RAGflow/api/db/services/document_service.py

#
#  Copyright 2024 The InfiniFlow Authors. All Rights Reserved.
#
#  Licensed under the Apache License, Version 2.0 (the "License");
#  you may not use this file except in compliance with the License.
#  You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
#  Unless required by applicable law or agreed to in writing, software
#  distributed under the License is distributed on an "AS IS" BASIS,
#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#  See the License for the specific language governing permissions and
#  limitations under the License.
#
import logging
import xxhash
import json
import random
import re
from concurrent.futures import ThreadPoolExecutor
from copy import deepcopy
from datetime import datetime
from io import BytesIO
import trio

from peewee import fn

from api.db.db_utils import bulk_insert_into_db
from api import settings
from api.utils import current_timestamp, get_format_time, get_uuid
from rag.settings import SVR_QUEUE_NAME
from rag.utils.storage_factory import STORAGE_IMPL
from rag.nlp import search, rag_tokenizer

from api.db import FileType, TaskStatus, ParserType, LLMType
from api.db.db_models import DB, Knowledgebase, Tenant, Task, UserTenant
from api.db.db_models import Document
from api.db.services.common_service import CommonService
from api.db.services.knowledgebase_service import KnowledgebaseService
from api.db import StatusEnum
from rag.utils.redis_conn import REDIS_CONN


class DocumentService(CommonService):
    model = Document

    @classmethod
    @DB.connection_context()
    def get_list(cls, kb_id, page_number, items_per_page, orderby, desc, keywords, id, name):
        docs = cls.model.select().where(cls.model.kb_id == kb_id)
        if id:
            docs = docs.where(cls.model.id == id)
        if name:
            docs = docs.where(cls.model.name == name)
        if keywords:
            docs = docs.where(fn.LOWER(cls.model.name).contains(keywords.lower()))
        if desc:
            docs = docs.order_by(cls.model.getter_by(orderby).desc())
        else:
            docs = docs.order_by(cls.model.getter_by(orderby).asc())

        count = docs.count()
        docs = docs.paginate(page_number, items_per_page)
        return list(docs.dicts()), count

    @classmethod
    @DB.connection_context()
    def get_by_kb_id(cls, kb_id, page_number, items_per_page, orderby, desc, keywords):
        if keywords:
            docs = cls.model.select().where((cls.model.kb_id == kb_id), (fn.LOWER(cls.model.name).contains(keywords.lower())))
        else:
            docs = cls.model.select().where(cls.model.kb_id == kb_id)
        count = docs.count()
        if desc:
            docs = docs.order_by(cls.model.getter_by(orderby).desc())
        else:
            docs = docs.order_by(cls.model.getter_by(orderby).asc())

        docs = docs.paginate(page_number, items_per_page)

        return list(docs.dicts()), count

    @classmethod
    @DB.connection_context()
    def insert(cls, doc):
        if not cls.save(**doc):
            raise RuntimeError("Database error (Document)!")
        e, kb = KnowledgebaseService.get_by_id(doc["kb_id"])
        if not KnowledgebaseService.update_by_id(kb.id, {"doc_num": kb.doc_num + 1}):
            raise RuntimeError("Database error (Knowledgebase)!")
        return Document(**doc)

    @classmethod
    @DB.connection_context()
    def remove_document(cls, doc, tenant_id):
        cls.clear_chunk_num(doc.id)
        try:
            settings.docStoreConn.delete({"doc_id": doc.id}, search.index_name(tenant_id), doc.kb_id)
            settings.docStoreConn.update(
                {"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "community_report"], "source_id": doc.id},
                {"remove": {"source_id": doc.id}},
                search.index_name(tenant_id),
                doc.kb_id,
            )
            settings.docStoreConn.update({"kb_id": doc.kb_id, "knowledge_graph_kwd": ["graph"]}, {"removed_kwd": "Y"}, search.index_name(tenant_id), doc.kb_id)
            settings.docStoreConn.delete(
                {"kb_id": doc.kb_id, "knowledge_graph_kwd": ["entity", "relation", "graph", "community_report"], "must_not": {"exists": "source_id"}}, search.index_name(tenant_id), doc.kb_id
            )
        except Exception:
            pass
        return cls.delete_by_id(doc.id)

