app.fojin由 fojin.app 发布v0.4.0更新于 Oct 24, 2026
Buddhist canon tools: search, passages, cross-canon parallels, dictionaries — all URN-cited.
托管在 FoJin 佛学语料库之上的只读 MCP 服务器，让助手检索经文、读取段落、对照跨藏平行文本并核验引文。
Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。
把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。
阅读 https://sourceweft.com/mcp/app.fojin%2Ffojin-mcp/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。把它添加到你客户端的 mcpServers 配置中。
{
"mcpServers": {
"fojin-buddhist-canon-tools": {
"type": "http",
"url": "https://mcp.fojin.ai/mcp"
}
}
}AI Q&A grounded in the world's largest open aggregation of Buddhist texts: 600+ sources, 30+ languages, trilingual cross-canon retrieval — every answer linked back to its source passage.
Ask a question in plain language and FoJin's assistant "XiaoJin" answers from the canon itself — Retrieval-Augmented Generation over 670K+ embedded passages, optional cross-encoder reranking and root-sutra recall, clickable 【《sutra》juan N】 citations that open the exact source text, anti-hallucination guards, and a citation drawer with side-by-side 汉 / 巴利 / 藏文 cross-canon parallels. You can also ask in the voice of 15 historical Buddhist masters, each scoped to their own tradition's scriptures.
What makes those answers trustworthy is the corpus underneath. FoJin aggregates 612 data sources into one searchable platform — 10,500+ texts with 19,000+ volumes of full content in Classical Chinese, Pali, Tibetan and Sanskrit, the first LLM-driven trilingual cross-canon parallel reading platform (CBETA × SuttaCentral × 84000) with LLM-verified chunk-level alignment, a 110K+ entity knowledge graph on a Deck.GL geo map, and 39 dictionaries with 747K entries. Every feature exists to make the answers more grounded — and to let you go deeper once you have one.
FoJin is built to be open, cross-canon, verifiable Buddhist knowledge infrastructure — not just a site to read, but a corpus other tools can call. Every passage carries a stable, resolvable cross-canon URN (fojin:cbeta/T0001.1), and the fojin-mcp server lets AI assistants (Claude, ChatGPT, any MCP client) answer from FoJin's cited passages directly — hosted at mcp.fojin.ai, anonymous, no key required. Any model can also check its own work through the open /api/verify/quote endpoint: hand it a sentence and a citation, and it answers is this quote real and is the citation right as two separate questions. See fojin.app/agents.
Live Demo · API Docs · 中文文档 · Discussions · Discord · Report Bug
[CI] [Security Scan] [fojin-mcp on PyPI] [License] [GitHub stars]
Buddhist texts are scattered across hundreds of databases worldwide — CBETA, SuttaCentral, BDRC, SAT, 84000, GRETIL, and many more. Each has its own interface, language, and data format. When you have a question — "what does the Heart Sutra mean by 'form is emptiness'?", "how do the Pali and Chinese versions of this passage differ?" — you spend more time hunting for the right passage than understanding it.
FoJin answers the question for you. Ask in plain language; XiaoJin retrieves the relevant passages from 612 sources and answers with clickable citations you can verify. Everything else FoJin does — full-text reading, cross-canon alignment, the knowledge graph, 39 dictionaries — exists to make those answers more grounded, and to let you go deeper once you have one:
Then visit: http://localhost:3000
API docs at http://localhost:8000/docs
After first startup, the platform has the database schema and source metadata but no text content. To import texts from public data sources:
Each importer downloads data directly from the original source (CBETA, SuttaCentral, etc.) — no data is bundled in this repository.
FoJin's verified, cross-canon corpus is callable directly from AI assistants (Claude Desktop, ChatGPT, or any MCP client) via the published fojin-mcp server — so the assistant answers Buddhist questions from FoJin's cited passages instead of hallucinating.
Point any MCP client at the hosted endpoint — nothing to install, no key to get:
Or run it locally over stdio against your own instance:
It exposes eight read-only tools over the public API, each returning passages with a stable, resolvable cross-canon URN (fojin:cbeta/T0001.1):
Claude Desktop config and ChatGPT setup are in mcp-server/README.md. The server is a thin, read-only client: it holds no credentials, bundles no corpus, and only calls FoJin's public endpoints — its default target is https://fojin.app/api, overridable via FOJIN_API_BASE_URL to point at a self-hosted instance.
