysis_5d9c9ae9a129bff5:dep_52, analysis_5d9c9ae9a129bff5:dep_53, analysis_5d9c9ae9a129bff5:dep_54
c27 markitdown-ocr lists pdfminer.six, pdfplumber, PyMuPDF, mammoth, python-docx, python-pptx, pandas, openpyxl and Pillow as runtime dependencies, and openai as optional. observed supported analysis_5d9c9ae9a129bff5:dep_4, analysis_5d9c9ae9a129bff5:dep_5, analysis_5d9c9ae9a129bff5:dep_6, analysis_5d9c9ae9a129bff5:dep_7, analysis_5d9c9ae9a129bff5:dep_8, analysis_5d9c9ae9a129bff5:dep_9, analysis_5d9c9ae9a129bff5:dep_10, analysis_5d9c9ae9a129bff5:dep_11, analysis_5d9c9ae9a129bff5:dep_12, analysis_5d9c9ae9a129bff5:dep_13
c28 markitdown-mcp adds mcp as a runtime dependency, and markitdown-sample-plugin adds striprtf. observed supported analysis_5d9c9ae9a129bff5:dep_1, analysis_5d9c9ae9a129bff5:dep_15
c29 The analyzer extracted 925 import relations and 5,891 call relations. observed supported analysis_5d9c9ae9a129bff5:relation_counts
c30 Dependency records come only from requirements files, package.json and pyproject.toml. setup.py, setup.cfg, Pipfile and lock files are not parsed. observed supported analysis_5d9c9ae9a129bff5:lim_3a
c31 The analyzer does not resolve reflection, runtime dependency injection, dynamic imports, monkey-patching, generated code, framework runtime wiring or dynamic dispatch. observed supported analysis_5d9c9ae9a129bff5:lim_1
c32 Install, run and build inference is partial, and installation commands are not verified. observed supported analysis_5d9c9ae9a129bff5:lim_2
c33 Files over 400 kB, minified bundles and vendored directories are inventoried but not parsed. Relation extraction is capped at 2500 per file and 150000 per repository. observed supported analysis_5d9c9ae9a129bff5:lim_6
c34 The analyzer counts 112 test-like files, which suggests at least some expected behaviour is encoded in tests; static analysis reads them only as code and does not report whether they pass. inferred supported analysis_5d9c9ae9a129bff5:summary_repository, analysis_5d9c9ae9a129bff5:lim_1
c35 A suggested start is the core entry candidate markitdown/__main__.py, then the hub _markitdown.py, then the heavily called leaf modules _base_converter.py and _stream_info.py. inferred supported analysis_5d9c9ae9a129bff5:py_entry_d19b256901ab, analysis_5d9c9ae9a129bff5:role_10, analysis_5d9c9ae9a129bff5:role_13, analysis_5d9c9ae9a129bff5:role_11, analysis_5d9c9ae9a129bff5:role_14
c36 After the core modules, reading markitdown_mcp/__main__.py and the OCR package's _plugin.py and _ocr_service.py shows how the outer packages sit on top of the core. inferred supported analysis_5d9c9ae9a129bff5:py_entry_2f852a8015d4, analysis_5d9c9ae9a129bff5:role_6, analysis_5d9c9ae9a129bff5:role_4, analysis_5d9c9ae9a129bff5:dep_2, analysis_5d9c9ae9a129bff5:dep_3

來源、權利與說明 · attribution-license-templates/v0.1

權利聲明。原始 repository 託管於 GitHub。其原始碼、文件、名稱、媒體與相關素材的權利,仍屬各自的作者、貢獻者與其他權利人所有,並受該 repository 的授權條款約束。

Original repository hosted on GitHub. Repository source code, documentation, names, media, and related project materials remain subject to the rights of their respective authors, contributors, and other rights holders and to applicable repository license terms.

平台聲明。GitHub 是所連結 repository 的來源託管平台。GitHub 與相關標誌為 GitHub, Inc. 的商標。除非另有明確說明,EVEMISS Technology 與 GitHub 之間沒有隸屬或背書關係。

GitHub is the source hosting platform for the linked repository. GitHub and related marks are trademarks of GitHub, Inc. EVEMISS Technology is not affiliated with or endorsed by GitHub unless explicitly stated otherwise.

這一頁怎麼來的。本頁由對應版本的 repository 分析與 AI 輔助的編輯工具產生。技術陳述綁定所分析的版本;原始 repository 更新後,可能重新驗證。

This page was produced using revision-aware repository analysis and AI-assisted editorial tooling. Technical claims are tied to the analyzed repository revision and may be revalidated when the source repository changes.

本頁包含 AI 輔助分析,並經 EVEMISS Technology 人工與 AI 協力審閱。 · 回報權利疑慮 · 回到概覽