    @classmethod
    @DB.connection_context()
    def get_newly_uploaded(cls):
        fields = [
            cls.model.id,
            cls.model.kb_id,
            cls.model.parser_id,
            cls.model.parser_config,
            cls.model.name,
            cls.model.type,
            cls.model.location,
            cls.model.size,
            Knowledgebase.tenant_id,
            Tenant.embd_id,
            Tenant.img2txt_id,
            Tenant.asr_id,
            cls.model.update_time,
        ]
        docs = (
            cls.model.select(*fields)
            .join(Knowledgebase, on=(cls.model.kb_id == Knowledgebase.id))
            .join(Tenant, on=(Knowledgebase.tenant_id == Tenant.id))
            .where(
                cls.model.status == StatusEnum.VALID.value,
                ~(cls.model.type == FileType.VIRTUAL.value),
                cls.model.progress == 0,
                cls.model.update_time >= current_timestamp() - 1000 * 600,
                cls.model.run == TaskStatus.RUNNING.value,
            )
            .order_by(cls.model.update_time.asc())
        )
        return list(docs.dicts())

    @classmethod
    @DB.connection_context()
    def get_unfinished_docs(cls):
        fields = [cls.model.id, cls.model.process_begin_at, cls.model.parser_config, cls.model.progress_msg, cls.model.run, cls.model.parser_id]
        docs = cls.model.select(*fields).where(cls.model.status == StatusEnum.VALID.value, ~(cls.model.type == FileType.VIRTUAL.value), cls.model.progress < 1, cls.model.progress > 0)
        return list(docs.dicts())

    @classmethod
    @DB.connection_context()
    def increment_chunk_num(cls, doc_id, kb_id, token_num, chunk_num, duation):
        num = (
            cls.model.update(token_num=cls.model.token_num + token_num, chunk_num=cls.model.chunk_num + chunk_num, process_duation=cls.model.process_duation + duation)
            .where(cls.model.id == doc_id)
            .execute()
        )
        if num == 0:
            raise LookupError("Document not found which is supposed to be there")
        num = Knowledgebase.update(token_num=Knowledgebase.token_num + token_num, chunk_num=Knowledgebase.chunk_num + chunk_num).where(Knowledgebase.id == kb_id).execute()
        return num

    @classmethod
    @DB.connection_context()
    def decrement_chunk_num(cls, doc_id, kb_id, token_num, chunk_num, duation):
        num = (
            cls.model.update(token_num=cls.model.token_num - token_num, chunk_num=cls.model.chunk_num - chunk_num, process_duation=cls.model.process_duation + duation)
            .where(cls.model.id == doc_id)
            .execute()
        )
        if num == 0:
            raise LookupError("Document not found which is supposed to be there")
        num = Knowledgebase.update(token_num=Knowledgebase.token_num - token_num, chunk_num=Knowledgebase.chunk_num - chunk_num).where(Knowledgebase.id == kb_id).execute()
        return num

    @classmethod
    @DB.connection_context()
    def clear_chunk_num(cls, doc_id):
        doc = cls.model.get_by_id(doc_id)
        assert doc, "Can't fine document in database."

        num = (
            Knowledgebase.update(token_num=Knowledgebase.token_num - doc.token_num, chunk_num=Knowledgebase.chunk_num - doc.chunk_num, doc_num=Knowledgebase.doc_num - 1)
            .where(Knowledgebase.id == doc.kb_id)
            .execute()
        )
        return num

    @classmethod
    @DB.connection_context()
    def get_tenant_id(cls, doc_id):
        docs = cls.model.select(Knowledgebase.tenant_id).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(cls.model.id == doc_id, Knowledgebase.status == StatusEnum.VALID.value)
        docs = docs.dicts()
        if not docs:
            return
        return docs[0]["tenant_id"]

    @classmethod
    @DB.connection_context()
    def get_knowledgebase_id(cls, doc_id):
        docs = cls.model.select(cls.model.kb_id).where(cls.model.id == doc_id)
        docs = docs.dicts()
        if not docs:
            return
        return docs[0]["kb_id"]

    @classmethod
    @DB.connection_context()
    def get_tenant_id_by_name(cls, name):
        docs = cls.model.select(Knowledgebase.tenant_id).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(cls.model.name == name, Knowledgebase.status == StatusEnum.VALID.value)
        docs = docs.dicts()
        if not docs:
            return
        return docs[0]["tenant_id"]

    @classmethod
    @DB.connection_context()
    def accessible(cls, doc_id, user_id):
        docs = (
            cls.model.select(cls.model.id)
            .join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id))
            .join(UserTenant, on=(UserTenant.tenant_id == Knowledgebase.tenant_id))
            .where(cls.model.id == doc_id, UserTenant.user_id == user_id)
            .paginate(0, 1)
        )
        docs = docs.dicts()
        if not docs:
            return False
        return True