Search across Buddhist canons by title, translator, catalog number, or full-text keyword. Powered by Elasticsearch with ICU tokenizer for multi-language support.
[Search results for Avatamsaka Sutra]
Read 8,900+ Buddhist texts with 19,000+ volumes of full content online. CBETA-style typography with intelligent verse/prose detection, paragraph reflow, and adjustable font size. Navigate by volume, scroll through content, and jump between related texts.
Compare translations side by side — Classical Chinese, Sanskrit, Pali, Tibetan, English, Japanese, Korean, Gandhari, and 21 more languages.
32 authoritative dictionaries with 748,000+ entries across Chinese, Pali, Sanskrit, Tibetan, and English:
Chinese Buddhist Dictionaries (14)
Pali Dictionaries (5)
Sanskrit Dictionaries (4)
Tibetan Dictionaries (2)
Multilingual Reference (4)
Specialized (3)
110,000+ entities (monasteries, persons, texts, schools, concepts) and 28,000+ relationships — including 22,000+ teacher-student lineage chains from the DILA Authority Database — visualized as an interactive force-directed graph. Click any node to explore connections.
The first LLM-driven cross-canon parallel reading system for Buddhist texts. No other platform provides this: CBETA (汉文), SuttaCentral (Pali), and 84000 (Tibetan) each operate in their own language silo. FoJin bridges them via LLM-verified chunk-level alignment.
Current coverage (4,000+ chunk-level alignments across 11 pair definitions):
Confidence distribution: all pairs ≥ 0.75. Hand-verified precision on the original MVP sample: 100%. Mahāyāna 汉藏 batches: Lotus Sutra (2026-06-08) at $1.70 / 259 pairs (8.6% accept rate); 8,000-verse Prajñāpāramitā (2026-06-09) at $3.64 / 127 pairs (3.4% accept rate, lower because the sūtra's repetitive paratactic style yields fewer 1:1 chunk-level correspondences).
How to use:
In AI Q&A — When XiaoJin cites an aligned sutra, the citation drawer shows tabs [ 汉文 ] [ 巴利 (5) ] [ 藏文 (3) ]. Click a tab to see the corresponding passage in another canon, rendered with proper Devanagari / Tibetan fonts.
In the reader — Click the 🌐 「多语对读」 (Multilingual Parallel) button in the toolbar. Default tab 「按经对读」 shows sutta-level parallels from SuttaCentral's authoritative Akanuma-style table (3,293 pairs covering all 4 Āgamas ↔ Nikāyas), with Pāli original + Sujato English previews and "read full text" links. Alternate tab 「按段对读」 retains the experimental embedding+LLM chunk-level alignment (pipeline-generated, known to have noise). The panel sits to the left of the AI reading panel; both can be open simultaneously and independently resized.
Pipeline (backend/scripts/build_alignments.py):
{is_parallel, confidence, reason} JSON per candidateconfidence ≥ 0.75 persisted to alignment_pairs with unique (text_a, text_b) chunk tuple for idempotent re-runs$50 cost ceiling guard (actual MVP spend: ~$0.15)RAG layer automatically includes parallel_chunks in the LLM context when a retrieved chunk has alignments, so answers can naturally reference "the Pali version says…" without hallucinating.
Growing the alignment set — the flywheel: beyond the batch pipeline above, an alignment flywheel (backend/app/services/alignment_flywheel.py) mines new candidate parallels by expanding outward from already-verified pairs — neighbouring chunks tend to align too, so this is both fast and precise where blind nearest-neighbour search drowns in same-language matches. Candidates are staged for human review (admin UI at /admin/alignment/review) and only promoted into the ground-truth alignment_pairs once accepted (method='flywheel-verified'). Nothing is auto-promoted — human review is the precision gate — and each verified alignment makes the next round of mining better. A margin-based candidate router in build_alignments.py routes recalled candidates into auto-accept / LLM-verify / auto-reject bands by ratio-margin, cutting LLM cost while keeping auto-accept off by default so the precision guarantee holds.
