Bid document product parameter information matching system and method based on large model in service industry

Abstract

The invention relates to the technical field of AI large models, in particular to a bidding document product parameter information matching system and method based on a large model in the service industry, and the system comprises a human-computer interface module, a product parameter information uploading module, a bidding document uploading module, a bidding document product parameter information matching module, a security management module and a database module. The method has the beneficial effects that an efficient, accurate and flexible bidding file product parameter information matching system based on the large model is constructed by integrating a plurality of modules of product parameter information uploading, bidding file uploading, product parameter and bidding file matching and the like, and remarkable convenience and benefits are brought to purchasing or bidding work of users.

Description

Technical Field

[0001] The present invention relates to the technical field of AI large models, and specifically provides a system and method for matching product parameter information of tender documents based on large models in the service industry. Background Art

[0002] In the current digital age, document processing and information retrieval have become an indispensable part of all industries. Especially in the field of bidding, the accurate extraction and analysis of document content are particularly important. Traditional document processing methods often rely on manual reading, marking, and extraction, which are not only inefficient but also easily affected by human factors, resulting in information omission or errors. Especially when dealing with a large number of documents, such as tender documents in various formats like docx and pdf, how to quickly and accurately locate and extract key information, such as control points, has become an urgent problem to be solved.

[0003] In the prior art, although there are some document parsing tools and search engines, most of them focus on full-text retrieval or simple content extraction, and have limited ability to accurately locate and fragmentarily extract specific control points, making it difficult to meet the efficient and accurate processing requirements in complex business scenarios. In addition, these tools often lack the unified processing ability across different formats, forcing users to frequently switch tools or methods when dealing with files of different formats, increasing the operation complexity and time cost. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for matching product parameter information of tender documents based on large models in the service industry. By introducing the advanced deep analysis ability of large models and the retrieval ability of the bge model, it aims to achieve efficient and accurate matching and retrieval between user-defined product parameter information and tender document content, so as to solve the problems proposed in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: A system for matching product parameter information of tender documents based on large models in the service industry, the system includes:

[0006] A human-machine interface module, which enables business users and administrators to access the system through a browser and enter system pages with different functions and layouts according to the user type;

[0007] A product parameter information upload module, which allows users to import a self-defined product parameter information table in a specified area, and parse, verify, and store the information in a database;

[0008] A tender document upload module, which supports users to upload tender documents in PDF or Word format, perform format analysis, preprocessing, segmentation, and metadata update, and create a vector database;

[0009] The product parameter information matching module of the tender document uses large model technology to enable users to select columns in the product parameter information table as the matching basis, perform efficient retrieval with the tender document vector library, introduce the large model to evaluate similarity, store the matching information in the database, and finally display the matching results in the original format.

[0010] Preferably, the human-machine interface module includes:

[0011] The business user interface provides functions for uploading tender documents, importing product parameter information tables, and viewing matching results;

[0012] The administrator interface is used for system maintenance, including model library update, database management, and user account management and permission allocation.

[0013] Preferably, the tender document upload module further includes:

[0014] The file format analysis unit is used to determine whether the uploaded file is in Word or PDF format, and perform text or picture type recognition on PDF files;

[0015] The PDF processing unit directly converts text PDFs, and first converts picture PDFs to PNG pictures and then to Word documents;

[0016] The file splitting unit uses the Loader in the langchain library to load and split Word-version tender documents, supporting splitting in two ways: by character count and by paragraph;

[0017] The vector database creation unit creates a FAISS vector database for tender documents using the selected embedding large model.

[0018] Preferably, the product parameter information matching module of the tender document includes:

[0019] The matching basis selection unit allows users to freely select any column or multiple columns in the product parameter information table as the matching basis;

[0020] The efficient retrieval unit uses the bge-es-reranker retrieval technology to perform cyclic retrieval of the selected product parameter information with the tender document vector library;

[0021] The similarity evaluation unit introduces a large model, merges the product parameter information with the retrieved tender document fragments through refined guiding text, and performs in-depth analysis and similarity judgment;

[0022] The result storage and display unit stores the matching information of each row in the database in dictionary form, and after all processes are completed, displays the matching results in the original format of the product parameter information table.

[0023] Preferably, the system further includes:

[0024] A security management module that implements a strict permission management mechanism to ensure that each user account can only view the materials uploaded by itself, protecting the security and privacy of the bidding documents;

[0025] A database module that supports multiple types such as mongodb and redis, and is used to store the parsed product parameter information, vector database, and matching results, with the design considering data scalability and query efficiency.

[0026] A matching method for a matching system of product parameter information of bidding documents based on large models in the service industry, the method includes the following steps:

[0027] Receive the product parameter information table and the bidding document uploaded by the user, where the product parameter information table contains the key information for matching, and the bidding document is in Word or PDF format;

[0028] Use the parsing mechanism to parse the product parameter information table, extract the column names and their corresponding values, and verify the integrity of the data and the correctness of the format, and then store the extracted data in the database;

[0029] Preprocess the bidding document, including format analysis, PDF file conversion, file splitting, metadata update, and creating a FAISS vector database using the selected large model;

[0030] Provide a user interface that allows the user to select any one or more columns in the product parameter information table as the matching basis, and select the bidding documents to be analyzed;

[0031] Start the matching mechanism, use the bge-es-reranker retrieval technology of the large model to efficiently retrieve the selected product parameter information from the bidding document vector library, and introduce the large model to deeply evaluate the retrieval results, give the similarity score, and finally store the matching information in the database and display the matching results in the original format.