    @classmethod
    @DB.connection_context()
    def accessible4deletion(cls, doc_id, user_id):
        docs = cls.model.select(cls.model.id).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(cls.model.id == doc_id, Knowledgebase.created_by == user_id).paginate(0, 1)
        docs = docs.dicts()
        if not docs:
            return False
        return True

    @classmethod
    @DB.connection_context()
    def get_embd_id(cls, doc_id):
        docs = cls.model.select(Knowledgebase.embd_id).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(cls.model.id == doc_id, Knowledgebase.status == StatusEnum.VALID.value)
        docs = docs.dicts()
        if not docs:
            return
        return docs[0]["embd_id"]

    @classmethod
    @DB.connection_context()
    def get_chunking_config(cls, doc_id):
        configs = (
            cls.model.select(
                cls.model.id,
                cls.model.kb_id,
                cls.model.parser_id,
                cls.model.parser_config,
                Knowledgebase.language,
                Knowledgebase.embd_id,
                Tenant.id.alias("tenant_id"),
                Tenant.img2txt_id,
                Tenant.asr_id,
                Tenant.llm_id,
            )
            .join(Knowledgebase, on=(cls.model.kb_id == Knowledgebase.id))
            .join(Tenant, on=(Knowledgebase.tenant_id == Tenant.id))
            .where(cls.model.id == doc_id)
        )
        configs = configs.dicts()
        if not configs:
            return None
        return configs[0]

    @classmethod
    @DB.connection_context()
    def get_doc_id_by_doc_name(cls, doc_name):
        fields = [cls.model.id]
        doc_id = cls.model.select(*fields).where(cls.model.name == doc_name)
        doc_id = doc_id.dicts()
        if not doc_id:
            return
        return doc_id[0]["id"]

    @classmethod
    @DB.connection_context()
    def get_thumbnails(cls, docids):
        fields = [cls.model.id, cls.model.kb_id, cls.model.thumbnail]
        return list(cls.model.select(*fields).where(cls.model.id.in_(docids)).dicts())

    @classmethod
    @DB.connection_context()
    def update_parser_config(cls, id, config):
        e, d = cls.get_by_id(id)
        if not e:
            raise LookupError(f"Document({id}) not found.")

        def dfs_update(old, new):
            for k, v in new.items():
                if k not in old:
                    old[k] = v
                    continue
                if isinstance(v, dict):
                    assert isinstance(old[k], dict)
                    dfs_update(old[k], v)
                else:
                    old[k] = v

        dfs_update(d.parser_config, config)
        if not config.get("raptor") and d.parser_config.get("raptor"):
            del d.parser_config["raptor"]
        cls.update_by_id(id, {"parser_config": d.parser_config})

    @classmethod
    @DB.connection_context()
    def get_doc_count(cls, tenant_id):
        docs = cls.model.select(cls.model.id).join(Knowledgebase, on=(Knowledgebase.id == cls.model.kb_id)).where(Knowledgebase.tenant_id == tenant_id)
        return len(docs)

    @classmethod
    @DB.connection_context()
    def begin2parse(cls, docid):
        cls.update_by_id(docid, {"progress": random.random() * 1 / 100.0, "progress_msg": "Task is queued...", "process_begin_at": get_format_time()})

    @classmethod
    @DB.connection_context()
    def update_meta_fields(cls, doc_id, meta_fields):
        return cls.update_by_id(doc_id, {"meta_fields": meta_fields})

    @classmethod
    @DB.connection_context()
    def update_progress(cls):
        docs = cls.get_unfinished_docs()
        for d in docs:
            try:
                tsks = Task.query(doc_id=d["id"], order_by=Task.create_time)
                if not tsks:
                    continue
                msg = []
                prg = 0
                finished = True
                bad = 0
                has_raptor = False
                has_graphrag = False
                e, doc = DocumentService.get_by_id(d["id"])
                status = doc.run  # TaskStatus.RUNNING.value
                for t in tsks:
                    if 0 <= t.progress < 1:
                        finished = False
                    if t.progress == -1:
                        bad += 1
                    prg += t.progress if t.progress >= 0 else 0
                    msg.append(t.progress_msg)
                    if t.task_type == "raptor":
                        has_raptor = True
                    elif t.task_type == "graphrag":
                        has_graphrag = True
                prg /= len(tsks)
                if finished and bad:
                    prg = -1
                    status = TaskStatus.FAIL.value
                elif finished:
                    if d["parser_config"].get("raptor", {}).get("use_raptor") and not has_raptor:
                        queue_raptor_o_graphrag_tasks(d, "raptor")
                        prg = 0.98 * len(tsks) / (len(tsks) + 1)
                    elif d["parser_config"].get("graphrag", {}).get("use_graphrag") and not has_graphrag:
                        queue_raptor_o_graphrag_tasks(d, "graphrag")
                        prg = 0.98 * len(tsks) / (len(tsks) + 1)
                    else:
                        status = TaskStatus.DONE.value

