Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

7.2 KB
Raw Blame
#
#  Copyright 2026 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.
#
from typing import Any

from .base import Base
from .document import Document


class DataSet(Base):
    class ParserConfig(Base):
        def __init__(self, rag, res_dict):
            super().__init__(rag, res_dict)

    def __init__(self, rag, res_dict):
        self.id = ""
        self.name = ""
        self.avatar = ""
        self.tenant_id = None
        self.description = ""
        self.embedding_model = ""
        self.permission = "me"
        self.document_count = 0
        self.chunk_count = 0
        self.parser_id = ""
        self.pipeline_id = None
        self.language = None
        self.parser_config = None
        self.pagerank = 0
        for k in list(res_dict.keys()):
            if k not in self.__dict__:
                res_dict.pop(k)
        super().__init__(rag, res_dict)

    @property
    def parse_type(self) -> int:
        """1 when parsed by a builtin parser, 2 when parsed by an ingestion pipeline."""
        return 2 if self.pipeline_id else 1

    def update(self, update_message: dict):
        if "chunk_method" in update_message:
            raise ValueError("chunk_method is not supported by RAGFlow 1.0; use parse_type with parser_id or pipeline_id")
        res = self.put(f"/datasets/{self.id}", update_message)
        res = res.json()
        if res.get("code") != 0:
            raise Exception(res["message"])

        data = res.get("data", {})
        # BuiltIn datasets omit pipeline_id from the response; clear the previous
        # mode before merging so a Pipeline -> BuiltIn update cannot leave stale state.
        self.pipeline_id = data.get("pipeline_id") or None
        self._update_from_dict(self.rag, data)
        return self

    def upload_documents(self, document_list: list[dict]):
        url = f"/datasets/{self.id}/documents"
        files = [("file", (ele["display_name"], ele["blob"])) for ele in document_list]
        res = self.post(path=url, json=None, files=files)
        res = res.json()
        if res.get("code") == 0:
            doc_list = []
            for doc in res["data"]:
                document = Document(self.rag, doc)
                doc_list.append(document)
            return doc_list
        raise Exception(res.get("message"))

    def list_documents(
        self,
        id: str | None = None,
        ids: list[str] | None = None,
        name: str | None = None,
        keywords: str | None = None,
        page: int = 1,
        page_size: int = 30,
        orderby: str = "create_time",
        desc: bool = True,
        create_time_from: int = 0,
        create_time_to: int = 0,
    ):
        # Validate that id and ids are not used together
        if id and ids:
            raise ValueError("Cannot use both 'id' and 'ids' parameters at the same time.")

        params = {
            "id": id,
            "name": name,
            "keywords": keywords,
            "page": page,
            "page_size": page_size,
            "orderby": orderby,
            "desc": desc,
            "create_time_from": create_time_from,
            "create_time_to": create_time_to,
        }
        # Handle ids parameter - requests expands list values into multiple query params
        if ids:
            params["ids"] = ids
        res = self.get(f"/datasets/{self.id}/documents", params=params)
        res = res.json()
        documents = []
        if res.get("code") == 0:
            for document in res["data"].get("docs"):
                documents.append(Document(self.rag, document))
            return documents
        raise Exception(res["message"])

    def delete_documents(self, ids: list[str] | None = None, delete_all: bool = False):
        res = self.rm(f"/datasets/{self.id}/documents", {"ids": ids, "delete_all": delete_all})
        res = res.json()
        if res.get("code") != 0:
            raise Exception(res["message"])

    def _get_documents_status(self, document_ids):
        import time

        terminal_states = {"DONE", "FAIL", "CANCEL"}
        ingestion_terminal_states = {"COMPLETED": "DONE", "FAILED": "FAIL", "STOPPED": "CANCEL"}
        interval_sec = 1
        pending = set(document_ids)
        finished = []
        while pending:
            for doc_id in list(pending):
                docs = self.list_documents(id=doc_id)
                if not docs:
                    raise RuntimeError(f"Document {doc_id} not found while waiting for parsing to finish")
                doc = docs[0]
                status = ingestion_terminal_states.get(doc.ingestion_status)
                if status:
                    finished.append((doc_id, status, doc.chunk_count, doc.token_count))
                    pending.discard(doc_id)
                elif isinstance(doc.run, str) and doc.run.upper() in terminal_states:
                    finished.append((doc_id, doc.run, doc.chunk_count, doc.token_count))
                    pending.discard(doc_id)
                elif float(doc.progress or 0.0) >= 1.0:
                    finished.append((doc_id, "DONE", doc.chunk_count, doc.token_count))
                    pending.discard(doc_id)
            if pending:
                time.sleep(interval_sec)
        return finished

    def async_parse_documents(self, document_ids):
        res = self.post(f"/datasets/{self.id}/chunks", {"document_ids": document_ids})
        res = res.json()
        if res.get("code") != 0:
            raise Exception(res.get("message"))

    def parse_documents(self, document_ids):
        try:
            self.async_parse_documents(document_ids)
            return self._get_documents_status(document_ids)
        except KeyboardInterrupt:
            self.async_cancel_parse_documents(document_ids)

        return self._get_documents_status(document_ids)

    def async_cancel_parse_documents(self, document_ids):
        res = self.rm(f"/datasets/{self.id}/chunks", {"document_ids": document_ids})
        res = res.json()
        if res.get("code") != 0:
            raise Exception(res.get("message"))

    def get_auto_metadata(self) -> dict[str, Any]:
        """
        Retrieve auto-metadata configuration for a dataset via SDK.
        """
        res = self.get(f"/datasets/{self.id}/metadata/config")
        res = res.json()
        if res.get("code") == 0:
            return res["data"]
        raise Exception(res["message"])

    def update_auto_metadata(self, **config: Any) -> dict[str, Any]:
        """
        Update auto-metadata configuration for a dataset via SDK.
        """
        res = self.put(f"/datasets/{self.id}/metadata/config", config)
        res = res.json()
        if res.get("code") == 0:
            return res["data"]
        raise Exception(res["message"])