MCP server for the BHt (Biblia Hebraica transcripta) Hebrew Bible database
Project description
bht-mcp
MCP server for the BHt (Biblia Hebraica transcripta) Hebrew Bible database at LMU Munich.
Ask your AI assistant to search, analyze, and compare 489,000+ morphologically analyzed Hebrew Bible tokens โ directly from your chat interface.
You: "Compare the use of root BRสพ (create) in Genesis vs Isaiah"
AI: [calls bht_search with Wurzel filter for each book, then bht_token_detail
on selected tokens โ returns morphological breakdowns, verbal stems,
and syntactic contexts from both books]
No Hebrew expertise is needed to get started. The AI reads the linguistic data and explains it to you.
About BHt
BHt (Biblia Hebraica transcripta) is a digital transcription of the Hebrew Bible maintained at LMU Munich under the direction of Prof. Wolfgang Richter. It features:
- 489,437 tokens with full morphological analysis
- Richter transcription system โ Latin-script phonemic representation (e.g.,
barฤ(สพ)for ืึธึผืจึธื) - 5-class part-of-speech system (Hauptwortart, Nebenwortart, Fรผgwortart, etc.)
- Syntactic trees at word-level (Wortfรผgungsebene) and sentence-level (Satzfรผgungsebene)
- Textual criticism annotations with Greek/Masoretic manuscript variant data
- 8 Sirach manuscript fragments (Ben Sira, from Cairo Genizah and Masada)
Table of Contents
- About BHt
- How It Works
- Installation & Setup
- Setting Up Your MCP Client
- Research Scenarios
- Tool Reference
- Rate Limits
- License
- ๐ฐ๐ท ํ๊ตญ์ด
How It Works
โโโโโโโโโโโโโโโโ stdio โโโโโโโโโโโโโโโโโโโโ cache miss โโโโโโโโโโโโโโโโ
โ MCP Client โ โโโโโโโโโ bht-mcp โ โโโโโโโโโโโโโโ BHt Website โ
โ (Claude, โ โโโโโโโโโ (local process) โ โโโโโโโโโโโโโโ (LMU Munich) โ
โ local LLM, โ โ โโโโโโโโโโโโโโ โ โโโโโโโโโโโโโโโโ
โ etc.) โ โ โ ~/.bht/ โ โ
โโโโโโโโโโโโโโโโ โ โ cache.db โ โ โ local SQLite cache
โ โโโโโโโโโโโโโโ โ
โโโโโโโโโโโโโโโโโโโโ
- Runs entirely on your machine. No remote server, no API keys, no cost.
- Progressive caching. The first search for a book fetches all its tokens in one request. After that, searches in the same book are instant โ zero network calls.
- Respectful. Built-in rate limits (1 req/s, 150 HTML pages/day) protect the university server. The tool never bulk-scrapes.
Installation & Setup
Option A: Using uv (recommended)
uv is the standard tool for running MCP servers. No separate install step needed โ just configure your client.
Install uv (one-time):
| Platform | Command |
|---|---|
| macOS | brew install uv |
| Windows | powershell -c "irm https://astral.sh/uv/install.ps1 | iex" |
| Linux | curl -LsSf https://astral.sh/uv/install.sh | sh |
Then go directly to Setting Up Your MCP Client below.
Option B: Using pip
If you prefer a traditional install:
pip install bht-mcp
Requires Python 3.11+. If Python is not installed: python.org/downloads (Windows: check "Add Python to PATH"). macOS:
brew install python. Linux:sudo apt install python3 python3-pip.
Setting Up Your MCP Client
MCP (Model Context Protocol) lets AI assistants use external tools. Choose your client below:
Claude Desktop
Edit your MCP settings (Settings โ Developer โ MCP Servers) and add:
If you installed with uv:
{
"mcpServers": {
"bht": {
"command": "uvx",
"args": ["bht-mcp"]
}
}
}
If you installed with pip:
{
"mcpServers": {
"bht": {
"command": "python",
"args": ["-m", "bht_mcp"]
}
}
}
Claude Code
claude mcp add bht -- uvx bht-mcp
Or with pip:
claude mcp add bht -- python -m bht_mcp
Local LLMs (Open WebUI, llama.cpp, Ollama, etc.)
