diff --git a/.env.example b/.env.example
new file mode 100644
index 0000000..117d133
--- /dev/null
+++ b/.env.example
@@ -0,0 +1,11 @@
+LLM_API_KEY=github_pat_...
+LLM_BASE_URL=https://models.github.ai/inference
+LLM_MODEL=openai/gpt-4.1-mini
+LLM_RPM=60
+
+EMBEDDER_API_KEY=github_pat_...
+EMBEDDER_BASE_URL=https://models.github.ai/inference
+EMBEDDER_MODEL=text-embedding-3-large
+EMBEDDER_DIM=3072
+
+RAGU_STORAGE=ragu_data
diff --git a/.gitattributes b/.gitattributes
new file mode 100644
index 0000000..a2da00f
--- /dev/null
+++ b/.gitattributes
@@ -0,0 +1,6 @@
+* -text
+*.woff binary
+*.woff2 binary
+*.ttf binary
+*.eot binary
+*.otf binary
diff --git a/.gitignore b/.gitignore
index c3e69b5..d7c3f2f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,3 +1,7 @@
+# $mol
+-*
+.DS_Store
+
checkpoints/
benchmark/*.json
ragu_working_dir/
@@ -175,4 +179,4 @@ cython_debug/
.ruff_cache/
# PyPI configuration file
-.pypirc
\ No newline at end of file
+.pypirc
diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 0000000..c7b74e0
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,33 @@
+# --- Frontend ---
+FROM node:20-alpine AS frontend
+
+RUN apk add --no-cache git
+
+WORKDIR /app
+RUN git clone --depth 1 https://github.com/hyoo-ru/mam.git . \
+ && npm install
+
+COPY front/ bog/RAGU/front/
+
+RUN npx mam bog/RAGU/front/app
+
+EXPOSE 9080
+
+CMD ["npm", "start"]
+
+
+# --- API ---
+FROM python:3.12-slim AS api
+
+WORKDIR /app
+
+COPY pyproject.toml ./
+COPY ragu/ ./ragu/
+RUN pip install --no-cache-dir .
+
+COPY server/ ./server/
+RUN pip install --no-cache-dir -r server/requirements.txt
+
+EXPOSE 8000
+
+CMD ["uvicorn", "server.main:app", "--host", "0.0.0.0", "--port", "8000"]
diff --git a/docker-compose.yml b/docker-compose.yml
new file mode 100644
index 0000000..3b18626
--- /dev/null
+++ b/docker-compose.yml
@@ -0,0 +1,43 @@
+services:
+ web:
+ build:
+ context: .
+ dockerfile: Dockerfile
+ target: frontend
+ ports:
+ - "9081:9080"
+ restart: unless-stopped
+ tty: true
+ stdin_open: true
+ develop:
+ watch:
+ - action: sync
+ path: ./front
+ target: /app/bog/RAGU/front
+
+ # http://localhost:9081/bog/RAGU/front/app/-/test.html
+
+ api:
+ platform: linux/arm64
+ build:
+ context: .