From chunks to sentences — the depth play: verified chunk pairs seed sentence-level alignment (backend/scripts/refine_sentence_alignments.py + services/sentence_align.py): a pure-Python bertalign-style dynamic program subdivides each aligned paragraph into 1-1 / 1-2 / 2-1 sentence pairs over BGE-M3 cosine similarity, anchored to stable character offsets and stored in sentence_alignments. This backs precise 逐句对读 and cross-lingual sentence lookup. Two consumption surfaces already ship: cross-lingual sentence search (GET /api/search/parallel-sentences — query in Chinese, get aligned Sanskrit/Tibetan sentences over the ~896K-pair MITRA store) and a versioned, license-stamped public dataset export (GET /exports/alignments.jsonl) that turns the alignment corpus into a citable research artifact. Alignment quality is guarded by a dedicated eval harness (gold set + precision/recall/calibration metrics + regression gate) alongside the RAG eval, and MITRA cross-canon parallels feed the RAG context gated by a proxy quality score (mitra_e_score).
This is FoJin's core. Ask questions in natural language; XiaoJin answers from canonical Buddhist texts using RAG (Retrieval-Augmented Generation) over 670K+ embedding vectors with an HNSW index for fast semantic search. Answers stay grounded because retrieval combines vector similarity, keyword reranking, and root-sutra recall (the sutra you asked about is always pulled into context), and every quoted passage is checked against the retrieved sources before it can become a clickable citation. Features include:
[AI Q&A answering about Xuanzang's disciples]
Trust is the point. Every answer passes three deterministic guards before it reaches you:
verified / citation_corrected / quote_relaxed / no_sources so you can see how grounded it is.Measured on the eval harness at temperature 0, the raw model is verbatim-faithful only a small fraction of the time — but after the guards ~98% of citing answers are trustworthy: FoJin either points you to a real source or honestly hedges, and never fabricates scripture. The metric (served_trustworthy_rate) is tracked in backend/eval/faithfulness.py as a regression gate.
Read that denominator precisely, because we'd rather you knew than were impressed: it counts answers that cite at all. An answer that reaches you with no 【…】 citation cites nothing, so it verifies nothing, and it is not in that 98% — the guards can only vouch for claims that name a source. Quote fidelity is reported separately (verbatim_quote_rate), on purpose: a run must not be able to improve its score by quoting less.
For multi-step questions a single search can't answer — "how is śūnyatā treated across Prajñāpāramitā, Madhyamaka and Yogācāra, with cited cross-canon parallels?" — the research assistant plans the question into steps, retrieves across the corpus, dictionaries and knowledge graph, then synthesises a grounded answer. The synthesis runs through the same citation guards as chat, so the agent can plan freely but cannot cite what it didn't retrieve. Available at /research (sign-in required); API at POST /api/research/query.
Select a specific Buddhist master to receive answers in their teaching style, grounded in their tradition's core scriptures. 15 historical masters available:
Each master has a 100-150 line enriched system prompt with lineage, core doctrines, speaking style, teaching methods, key allusions, and terminology table. When a master is selected, RAG retrieval is scoped to their core scriptures (e.g., selecting Zhiyi only searches 《摩诃止观》《法华玄义》 etc.), providing more precise citations.
Powered by Master-skill — the open-source Buddhist master AI persona framework.
Visualize 50,000+ geo-enabled Buddhist entities on an interactive world map — monasteries, historical places, persons, and schools. Built with Deck.GL + MapLibre.
Track real-time updates from 612 data sources — new texts added, translation releases, manuscript scans, and schema changes. Includes academic content aggregation and platform-wide activity summary.
When reading any text, the sidebar automatically finds semantically similar passages from other texts using pgvector cosine similarity. Discover cross-textual parallels, related commentaries, and thematic connections across the entire canon.
Visualize Buddhist textual history with interactive D3 charts — dynasty distribution, translation trends, language breakdown, category treemap, and top translators. Toggle between scholarly and popular presentation modes.
Save texts to personal collections, bookmark specific passages, and add annotations for study and research.
Export citations in BibTeX, RIS, and APA formats for academic papers and reference managers.
Available in 9 languages: Simplified Chinese, Traditional Chinese, English, Japanese, Korean, Thai, Vietnamese, Sinhala, and Burmese.