If your local LLM setup supports MCP, configure it to launch bht-mcp as a stdio subprocess:
With uv:
{
"mcpServers": {
"bht": {
"command": "uvx",
"args": ["bht-mcp"]
}
}
}
With pip:
{
"mcpServers": {
"bht": {
"command": "python",
"args": ["-m", "bht_mcp"]
}
}
}
Note: MCP tool calling requires a model that supports function/tool calling. Most 7B+ instruction-tuned models work (Llama 3, Mistral, Qwen, etc.).
Other MCP-compatible clients
Any client that supports MCP stdio transport can use bht-mcp. Use uvx bht-mcp or python -m bht_mcp as the server command.
After adding, restart your client. You should see 7 BHt tools available.
Research Scenarios
You don't need to know the tool names or parameters โ just describe what you want in natural language. The AI figures out which tools to call.
Read a verse
"Show me Genesis 1:1 word by word with grammatical analysis"
The AI finds all tokens in the verse, then retrieves morphology for each:
barฤ(สพ) โ Suffixkonjugation, 3rd person masc. singular, Qal stem, root BRสพ
สพรฏlล*hฤซm โ Substantiv, masc. plural, base สพl-h
สพศงt โ Praeposition
ha=ลกamaym โ Substantiv with article, masc. plural
Explore a root
"Find all forms of root สพMR (say/speak) in the Psalms"
The AI searches for all tokens with this root in the specified book and presents the results grouped by form.
Check manuscript variants
"Are there textual variants in Genesis 1?"
โ Gen 1,7c โ "kin": differs in Greek (G); G adds 7c text
โ Gen 1,9b โ "maqลm": G reads miqwว, MT reads miqwฤ-m?
Verbal stem distribution
"What verbal stems appear in Ecclesiastes, and how frequent are they?"
The AI searches for all verbs in Kohelet, retrieves their stem information, and presents frequency statistics:
โ G (Qal): 312, D (Piel): 45, H (Hiphil): 67, N (Niphal): 28, ...
Compare sentence structure
"Compare the syntactic structure of Gen 1:1 and the opening of Isaiah 40:1"
The AI retrieves sentence-level analysis for both verses and highlights structural differences โ sentence type, syntagm patterns, and clause relationships.
Find rare verb forms
"Show me all Hophal (HP) passive verbs in the Torah"
The AI searches each Torah book for Hophal stems and compiles a comprehensive list across Genesis through Deuteronomy.
Identify Aramaic sections
"Which tokens in Daniel are marked as Aramaic rather than Hebrew?"
The AI discovers the language codes available in BHt, then filters Daniel's tokens by language โ revealing the Aramaic sections (chapters 2:4bโ7:28).
Hapax legomena search
"Find lexemes that appear only once in the Hebrew Bible"
The AI iterates through lexemes from a book, checking each one's frequency across the entire corpus. Lexemes with exactly one occurrence are flagged as hapax legomena.
Cross-book morphological comparison
"Compare the construct chain patterns (Constructus) between Proverbs and Job"
The AI searches both books for construct-state nouns, then compares frequency, lexeme variety, and syntactic positions to reveal stylistic differences.
Syntax tree comparison for parallel passages
"Compare the word-level syntax of Psalm 18 and 2 Samuel 22 (parallel texts)"
The AI retrieves syntactic trees for matching verses in both books and highlights structural divergences between the two textual traditions.
Ben Sira manuscript comparison
"Which chapters of Sirach are preserved in manuscript A vs manuscript B?"