+ dockerfile: Dockerfile
+ target: api
+ ports:
+ - "8100:8000"
+ restart: unless-stopped
+ env_file: .env
+ environment:
+ - NUMBA_CPU_NAME=generic
+ volumes:
+ - ragu_data:/app/ragu_data
+ develop:
+ watch:
+ - action: sync+restart
+ path: ./server
+ target: /app/server
+
+ # http://localhost:8100/api/status
+
+volumes:
+ ragu_data:
diff --git a/front/app/app.meta.tree b/front/app/app.meta.tree
new file mode 100644
index 0000000..49489eb
--- /dev/null
+++ b/front/app/app.meta.tree
@@ -0,0 +1 @@
+include \/mol/offline/install
diff --git a/front/app/app.view.css.ts b/front/app/app.view.css.ts
new file mode 100644
index 0000000..712a8f9
--- /dev/null
+++ b/front/app/app.view.css.ts
@@ -0,0 +1,42 @@
+namespace $.$$ {
+
+ $mol_style_define( $bog_RAGU_front_app, {
+
+ Documents: {
+ flex: {
+ basis: '30rem',
+ grow: 1,
+ },
+ Body: {
+ flex: {
+ grow: 1,
+ },
+ },
+ },
+
+ Doc_text: {
+ flex: {
+ grow: 1,
+ },
+ minHeight: '20rem',
+ },
+
+ Doc_file: {
+ alignItems: 'center',
+ gap: '.5rem',
+ },
+
+ Index_record: {
+ alignItems: 'center',
+ gap: '.5rem',
+ },
+
+ Settings_page: {
+ flex: {
+ basis: '25rem',
+ },
+ },
+
+ } )
+
+}
diff --git a/front/app/app.view.tree b/front/app/app.view.tree
new file mode 100644
index 0000000..38f13d3
--- /dev/null
+++ b/front/app/app.view.tree
@@ -0,0 +1,94 @@
+$bog_RAGU_front_app $giper_bot
+ dialog_title @ \RAGU
+ api_url \http://localhost:8100
+ history? /
+ doc_text? \
+ doc_files? /
+ index_message? \
+ index_records? /
+ config_message? \
+ llm_api_key? \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
+ llm_base_url? \https://models.github.ai/inference
+ llm_model? \openai/gpt-4.1-mini
+ llm_rpm? \60
+ embedder_api_key? \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
+ embedder_base_url? \https://models.github.ai/inference
+ embedder_model? \text-embedding-3-large
+ embedder_dim? \3072
+ Doc_file* $mol_row
+ sub /
+ <= Doc_file_icon* $mol_icon_file
+ <= Doc_file_name* $mol_view
+ sub /
+ <= doc_file_name* \
+ <= Doc_file_remove* $mol_button_minor
+ click? <=> doc_file_remove*? null
+ sub /
+ <= Doc_file_remove_icon* $mol_icon_close
+ Index_record* $mol_row
+ sub /
+ <= Index_record_icon* $mol_icon_check
+ <= Index_record_info* $mol_view
+ sub /
+ <= index_record_text* \
+ Documents $mol_page
+ title @ \Documents
+ body /
+ <= Doc_text $mol_textarea
+ hint @ \Paste text to build knowledge graph...
+ value? <=> doc_text?
+ <= Doc_open $mol_button_open
+ title @ \Add Files
+ accept \.txt,.md,.csv,.json,.xml,.html,.docx
+ files? <=> doc_files_add? null
+ <= Doc_file_list $mol_list
+ rows <= doc_file_rows /
+ <= Index_submit $mol_button_major
+ title @ \Build Knowledge Graph
+ click? <=> index_submit? null
+ <= Index_message $mol_text
+ text <= index_message?
+ <= Index_record_list $mol_list
+ rows <= index_record_rows /
+ Settings_page $mol_page
+ title @ \Settings
+ body /
+ <= Llm_api_key $mol_form_field
+ name @ \LLM API Key
+ Content <= Llm_api_key_input $mol_string
+ hint \sk-...
+ value? <=> llm_api_key?
+ <= Llm_base_url $mol_form_field
+ name @ \LLM Base URL
+ Content <= Llm_base_url_input $mol_string
+ value? <=> llm_base_url?
+ <= Llm_model $mol_form_field
+ name @ \LLM Model
+ Content <= Llm_model_input $mol_string
+ value? <=> llm_model?
+ <= Llm_rpm $mol_form_field
+ name @ \LLM RPM
+ Content <= Llm_rpm_input $mol_string
+ value? <=> llm_rpm?
+ <= Embedder_api_key $mol_form_field
+ name @ \Embedder API Key
+ Content <= Embedder_api_key_input $mol_string
+ hint \sk-...
+ value? <=> embedder_api_key?
+ <= Embedder_base_url $mol_form_field
+ name @ \Embedder Base URL
+ Content <= Embedder_base_url_input $mol_string
+ value? <=> embedder_base_url?
+ <= Embedder_model $mol_form_field
+ name @ \Embedder Model
+ Content <= Embedder_model_input $mol_string
+ value? <=> embedder_model?