[612 data sources from 30 countries]
FoJin aggregates data from major Buddhist digital projects worldwide. Sources are categorized by research field (Han, Theravada, Tibetan, Sanskrit, Dunhuang, Art, Dictionary, Digital Humanities) and filterable by region, language, and type:
app, frontend: nginx)127.0.0.1 onlyPOST /api/auth/logout-all revokes every outstanding token. Production requires a strong secretVITE_UMAMI_URL + VITE_UMAMI_WEBSITE_ID; if either is unset, the frontend ships without telemetry.LLM_API_URL / EMBEDDING_API_URL at a local OpenAI-compatible server (vLLM, Ollama, LM Studio) for a fully offline setup, or have each user provide their own key via the in-app BYOK flow.127.0.0.1 by default (set via FRONTEND_BIND), so out of the box it's reachable only from the host — put an authenticated reverse proxy or CDN in front of it as the public entry. If you need direct LAN/remote access without a proxy, set FRONTEND_BIND=0.0.0.0; note that Docker publishes ports through its own iptables rules that bypass host firewalls like ufw, so anyone who can reach the host on that port hits your instance.See SECURITY.md for the full picture.
Contributions are welcome! Whether it's adding a new data source, improving search, fixing bugs, or translating the UI — we'd love your help.
git checkout -b feat/amazing-feature)git commit -m 'Add amazing feature')git push origin feat/amazing-feature)See CONTRIBUTING.md for detailed guidelines.
/docs, ReDoc at /redoc)/texts/{id} pages — proper <title> / <description> per sutra for search engines (SPA route crawlability)served_trustworthy_rate ~98% of citing answers at temp 0 as an eval regression gatefojin:cbeta/T0001.1) interoperable with CBETA / SuttaCentral / 84000 / GRETIL / VRIfojin-mcp — 8 read-only, URN-addressable tools; callable from Claude Desktop / ChatGPThttps://mcp.fojin.ai/mcp over streamable HTTP; anonymous, rate-limited, no key to obtainGET /api/verify/quote + verify_quote tool) — works on any quotation, not only FoJin's own answers: ES shortlists, Postgres decides, exact / near-miss / absent with a character-level diff; never silently accepts a match in a different juanGET /api/commentary/passage + commentaries tool) — historical commentaries keyed to a passage, anchored to juan and line; first corpus is the Diamond Sutra (52 commentators, 5,683 aligned segments)/agents) + llms.txt — one page telling an AI what FoJin exposes and how to call it/read-aloud) — sentence-synced audio with line highlighting and lock-screen playback (Heart Sutra first)/research) — plan → retrieve (corpus + dictionary + KG) → grounded synthesis behind the same citation guards/admin/alignment/review) into ground-truth alignmentssentence_alignments); pure-Python, no new depsmitra_e_score, NULL-permissive)GET /api/search/parallel-sentences)GET /exports/alignments.jsonl streaming endpointGET /alignment/sentences/..., ships dark behind enable_sentence_parallels); frontend pending prod sentence dataApache License 2.0 — applies to FoJin source code only. Third-party data sources retain their own licenses (CC BY-NC-SA, CC0, CC BY-NC-ND, etc.). See NOTICE for details.
FoJin is built on the generous work of the global Buddhist digital humanities community. Special thanks to:
If FoJin is useful for your research, please consider giving it a star!
Discussions · Issues · Contributing · [email protected]
Made with care for the Buddhist studies community.
来源：README.md，提交 b38d9cc
Semantic search across fojin's Buddhist corpus (10K+ texts, 30+ langs). Returns the most relevant passages, each with a `urn`, title, snippet and similarity score. `lang` optionally filters by language code (lzh=Classical Chinese, pi=Pali, sa=Sanskrit, bo=Tibetan, en=English).
langlimitquery必填Read the full content of one fascicle (卷) of a text, with its `urn`. Use the `text_id`/`juan_num` from a search_corpus hit. Returns the actual canonical text — quote from this, not from memory.