The AI queries the book list to show manuscript coverage:
โ ASir: chapters 2โ16,23,27 (17 chapters)
โ BSir: chapters 10โ11,15โ16,20,30โ51 (26 chapters)
โ Overlap: chapters 10,11,15,16 โ compare these
Tool Reference
| Tool | Purpose | BHt Requests |
|---|---|---|
| bht_list_books | List all 47 books with codes and chapter counts | 0 |
| bht_field_info | Get valid values for any of 42 search fields | 0โ1 |
| bht_search | Search tokens by location or morphological filters | 0โ1 |
| bht_token_detail | Full morphological analysis of a single token | 0โ1 |
| bht_syntax_tree | Word-level syntactic tree (Wortfรผgungsebene) | 0โ2 |
| bht_sentence_analysis | Sentence-level analysis (Satzfรผgungsebene) | 0โ2 |
| bht_text_annotations | Textual criticism annotations for a chapter | 0โ1 |
Typical workflow: bht_list_books โ bht_search โ bht_token_detail โ bht_syntax_tree
Every response includes a quota field showing daily usage:
{
"data": [...],
"quota": {"daily_html_used": 12, "daily_html_limit": 150, "daily_html_remaining": 138}
}
Rate Limits
This tool accesses a university research server. Built-in limits protect it:
| Limit | Value | Reason |
|---|---|---|
| Request interval | 1 req/s | Server capacity protection |
| Daily HTML requests | 150/day | Token detail, syntax, sentence, annotation pages |
| Daily JSON requests | Unlimited | Search API is lightweight |
These limits are comparable to a researcher manually browsing the site (typically 25โ100 pages/day). The daily limit resets at midnight.
When the limit is reached, cached data remains fully accessible โ only new HTML fetches are blocked.
License
Code: MIT
BHt data: CC BY-SA 4.0 (LMU Munich)
๐ฐ๐ท ํ๊ตญ์ด
AI ์ด์์คํดํธ์๊ฒ ํ๋ธ๋ฆฌ์ด ์ฑ๊ฒฝ 489,000๊ฐ ์ด์์ ํํ๋ก ๋ถ์ ํ ํฐ์ ๊ฒ์, ๋ถ์, ๋น๊ตํ๋๋ก ์์ฒญํ์ธ์ โ ์ฑํ ์ธํฐํ์ด์ค์์ ๋ฐ๋ก.
์ฌ์ฉ์: "์ฐฝ์ธ๊ธฐ์ ์ด์ฌ์ผ์์ ์ด๊ทผ BRสพ(์ฐฝ์กฐํ๋ค)์ ์ฌ์ฉ์ ๋น๊ตํด์ค"
AI: [๊ฐ ์ฑ
์์ ์ด๊ทผ ํํฐ๋ก bht_search๋ฅผ ํธ์ถํ๊ณ , ์ ํ๋ ํ ํฐ์ ๋ํด
bht_token_detail์ ํธ์ถ โ ๋ ์ฑ
์ ํํ๋ก ๋ถ์, ๋์ฌ ์ด๊ฐ,
ํต์ฌ์ ๋งฅ๋ฝ์ ๋ฐํํฉ๋๋ค]
ํ๋ธ๋ฆฌ์ด ์ ๋ฌธ ์ง์ ์์ด๋ ์์ํ ์ ์์ต๋๋ค. AI๊ฐ ์ธ์ดํ ๋ฐ์ดํฐ๋ฅผ ์ฝ๊ณ ์ค๋ช ํด์ค๋๋ค.
BHt ์๊ฐ
BHt (Biblia Hebraica transcripta)๋ ๋ฎํจ ๋ํ๊ต(LMU Munich)์์ Wolfgang Richter ๊ต์์ ์ง๋ ์๋ ๊ด๋ฆฌ๋๋ ํ๋ธ๋ฆฌ์ด ์ฑ๊ฒฝ ๋์งํธ ์ ์ฌ๋ณธ์ ๋๋ค.