+ <= Embedder_dim $mol_form_field
+ name @ \Embedder Dim
+ Content <= Embedder_dim_input $mol_string
+ value? <=> embedder_dim?
+ <= Config_save $mol_button_major
+ title @ \Save
+ click? <=> config_save? null
+ <= Config_message $mol_text
+ text <= config_message?
diff --git a/front/app/app.view.ts b/front/app/app.view.ts
new file mode 100644
index 0000000..c413099
--- /dev/null
+++ b/front/app/app.view.ts
@@ -0,0 +1,149 @@
+namespace $.$$ {
+
+ type Request = {
+ message: string
+ files: string[]
+ }
+
+ type IndexRecord = {
+ count: number
+ names: string[]
+ }
+
+ export class $bog_RAGU_front_app extends $.$bog_RAGU_front_app {
+
+ @ $mol_mem
+ config_synced() {
+ this.push_config()
+ return true
+ }
+
+ @ $mol_mem
+ override pages() {
+ this.config_synced()
+ return [
+ this.Settings_page(),
+ this.Documents(),
+ this.Dialog(),
+ ... this.result() ? [ this.Result_page( this.version() ) ] : [],
+ ]
+ }
+
+ @ $mol_action
+ override doc_files_add( next: readonly File[] ) {
+ if( !next?.length ) return
+ this.doc_files([ ... this.doc_files(), ... next ])
+ }
+
+ override doc_file_rows() {
+ return this.doc_files().map( ( _, i ) => this.Doc_file( i ) )
+ }
+
+ override doc_file_name( index: number ) {
+ return ( this.doc_files()[ index ] as File ).name
+ }
+
+ @ $mol_action
+ override doc_file_remove( index: number ) {
+ const files = [ ... this.doc_files() ]
+ files.splice( index, 1 )
+ this.doc_files( files )
+ }
+
+ override index_record_rows() {
+ return this.index_records().map( ( _, i ) => this.Index_record( i ) )
+ }
+
+ override index_record_text( index: number ) {
+ const rec = this.index_records()[ index ] as IndexRecord
+ return `${ rec.count } doc(s): ${ rec.names.join( ', ' ) }`
+ }
+
+ @ $mol_mem
+ override communication() {
+
+ const history = this.history()
+ if( history.length % 2 === 0 ) return
+
+ const last = history[ history.length - 1 ] as Request
+
+ try {
+ const resp = $mol_fetch.json(
+ this.api_url() + '/api/query',
+ {
+ method: 'POST',
+ headers: { 'Content-Type': 'application/json' },
+ body: JSON.stringify({ query: last.message }),
+ },
+ )
+ this.history([ ... history, resp ])
+ } catch( error: any ) {
+ if( $mol_promise_like( error ) ) $mol_fail_hidden( error )
+ if( $mol_fail_log( error ) ) {
+ this.history([ ... history, { message: '\u{1F6D1}' + error.message, files: [] } ])
+ }
+ }
+
+ }
+
+ @ $mol_action
+ override index_submit() {
+ const text = this.doc_text()
+ const files = this.doc_files() as File[]
+
+ if( !text && !files.length ) return
+
+ const form = new FormData()
+ if( text ) form.append( 'text', text )
+ for( const file of files ) {
+ form.append( 'files', file )
+ }
+
+ const resp = $mol_fetch.json(
+ this.api_url() + '/api/index',
+ {
+ method: 'POST',
+ body: form,
+ },
+ ) as { status: string; documents_count: number; names: string[]; total_documents: number }
+
+ this.index_records([
+ ... this.index_records(),
+ { count: resp.documents_count, names: resp.names } as IndexRecord,
+ ])
+
+ this.index_message( `Indexed ${ resp.documents_count } doc(s). Total: ${ resp.total_documents }` )
+ this.doc_text( '' )
+ this.doc_files( [] )
+ }
+
+ @ $mol_action
+ push_config() {
+ $mol_fetch.json(
+ this.api_url() + '/api/config',