                msg = "\n".join(sorted(msg))
                info = {"process_duation": datetime.timestamp(datetime.now()) - d["process_begin_at"].timestamp(), "run": status}
                if prg != 0:
                    info["progress"] = prg
                if msg:
                    info["progress_msg"] = msg
                cls.update_by_id(d["id"], info)
            except Exception as e:
                if str(e).find("'0'") < 0:
                    logging.exception("fetch task exception")

    @classmethod
    @DB.connection_context()
    def get_kb_doc_count(cls, kb_id):
        return len(cls.model.select(cls.model.id).where(cls.model.kb_id == kb_id).dicts())

    @classmethod
    @DB.connection_context()
    def do_cancel(cls, doc_id):
        try:
            _, doc = DocumentService.get_by_id(doc_id)
            return doc.run == TaskStatus.CANCEL.value or doc.progress < 0
        except Exception:
            pass
        return False


def queue_raptor_o_graphrag_tasks(doc, ty):
    chunking_config = DocumentService.get_chunking_config(doc["id"])
    hasher = xxhash.xxh64()
    for field in sorted(chunking_config.keys()):
        hasher.update(str(chunking_config[field]).encode("utf-8"))

    def new_task():
        nonlocal doc
        return {"id": get_uuid(), "doc_id": doc["id"], "from_page": 100000000, "to_page": 100000000, "task_type": ty, "progress_msg": datetime.now().strftime("%H:%M:%S") + " created task " + ty}

    task = new_task()
    for field in ["doc_id", "from_page", "to_page"]:
        hasher.update(str(task.get(field, "")).encode("utf-8"))
    hasher.update(ty.encode("utf-8"))
    task["digest"] = hasher.hexdigest()
    bulk_insert_into_db(Task, [task], True)
    assert REDIS_CONN.queue_product(SVR_QUEUE_NAME, message=task), "Can't access Redis. Please check the Redis' status."


def doc_upload_and_parse(conversation_id, file_objs, user_id):
    """
    上传并解析文档，将内容存入知识库

    参数:
        conversation_id: 会话ID
        file_objs: 文件对象列表
        user_id: 用户ID

    返回:
        处理成功的文档ID列表

    处理流程:
        1. 验证会话和知识库
        2. 初始化嵌入模型
        3. 上传文件到存储
        4. 多线程解析文件内容
        5. 生成内容嵌入向量
        6. 存入文档存储系统
    """
    from rag.app import presentation, picture, naive, audio, email
    from api.db.services.dialog_service import DialogService
    from api.db.services.file_service import FileService
    from api.db.services.llm_service import LLMBundle
    from api.db.services.user_service import TenantService
    from api.db.services.api_service import API4ConversationService
    from api.db.services.conversation_service import ConversationService

    e, conv = ConversationService.get_by_id(conversation_id)
    if not e:
        e, conv = API4ConversationService.get_by_id(conversation_id)
    assert e, "Conversation not found!"

    e, dia = DialogService.get_by_id(conv.dialog_id)
    if not dia.kb_ids:
        raise LookupError("No knowledge base associated with this conversation. Please add a knowledge base before uploading documents")
    kb_id = dia.kb_ids[0]
    e, kb = KnowledgebaseService.get_by_id(kb_id)
    if not e:
        raise LookupError("Can't find this knowledgebase!")