- 489,437๊ฐ ํ ํฐ โ ์์ ํ ํํ๋ก ๋ถ์ ํฌํจ
- Richter ์ ์ฌ ์ฒด๊ณ โ ๋ผํด ๋ฌธ์ ๊ธฐ๋ฐ ์์ ํ๊ธฐ (์:
barฤ(สพ)= ืึธึผืจึธื) - 5๋ถ๋ฅ ํ์ฌ ์ฒด๊ณ (Hauptwortart, Nebenwortart, Fรผgwortart ๋ฑ)
- ํต์ฌ ํธ๋ฆฌ โ ๋จ์ด ์์ค(Wortfรผgungsebene) ๋ฐ ๋ฌธ์ฅ ์์ค(Satzfรผgungsebene)
- ํ ์คํธ ๋นํ ์ฃผ์ โ ๊ทธ๋ฆฌ์ค์ด/๋ง์๋ผ ์ฌ๋ณธ ์ด๋ณธ ๋ฐ์ดํฐ
- ์๋ผํฌ ์ฌ๋ณธ ๋จํธ 8์ข (๋ฒค ์๋ผ, ์นด์ด๋ก ๊ฒ๋์ ๋ฐ ๋ง์ฌ๋ค ์ถํ )
์๋ ๋ฐฉ์
โโโโโโโโโโโโโโโโ stdio โโโโโโโโโโโโโโโโโโโโ cache miss โโโโโโโโโโโโโโโโ
โ MCP Client โ โโโโโโโโโ bht-mcp โ โโโโโโโโโโโโโโ BHt Website โ
โ (Claude, โ โโโโโโโโโ (local process) โ โโโโโโโโโโโโโโ (LMU Munich) โ
โ local LLM, โ โ โโโโโโโโโโโโโโ โ โโโโโโโโโโโโโโโโ
โ etc.) โ โ โ ~/.bht/ โ โ
โโโโโโโโโโโโโโโโ โ โ cache.db โ โ โ local SQLite cache
โ โโโโโโโโโโโโโโ โ
โโโโโโโโโโโโโโโโโโโโ
- ์์ ํ ๋ก์ปฌ์์ ์คํ๋ฉ๋๋ค. ์๊ฒฉ ์๋ฒ ์์, API ํค ์์, ๋น์ฉ ์์.
- ์ ์ง์ ์บ์ฑ. ํน์ ์ฑ ์ ์ฒ์ ๊ฒ์ํ๋ฉด ํด๋น ์ฑ ์ ์ ์ฒด ํ ํฐ์ 1ํ ์์ฒญ์ผ๋ก ๊ฐ์ ธ์ต๋๋ค. ์ดํ ๊ฐ์ ์ฑ ์ ๊ฒ์์ ์ฆ์ โ ๋คํธ์ํฌ ์์ฒญ 0ํ.
- ์๋ฒ ๋ณดํธ. ๋ด์ฅ ์๋ ์ ํ(์ด๋น 1์์ฒญ, ์ผ์ผ HTML 150๊ฑด)์ด ๋ํ ์๋ฒ๋ฅผ ๋ณดํธํฉ๋๋ค. ๋๋ ์คํฌ๋ํ์ ํ์ง ์์ต๋๋ค.
์ค์น ๋ฐ ์ค์
๋ฐฉ๋ฒ A: uv ์ฌ์ฉ (๊ถ์ฅ)
uv๋ MCP ์๋ฒ ์คํ์ ํ์ค ๋๊ตฌ์ ๋๋ค. ๋ณ๋ ์ค์น ๋จ๊ณ ์์ด ํด๋ผ์ด์ธํธ ์ค์ ๋ง ํ๋ฉด ๋ฉ๋๋ค.
uv ์ค์น (์ต์ด 1ํ):
| ํ๋ซํผ | ๋ฐฉ๋ฒ |
|---|---|
| macOS | brew install uv |
| Windows | powershell -c "irm https://astral.sh/uv/install.ps1 | iex" |
| Linux | curl -LsSf https://astral.sh/uv/install.sh | sh |
์ค์น ํ ์๋ MCP ํด๋ผ์ด์ธํธ ์ค์ ์ผ๋ก ๋ฐ๋ก ์งํํ์ธ์.
๋ฐฉ๋ฒ B: pip ์ฌ์ฉ
์ ํต์ ์ธ ๋ฐฉ์์ ์ ํธํ๋ ๊ฒฝ์ฐ:
pip install bht-mcp
Python 3.11 ์ด์ ํ์. Python์ด ์๋ ๊ฒฝ์ฐ: python.org/downloads (Windows: "Add Python to PATH" ์ฒดํฌ). macOS:
brew install python. Linux:sudo apt install python3 python3-pip.