+ {
+ method: 'POST',
+ headers: { 'Content-Type': 'application/json' },
+ body: JSON.stringify({ env: {
+ LLM_API_KEY: this.llm_api_key(),
+ LLM_BASE_URL: this.llm_base_url(),
+ LLM_MODEL: this.llm_model(),
+ LLM_RPM: this.llm_rpm(),
+ EMBEDDER_API_KEY: this.embedder_api_key(),
+ EMBEDDER_BASE_URL: this.embedder_base_url(),
+ EMBEDDER_MODEL: this.embedder_model(),
+ EMBEDDER_DIM: this.embedder_dim(),
+ } }),
+ },
+ )
+ }
+
+ @ $mol_action
+ override config_save() {
+ this.push_config()
+ this.config_message( 'Saved' )
+ }
+
+ }
+
+}
diff --git a/front/app/index.html b/front/app/index.html
new file mode 100644
index 0000000..26dd154
--- /dev/null
+++ b/front/app/index.html
@@ -0,0 +1,14 @@
+
+
+
+
+ RAGU
+
+
+
+
+
+
+
+
+
diff --git a/server/main.py b/server/main.py
new file mode 100644
index 0000000..5a00fa3
--- /dev/null
+++ b/server/main.py
@@ -0,0 +1,221 @@
+import io
+import os
+from contextlib import asynccontextmanager
+
+from fastapi import FastAPI, UploadFile, File, Form
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
+from pydantic import BaseModel
+
+from ragu import (
+ KnowledgeGraph,
+ SimpleChunker,
+ BuilderArguments,
+ LocalSearchEngine,
+ GlobalSearchEngine,
+ NaiveSearchEngine,
+ Settings,
+)
+from ragu.llm import OpenAIClient
+from ragu.embedder import OpenAIEmbedder
+from ragu.triplet import ArtifactsExtractorLLM
+
+
+ENV_KEYS = [
+ "LLM_API_KEY", "LLM_BASE_URL", "LLM_MODEL", "LLM_RPM",
+ "EMBEDDER_API_KEY", "EMBEDDER_BASE_URL", "EMBEDDER_MODEL", "EMBEDDER_DIM",
+ "RAGU_STORAGE",
+]
+
+
+class State:
+ kg: KnowledgeGraph | None = None
+ client: OpenAIClient | None = None
+ embedder: OpenAIEmbedder | None = None
+ indexed: bool = False
+ all_documents: list[str] = []
+ all_names: list[str] = []
+
+
+state = State()
+
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+ reinit_clients()
+ yield
+ if state.client:
+ await state.client.async_close()
+ if state.embedder:
+ await state.embedder.aclose()
+
+
+app = FastAPI(title="RAGU API", lifespan=lifespan)
+
+app.add_middleware(
+ CORSMiddleware,
+ allow_origins=["*"],
+ allow_methods=["*"],
+ allow_headers=["*"],
+)
+
+
+class QueryRequest(BaseModel):
+ query: str
+ engine: str = "local"
+ top_k: int = 20
+
+
+class BotResponse(BaseModel):
+ message: str
+ files: list[str] = []
+ document: str | None = None
+ confidence: float = 0.0
+ digest: str = ""
+ title: str = ""
+
+
+class ConfigUpdate(BaseModel):
+ env: dict[str, str]
+
+
+def reinit_clients():
+ state.client = OpenAIClient(
+ model_name=os.getenv("LLM_MODEL", ""),
+ base_url=os.getenv("LLM_BASE_URL", ""),
+ api_token=os.getenv("LLM_API_KEY", ""),
+ max_requests_per_minute=int(os.getenv("LLM_RPM", "60")),
+ )
+ state.embedder = OpenAIEmbedder(
+ model_name=os.getenv("EMBEDDER_MODEL", ""),
+ base_url=os.getenv("EMBEDDER_BASE_URL", os.getenv("LLM_BASE_URL", "")),
+ api_token=os.getenv("EMBEDDER_API_KEY", os.getenv("LLM_API_KEY", "")),
+ dim=int(os.getenv("EMBEDDER_DIM", "3072")),
+ )
+
+
+@app.get("/api/status")
+async def get_status():
+ return {"indexed": state.indexed}
+
+
+@app.get("/api/config")
+async def get_config():
+ return {k: os.getenv(k, "") for k in ENV_KEYS}
+
+
+@app.post("/api/config")
+async def set_config(req: ConfigUpdate):