    embd_mdl = LLMBundle(kb.tenant_id, LLMType.EMBEDDING, llm_name=kb.embd_id, lang=kb.language)

    err, files = FileService.upload_document(kb, file_objs, user_id)
    assert not err, "\n".join(err)

    def dummy(prog=None, msg=""):
        pass

    FACTORY = {ParserType.PRESENTATION.value: presentation, ParserType.PICTURE.value: picture, ParserType.AUDIO.value: audio, ParserType.EMAIL.value: email}
    parser_config = {"chunk_token_num": 4096, "delimiter": "\n!?;。；！？", "layout_recognize": "Plain Text"}
    # 使用线程池执行解析任务
    exe = ThreadPoolExecutor(max_workers=12)
    threads = []
    doc_nm = {}
    for d, blob in files:
        doc_nm[d["id"]] = d["name"]
    for d, blob in files:
        kwargs = {"callback": dummy, "parser_config": parser_config, "from_page": 0, "to_page": 100000, "tenant_id": kb.tenant_id, "lang": kb.language}
        threads.append(exe.submit(FACTORY.get(d["parser_id"], naive).chunk, d["name"], blob, **kwargs))

    for (docinfo, _), th in zip(files, threads):
        docs = []
        doc = {"doc_id": docinfo["id"], "kb_id": [kb.id]}
        for ck in th.result():
            d = deepcopy(doc)
            d.update(ck)
            d["id"] = xxhash.xxh64((ck["content_with_weight"] + str(d["doc_id"])).encode("utf-8")).hexdigest()
            d["create_time"] = str(datetime.now()).replace("T", " ")[:19]
            d["create_timestamp_flt"] = datetime.now().timestamp()
            if not d.get("image"):
                docs.append(d)
                continue

            output_buffer = BytesIO()
            if isinstance(d["image"], bytes):
                output_buffer = BytesIO(d["image"])
            else:
                d["image"].save(output_buffer, format="JPEG")

            STORAGE_IMPL.put(kb.id, d["id"], output_buffer.getvalue())
            d["img_id"] = "{}-{}".format(kb.id, d["id"])
            d.pop("image", None)
            docs.append(d)

    parser_ids = {d["id"]: d["parser_id"] for d, _ in files}
    docids = [d["id"] for d, _ in files]
    chunk_counts = {id: 0 for id in docids}
    token_counts = {id: 0 for id in docids}
    es_bulk_size = 64

    def embedding(doc_id, cnts, batch_size=16):
        nonlocal embd_mdl, chunk_counts, token_counts
        vects = []
        for i in range(0, len(cnts), batch_size):
            vts, c = embd_mdl.encode(cnts[i : i + batch_size])
            vects.extend(vts.tolist())
            chunk_counts[doc_id] += len(cnts[i : i + batch_size])
            token_counts[doc_id] += c
        return vects

    idxnm = search.index_name(kb.tenant_id)
    try_create_idx = True

    _, tenant = TenantService.get_by_id(kb.tenant_id)
    llm_bdl = LLMBundle(kb.tenant_id, LLMType.CHAT, tenant.llm_id)
    for doc_id in docids:
        cks = [c for c in docs if c["doc_id"] == doc_id]

        if parser_ids[doc_id] != ParserType.PICTURE.value:
            from graphrag.general.mind_map_extractor import MindMapExtractor

            mindmap = MindMapExtractor(llm_bdl)
            try:
                mind_map = trio.run(mindmap, [c["content_with_weight"] for c in docs if c["doc_id"] == doc_id])
                mind_map = json.dumps(mind_map.output, ensure_ascii=False, indent=2)
                if len(mind_map) < 32:
                    raise Exception("Few content: " + mind_map)
                cks.append(
                    {
                        "id": get_uuid(),
                        "doc_id": doc_id,
                        "kb_id": [kb.id],
                        "docnm_kwd": doc_nm[doc_id],
                        "title_tks": rag_tokenizer.tokenize(re.sub(r"\.[a-zA-Z]+$", "", doc_nm[doc_id])),
                        "content_ltks": rag_tokenizer.tokenize("summary summarize 总结 概况 file 文件 概括"),
                        "content_with_weight": mind_map,
                        "knowledge_graph_kwd": "mind_map",
                    }
                )
            except Exception as e:
                logging.exception("Mind map generation error")

        vects = embedding(doc_id, [c["content_with_weight"] for c in cks])
        assert len(cks) == len(vects)
        for i, d in enumerate(cks):
            v = vects[i]
            d["q_%d_vec" % len(v)] = v
        for b in range(0, len(cks), es_bulk_size):
            if try_create_idx:
                if not settings.docStoreConn.indexExist(idxnm, kb_id):
                    settings.docStoreConn.createIdx(idxnm, kb_id, len(vects[0]))
                try_create_idx = False
            settings.docStoreConn.insert(cks[b : b + es_bulk_size], idxnm, kb_id)

        DocumentService.increment_chunk_num(doc_id, kb.id, token_counts[doc_id], chunk_counts[doc_id], 0)

    return [d["id"] for d, _ in files]