MCP ํด๋ผ์ด์ธํธ ์ค์
MCP (Model Context Protocol)๋ AI ์ด์์คํดํธ๊ฐ ์ธ๋ถ ๋๊ตฌ๋ฅผ ์ฌ์ฉํ ์ ์๊ฒ ํด์ค๋๋ค. ์๋์์ ์ฌ์ฉ ์ค์ธ ํด๋ผ์ด์ธํธ๋ฅผ ์ ํํ์ธ์:
Claude Desktop
MCP ์ค์ ์ ํธ์งํฉ๋๋ค (Settings โ Developer โ MCP Servers):
uv๋ก ์ค์นํ ๊ฒฝ์ฐ:
{
"mcpServers": {
"bht": {
"command": "uvx",
"args": ["bht-mcp"]
}
}
}
pip์ผ๋ก ์ค์นํ ๊ฒฝ์ฐ:
{
"mcpServers": {
"bht": {
"command": "python",
"args": ["-m", "bht_mcp"]
}
}
}
Claude Code
claude mcp add bht -- uvx bht-mcp
pip ์ฌ์ฉ ์:
claude mcp add bht -- python -m bht_mcp
๋ก์ปฌ LLM (Open WebUI, llama.cpp, Ollama ๋ฑ)
๋ก์ปฌ LLM ํ๊ฒฝ์ด MCP๋ฅผ ์ง์ํ๋ ๊ฒฝ์ฐ, bht-mcp๋ฅผ stdio ์๋ธํ๋ก์ธ์ค๋ก ์ค์ ํ์ธ์:
uv ์ฌ์ฉ:
{
"mcpServers": {
"bht": {
"command": "uvx",
"args": ["bht-mcp"]
}
}
}
pip ์ฌ์ฉ:
{
"mcpServers": {
"bht": {
"command": "python",
"args": ["-m", "bht_mcp"]
}
}
}
์ฐธ๊ณ : MCP tool calling์ ์ฌ์ฉํ๋ ค๋ฉด ํจ์/๋๊ตฌ ํธ์ถ์ ์ง์ํ๋ ๋ชจ๋ธ์ด ํ์ํฉ๋๋ค. ๋๋ถ๋ถ์ 7B+ instruction-tuned ๋ชจ๋ธ์ด ๋์ํฉ๋๋ค (Llama 3, Mistral, Qwen ๋ฑ).
๊ธฐํ MCP ํธํ ํด๋ผ์ด์ธํธ
MCP stdio ์ ์ก์ ์ง์ํ๋ ๋ชจ๋ ํด๋ผ์ด์ธํธ์์ ์ฌ์ฉํ ์ ์์ต๋๋ค. ์๋ฒ ๋ช
๋ น์ด๋ uvx bht-mcp ๋๋ python -m bht_mcp์
๋๋ค.
์ถ๊ฐ ํ ํด๋ผ์ด์ธํธ๋ฅผ ์ฌ์์ํ๋ฉด 7๊ฐ์ BHt ๋๊ตฌ๊ฐ ํ์ฑํ๋ฉ๋๋ค.
์ฐ๊ตฌ ์๋๋ฆฌ์ค
๋๊ตฌ ์ด๋ฆ์ด๋ ํ๋ผ๋ฏธํฐ๋ฅผ ์ ํ์ ์์ต๋๋ค โ ์ํ๋ ๊ฒ์ ์์ฐ์ด๋ก ์ค๋ช ํ๋ฉด AI๊ฐ ์ ์ ํ ๋๊ตฌ๋ฅผ ํธ์ถํฉ๋๋ค.
๊ตฌ์ ๋ถ์
"์ฐฝ์ธ๊ธฐ 1:1์ ๋จ์ด๋ณ๋ก ๋ฌธ๋ฒ ๋ถ์๊ณผ ํจ๊ป ๋ณด์ฌ์ฃผ์ธ์"
AI๊ฐ ํด๋น ์ ์ ๋ชจ๋ ํ ํฐ์ ์ฐพ๊ณ , ๊ฐ๊ฐ์ ํํ๋ก ์ ์กฐํํฉ๋๋ค:
barฤ(สพ) โ Suffixkonjugation, ๋จ์ฑ ๋จ์ 3์ธ์นญ, ์นผ(Qal) ์ด๊ฐ, ์ด๊ทผ BRสพ
สพรฏlล*hฤซm โ Substantiv, ๋จ์ฑ ๋ณต์, ๊ธฐ์ดํ สพl-h
สพศงt โ ์ ์น์ฌ(Praeposition)
ha=ลกamaym โ ๊ด์ฌ ํฌํจ Substantiv, ๋จ์ฑ ๋ณต์
์ด๊ทผ ํ์
"์ํธ์์ ์ด๊ทผ สพMR (๋งํ๋ค)์ ๋ชจ๋ ํํ๋ฅผ ์ฐพ์์ฃผ์ธ์"
AI๊ฐ ํด๋น ์ฑ ์์ ์ด ์ด๊ทผ์ ๋ชจ๋ ํ ํฐ์ ๊ฒ์ํ๊ณ , ํํ๋ณ๋ก ๊ทธ๋ฃนํํ์ฌ ๊ฒฐ๊ณผ๋ฅผ ์ ์ํฉ๋๋ค.