+ for k, v in req.env.items():
+ if k in ENV_KEYS:
+ os.environ[k] = v
+ reinit_clients()
+ return {k: os.getenv(k, "") for k in ENV_KEYS}
+
+
+async def _read_file(upload: UploadFile) -> str:
+ content = await upload.read()
+ ext = (upload.filename or "").rsplit(".", 1)[-1].lower()
+
+ if ext == "docx":
+ from docx import Document as DocxDocument
+ doc = DocxDocument(io.BytesIO(content))
+ return "\n".join(p.text for p in doc.paragraphs if p.text.strip())
+
+ return content.decode("utf-8")
+
+
+@app.post("/api/index")
+async def index_documents(
+ files: list[UploadFile] = File(default=[]),
+ text: str = Form(default=""),
+ language: str = Form(default="russian"),
+):
+ documents: list[str] = []
+ names: list[str] = []
+
+ if text.strip():
+ documents.append(text.strip())
+ names.append("(text)")
+
+ for f in files:
+ doc_text = await _read_file(f)
+ if doc_text.strip():
+ documents.append(doc_text)
+ names.append(f.filename or "unknown")
+
+ if not documents:
+ return {"status": "empty", "documents_count": 0, "names": []}
+
+ state.all_documents.extend(documents)
+ state.all_names.extend(names)
+
+ try:
+ Settings.storage_folder = os.getenv("RAGU_STORAGE", "ragu_data")
+ Settings.language = language
+
+ chunker = SimpleChunker(max_chunk_size=1000)
+ extractor = ArtifactsExtractorLLM(client=state.client, do_validation=False)
+
+ kg = KnowledgeGraph(
+ client=state.client,
+ embedder=state.embedder,
+ chunker=chunker,
+ artifact_extractor=extractor,
+ builder_settings=BuilderArguments(
+ use_llm_summarization=True,
+ vectorize_chunks=True,
+ ),
+ )
+
+ await kg.build_from_docs(state.all_documents)
+
+ state.kg = kg
+ state.indexed = True
+
+ except Exception as e:
+ # Roll back documents added in this request
+ state.all_documents = state.all_documents[:-len(documents)]
+ state.all_names = state.all_names[:-len(names)]
+ return JSONResponse(
+ status_code=500,
+ content={"error": str(e), "names": names, "documents_count": 0},
+ )
+
+ return {
+ "status": "indexed",
+ "documents_count": len(documents),
+ "names": names,
+ "total_documents": len(state.all_documents),
+ }
+
+
+@app.post("/api/query", response_model=BotResponse)
+async def query_graph(req: QueryRequest):
+ if not state.kg:
+ return BotResponse(
+ message="Knowledge graph not built yet. Please index documents first.",
+ )
+
+ try:
+ if req.engine == "local":
+ engine = LocalSearchEngine(state.client, state.kg, state.embedder)
+ answer = await engine.a_query(req.query, top_k=req.top_k)
+ elif req.engine == "global":
+ engine = GlobalSearchEngine(state.client, state.kg)
+ answer = await engine.a_query(req.query)
+ elif req.engine == "naive":
+ engine = NaiveSearchEngine(state.client, state.kg, state.embedder)
+ answer = await engine.a_query(req.query, top_k=req.top_k)
+ else:
+ return BotResponse(message=f"Unknown engine: {req.engine}")
+ except Exception as e:
+ return BotResponse(message=f"Error: {e}")
+
+ return BotResponse(
+ message=answer,
+ confidence=0.8,
+ title=req.query[:50],
+ )
+
+
diff --git a/server/requirements.txt b/server/requirements.txt
new file mode 100644
index 0000000..f6b3ebf
--- /dev/null
+++ b/server/requirements.txt
@@ -0,0 +1,4 @@
+fastapi
+uvicorn
+python-docx
+python-multipart