text_id必填juan_num必填Cross-canon parallel passages aligned to a fascicle. fojin's alignment moat: given a Chinese fascicle, returns the aligned Pali/Tibetan/Sanskrit parallels so you can compare how a passage is rendered across traditions. A fascicle returns parallels for the *whole* juan — often hundreds. To narrow them to one passage, match the passage against `source_chunks[].text` and keep the parallels whose `aligns_source_chunk` equals that chunk_index. `urn` is filled for parallels that are fojin works; rows with `source="mitra-parallel"` are inline Skt/Tib sentences with no fojin work to cite, so they carry `original_preview` instead and stay `urn: null`.
text_id必填juan_num必填Look up a Buddhist term across fojin's dictionaries (748K+ entries).
term必填limitSearch fojin's knowledge graph for entities — people, places, works, doctrinal terms — matching `query`.
limitquery必填Resolve a fojin URN (e.g. fojin:cbeta/T0001.1) to a reader URL and confirm it exists in the corpus. Use to verify or dereference a citation.
urn必填Verify that a Buddhist-canon quote exists VERBATIM in the corpus. Call this before presenting any quoted scripture to a reader: LLMs routinely invent plausible-looking quotes. Returns `verbatim` (bool), where it was found (`matches`, each with a resolvable `urn`), or the closest near-miss window when it wasn't. `cite` optionally narrows the search — a CBETA id ("T0374") or fojin URN ("fojin:cbeta/T0374.13") — and `cite_matched` reports honestly whether the quote is where you claimed (a hit in a different fascicle does NOT confirm your citation). Quote must be ≥4 CJK chars after normalisation; Classical Chinese only. Short quotes are answered but say less: a four-character phrase recurs across the canon, so read `cite_matched` rather than `verbatim`, and check `matches_capped` before treating the list as complete. Each match carries an absolute `reader_url` — cite that, not a reconstructed third-party link.
citejuanquote必填What the historical commentators said about a passage of Buddhist scripture. Give a line of a sūtra and get back the classical commentaries that gloss it — each with the commentator's own words, the work it comes from, a quality tier, and a resolvable citation. This is the layer no other tool exposes for the Chinese canon: CBETA marks up almost none of it, and the alignment behind this was derived text by text. Two things to read carefully rather than skim. `total` is how many commentators the passage actually has — `truncated` says the list you got is a sample. And coverage is partial: roughly half of what a commentator wrote is not aligned, so an absent commentator means "not in the data", never "he said nothing". When nothing matches, `available_sutras` tells you which sūtras have data at all — do not read a miss as silence from the tradition.
limitquote必填已满足适用检查项的 100%
该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。
git clone https://github.com/xr843/fojin.gitcd fojincp .env.example .env # edit POSTGRES_PASSWORD before startingdocker compose up -d # database migrations run automatically# Import CBETA catalog (auto-scans local xml-p5 directory or fetches from remote)docker exec fojin-backend python scripts/import_catalog.py
# Import CBETA full text content (requires xml-p5 repository)docker exec fojin-backend python scripts/import_content.py --all --xml-dir /data/xml-p5
# Generate embeddings for AI Q&A (supports incremental processing)docker exec fojin-backend python -m scripts.archive.misc.generate_embeddings --source cbeta
# Import SuttaCentral Early Buddhist Textsdocker exec fojin-backend python scripts/archive/imports/import_suttacentral.py
# See all available importers (one-off importers live under archive/)ls backend/scripts/archive/imports/claude mcp add --transport http fojin https://mcp.fojin.ai/mcpuvx fojin-mcp # zero-install run# or: pip install fojin-mcp && fojin-mcp +-------------+ | Cloudflare | (CDN, SSL, DDoS protection) +------+------+ | +------+------+ | Nginx | (gzip, security headers, static cache) +------+------+ | +-----------+-----------+ | | +-----+------+ +-----+------+ | React 18 | | FastAPI | | Vite + D3 | | async SSE | +-------------+ +------+------+ | +--------+---------+---------+ | | | | +-----+--+ +--+----+ +--+---+ +---+--------+ | PG 15 | | ES 8 | |Redis | | LLM APIs | | pgvector | | ICU | |cache | | (multi- | | HNSW idx | | | | | | provider) | +---------+ +-------+ +------+ +------------+# Backendcd backendpython -m venv .venv && source .venv/bin/activatepip install -r requirements-dev.txtalembic upgrade headuvicorn app.main:app --reload
# Frontendcd frontendnpm installnpm run dev
# Testscd backend && pytest tests/ -q