์ฌ๋ณธ ์ด๋ณธ ํ์ธ
"์ฐฝ์ธ๊ธฐ 1์ฅ์ ํ ์คํธ ๋นํ ์ฃผ์์ด ์๋์?"
โ Gen 1,7c โ "kin": ๊ทธ๋ฆฌ์ค์ด(G)์ ์ฐจ์ด; G์ 7c ํ
์คํธ ์ถ๊ฐ
โ Gen 1,9b โ "maqลm": G๋ miqwว, MT๋ miqwฤ-m?
๋์ฌ ์ด๊ฐ ๋ถํฌ ๋ถ์
"์ ๋์์ ์ด๋ค ๋์ฌ ์ด๊ฐ์ด ๋ํ๋๋ฉฐ, ๋น๋๋ ์ด๋ค๊ฐ์?"
AI๊ฐ ์ฝํฌ๋ ์ ๋ชจ๋ ๋์ฌ๋ฅผ ๊ฒ์ํ๊ณ , ์ด๊ฐ ์ ๋ณด๋ฅผ ์กฐํํ์ฌ ๋น๋ ํต๊ณ๋ฅผ ์ ์ํฉ๋๋ค:
โ G (์นผ/Qal): 312, D (ํผ์/Piel): 45, H (ํํ/Hiphil): 67, N (๋ํ/Niphal): 28, ...
์ฑ ๊ฐ ๋ฌธ์ฅ ๊ตฌ์กฐ ๋น๊ต
"์ฐฝ์ธ๊ธฐ 1:1๊ณผ ์ด์ฌ์ผ 40:1์ ํต์ฌ ๊ตฌ์กฐ๋ฅผ ๋น๊ตํด์ฃผ์ธ์"
AI๊ฐ ๋ ๊ตฌ์ ์ ๋ฌธ์ฅ ์์ค ๋ถ์์ ๊ฐ์ ธ์ ๊ตฌ์กฐ์ ์ฐจ์ด๋ฅผ ๊ฐ์กฐํฉ๋๋ค โ ๋ฌธ์ฅ ์ ํ, ํตํฉ์(Syntagmen) ํจํด, ์ ๊ด๊ณ.
ํฌ๊ท ๋์ฌํ ํ์
"ํ ๋ผ์์ ํธํ(Hophal, HP) ์๋ํ ๋์ฌ๋ฅผ ๋ชจ๋ ์ฐพ์์ฃผ์ธ์"
AI๊ฐ ์ฐฝ์ธ๊ธฐ๋ถํฐ ์ ๋ช ๊ธฐ๊น์ง ๊ฐ ํ ๋ผ ์ฑ ์์ ํธํ ์ด๊ฐ์ ๊ฒ์ํ๊ณ ์ข ํฉ ๋ชฉ๋ก์ ์์ฑํฉ๋๋ค.
์๋์ด ๊ตฌ๊ฐ ์๋ณ
"๋ค๋์์์์ ํ๋ธ๋ฆฌ์ด๊ฐ ์๋ ์๋์ด๋ก ํ์๋ ํ ํฐ์?"
AI๊ฐ BHt์์ ์ฌ์ฉ ๊ฐ๋ฅํ ์ธ์ด ์ฝ๋๋ฅผ ๋ฐ๊ฒฌํ ํ, ๋ค๋์์์ ํ ํฐ์ ์ธ์ด๋ณ๋ก ํํฐ๋งํฉ๋๋ค โ ์๋์ด ๊ตฌ๊ฐ(2:4bโ7:28์ฅ)์ด ๋๋ฌ๋ฉ๋๋ค.
ํํ์ค ๋ ๊ณ ๋ฉ๋ ผ(Hapax legomenon) ํ์
"ํ๋ธ๋ฆฌ์ด ์ฑ๊ฒฝ์์ ๋จ ํ ๋ฒ๋ง ๋ฑ์ฅํ๋ ์ดํ์๋ฅผ ์ฐพ์์ฃผ์ธ์"
AI๊ฐ ํน์ ์ฑ ์ ์ดํ์๋ฅผ ์ํํ๋ฉฐ ์ ์ฒด ์ฝํผ์ค์์ ๊ฐ๊ฐ์ ๋น๋๋ฅผ ํ์ธํฉ๋๋ค. ์ ํํ 1ํ ์ถํํ๋ ์ดํ์๊ฐ ํํ์ค ๋ ๊ณ ๋ฉ๋ ผ์ผ๋ก ํ์๋ฉ๋๋ค.
์ฑ ๊ฐ ํํ๋ก ๋น๊ต
"์ ์ธ๊ณผ ์ฅ๊ธฐ์ ์ฐ๊ณํ(Constructus) ํจํด์ ๋น๊ตํด์ฃผ์ธ์"
AI๊ฐ ๋ ์ฑ ์์ ์ฐ๊ณํ ๋ช ์ฌ๋ฅผ ๊ฒ์ํ ํ, ๋น๋, ์ดํ์ ๋ค์์ฑ, ํต์ฌ์ ์์น๋ฅผ ๋น๊ตํ์ฌ ๋ฌธ์ฒด์ ์ฐจ์ด๋ฅผ ๋๋ฌ๋ ๋๋ค.
๋ณํ ๋ณธ๋ฌธ์ ํต์ฌ ํธ๋ฆฌ ๋น๊ต
"์ํธ 18ํธ๊ณผ ์ฌ๋ฌด์ํ 22์ฅ์ ๋จ์ด ์์ค ํต์ฌ ๊ตฌ์กฐ๋ฅผ ๋น๊ตํด์ฃผ์ธ์ (๋ณํ ๋ณธ๋ฌธ)"
AI๊ฐ ๋ ์ฑ ์ ๋์ ์ ์์ ํต์ฌ ํธ๋ฆฌ๋ฅผ ๊ฐ์ ธ์ ๋ ํ ์คํธ ์ ์น ์ฌ์ด์ ๊ตฌ์กฐ์ ์ฐจ์ด๋ฅผ ๊ฐ์กฐํฉ๋๋ค.
๋ฒค ์๋ผ ์ฌ๋ณธ ๋น๊ต
"์๋ผํฌ ์ฌ๋ณธ A์ ์ฌ๋ณธ B์ ๋ณด์กด๋ ์ฅ์ ๊ฐ๊ฐ ์ด๋ค ๊ฒ์ธ๊ฐ์?"
AI๊ฐ ์ฑ ๋ชฉ๋ก์ ์กฐํํ์ฌ ์ฌ๋ณธ ์ปค๋ฒ๋ฆฌ์ง๋ฅผ ๋ณด์ฌ์ค๋๋ค:
โ ASir: 2โ16, 23, 27์ฅ (17๊ฐ ์ฅ)
โ BSir: 10โ11, 15โ16, 20, 30โ51์ฅ (26๊ฐ ์ฅ)
โ ๊ฒน์น๋ ์ฅ: 10, 11, 15, 16 โ ์ด๋ค์ ๋น๊ต
๋๊ตฌ ์ฐธ์กฐ
| ๋๊ตฌ | ์ฉ๋ | BHt ์์ฒญ ์ |
|---|---|---|
| bht_list_books | 47๊ฐ ์ฑ ๋ชฉ๋ก (์ฝ๋์ ์ฅ ์ ํฌํจ) | 0 |
| bht_field_info | 42๊ฐ ๊ฒ์ ํ๋์ ์ ํจํ ๊ฐ ์กฐํ | 0โ1 |
| bht_search | ์์น ๋๋ ํํ๋ก ํํฐ๋ก ํ ํฐ ๊ฒ์ | 0โ1 |
| bht_token_detail | ๋จ์ผ ํ ํฐ์ ์์ ํ ํํ๋ก ๋ถ์ | 0โ1 |
| bht_syntax_tree | ๋จ์ด ์์ค ํต์ฌ ํธ๋ฆฌ (Wortfรผgungsebene) | 0โ2 |
| bht_sentence_analysis | ๋ฌธ์ฅ ์์ค ๋ถ์ (Satzfรผgungsebene) | 0โ2 |
| bht_text_annotations | ์ฅ ๋จ์ ํ ์คํธ ๋นํ ์ฃผ์ | 0โ1 |
์ผ๋ฐ์ ์ธ ์ํฌํ๋ก์ฐ: bht_list_books โ bht_search โ bht_token_detail โ bht_syntax_tree
๋ชจ๋ ์๋ต์ ์ผ์ผ ์ฌ์ฉ๋์ ๋ณด์ฌ์ฃผ๋ quota ํ๋๊ฐ ํฌํจ๋ฉ๋๋ค:
{
"data": [...],
"quota": {"daily_html_used": 12, "daily_html_limit": 150, "daily_html_remaining": 138}
}
์ฌ์ฉ ์ ํ
์ด ๋๊ตฌ๋ ๋ํ ์ฐ๊ตฌ ์๋ฒ์ ์ ๊ทผํฉ๋๋ค. ๋ด์ฅ๋ ์ ํ์ด ์๋ฒ๋ฅผ ๋ณดํธํฉ๋๋ค:
| ์ ํ | ๊ฐ | ์ด์ |
|---|---|---|
| ์์ฒญ ๊ฐ๊ฒฉ | ์ด๋น 1ํ | ์๋ฒ ์ฉ๋ ๋ณดํธ |
| ์ผ์ผ HTML ์์ฒญ | 150๊ฑด/์ผ | ํ ํฐ ์์ธ, ํต์ฌ, ๋ฌธ์ฅ, ์ฃผ์ ํ์ด์ง |
| ์ผ์ผ JSON ์์ฒญ | ๋ฌด์ ํ | ๊ฒ์ API๋ ๊ฒฝ๋ |
์ด ์ ํ์ ์ฐ๊ตฌ์๊ฐ ์ง์ ์ฌ์ดํธ๋ฅผ ํ์ํ๋ ์์ค(๋ณดํต ํ๋ฃจ 25โ100 ํ์ด์ง)๊ณผ ๋๋ฑํฉ๋๋ค. ์ผ์ผ ์ ํ์ ์์ ์ ์ด๊ธฐํ๋ฉ๋๋ค.
์ ํ์ ๋๋ฌํ๋ฉด ์บ์๋ ๋ฐ์ดํฐ๋ ์ฌ์ ํ ์์ ํ ์ ๊ทผ ๊ฐ๋ฅํฉ๋๋ค โ ์๋ก์ด HTML ๊ฐ์ ธ์ค๊ธฐ๋ง ์ฐจ๋จ๋ฉ๋๋ค.
๋ผ์ด์ ์ค
์ฝ๋: MIT
BHt ๋ฐ์ดํฐ: CC BY-SA 4.0 (LMU Munich)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file bht_mcp-0.1.3.tar.gz.
File metadata
- Download URL: bht_mcp-0.1.3.tar.gz
- Upload date:
- Size: 33.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
8f3f40fb76a9959d15b7712207076f2f8f9d35014414d9176ec9d5bf73bcfcde
|
|
| MD5 |
258f086198d4d285b5be1cccca317105
|
|
| BLAKE2b-256 |
d534753d613a0161d93b35c22a9739a7da74d25870ae830b810f6fa50cc9fa87
|
File details
Details for the file bht_mcp-0.1.3-py3-none-any.whl.
File metadata
- Download URL: bht_mcp-0.1.3-py3-none-any.whl
- Upload date:
- Size: 40.3 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
a58c617daf458b9ab9631e054259bb9ff5888cb9d92ae5a2da90098e5b0543fd
|
|
| MD5 |
b13a42ed6cf70d3e7197fe8537b781e1
|
|
| BLAKE2b-256 |
5a0505f33aa0410ed90854f0f4352e0101f84a024b65b2a488e37a922b8afc97
|