[0032] Preferably, the steps of preprocessing the bidding document specifically include:

[0033] Analyze the format of the bidding document. If it is in PDF format, further determine whether it is a text PDF or a picture PDF;

[0034] For text PDF, directly perform conversion processing; for picture PDF, first convert it to a PNG picture, and then convert the picture to a Word document;

[0035] Use the Loader in the langchain library to load and split the Word version of the bidding document, supporting splitting in two ways: by character count and by paragraph.

[0036] Add a document number to the metadata of each tender document;

[0037] Create a FAISS vector database for tender documents using the selected large model for efficient retrieval later.

[0038] Preferably, the steps of starting the matching mechanism specifically include:

[0039] Extract the corresponding content as the matching basis according to the columns in the product parameter information table selected by the user;

[0040] Use the bge-es-reranker retrieval technology to circularly retrieve the extracted product parameter information from the tender document vector library, and return the tender document fragments with the highest similarity to the product parameters;

[0041] Merge the product parameter information and the retrieved tender document fragments with the refined guiding text, and introduce the large model for in-depth analysis and similarity judgment;

[0042] According to the evaluation results of the large model, give the similarity scores of each line of product parameter information and the corresponding tender document fragments.

[0043] Preferably, the method further includes:

[0044] Store the matching information of each line in the database in the form of a dictionary in an orderly manner to ensure the integrity and security of the data;

[0045] After all the retrieval and judgment processes are completed, extract the execution results of this time from the database and display them in the original format of the product parameter information table, providing users with a clear and intuitive experience of viewing the matching results.

[0046] Preferably, the method further includes:

[0047] Implement a strict permission management mechanism to ensure that each user account can only view the materials uploaded by themselves, protecting the security and privacy of tender documents;

[0048] Support the invocation of multiple leading large models, such as Inspur Cloud Sea Ruo large model, Alibaba Qianwen, and GPT, to ensure the accuracy and efficiency of analysis results;

[0049] The database supports mongodb and redis types, which are used to store the parsed product parameter information, vector database, and matching results. The design considers the scalability and query efficiency of the data.

[0050] Compared with the prior art, the beneficial effects of the present invention are:

[0051] The proposed system and method for matching product parameter information in tender documents in the service industry based on large models can, through an automated parsing and matching mechanism, complete the comparison of a large amount of product parameter information with tender documents within a short time, significantly improving the processing efficiency. At the same time, by using advanced retrieval techniques and large model evaluation, the accuracy of the matching results is ensured, reducing human errors and omissions.

[0052] Users can freely select the product parameter columns and tender documents to be matched in the system according to actual needs, realizing a highly customized matching process. This design gives users more operating space, enabling them to flexibly adjust the matching strategy according to their own needs.

[0053] The system uses a database to store key information, ensuring data scalability and query efficiency. The matching results are presented to users in an intuitive and clear manner, facilitating users to quickly understand the degree of association between product parameters and tender documents and providing an intuitive basis for decision-making. Brief Description of the Drawings

[0054] Figure 1 is the development flowchart of the present invention;

[0055] Figure 2 is the system block diagram of the present invention. Detailed Embodiments

[0056] In order to clearly and completely describe the purpose, technical solution of the present invention, and make the advantages more clearly understood, the following further details the embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0057] Example 1, please refer to Figures 1 to 2 , the present invention provides a technical solution: a system for matching product parameter information in tender documents in the service industry based on large models, which makes full use of the cutting-edge AI technologies of Inspur Cloud Sea Ruoyun large model and Inspur Cloud Sea Ruoyun bge model, uses Python language as the basic programming language, and combines the open-source library Streamlit to write an interactive web application.

[0058] The development process of this system is as shown in Figure 1 and is described in detail below:

[0059] The human-machine interface is an important part of this system, and the interaction between users and the system is realized through the main interface module. There are mainly two types of users of this system: business users and administrators. They access the system interface through a browser and enter system pages with different functions and layouts according to different user types.

[0060] For business users, they mainly perform the following operations through the system interface: uploading tender documents in the form of word or pdf; importing product parameter information tables in the form of excel; and viewing the excel table after the tender documents are processed by the large model and matched with the product parameter information. In terms of security, the system implements a strict permission management mechanism to ensure that each user account can only view the materials uploaded by itself, thus effectively guaranteeing the security and privacy of tender documents.

[0061] Administrators maintain the model library and database through the system maintenance page. This includes updating or adding new large models to meet changing needs, managing the database to ensure data accuracy and integrity, and managing and maintaining user accounts to ensure reasonable allocation and management of user permissions and protect the security of tender documents.

[0062] The model library is one of the core components of this system. Among them, the Hairo BGE model is responsible for accurately extracting corresponding segments from tender documents according to control points. Subsequently, the Hairo large model will deeply analyze these control points and the extracted segments, evaluate their similarity, and give specific similarity values. It is worth mentioning that the system also supports calling a variety of leading large models such as the Inspur Cloud Hairo large model, Alibaba Qianwen, and GPT to ensure the accuracy and efficiency of analysis results.

[0063] In terms of the database, this system supports multiple types such as mongodb and redis, which are used to store the dictionaries extracted by the BGE model, such as information on each row of control points in the excel table and the corresponding tender document segments returned by the model.

[0064] Specifically, for business users, the human-machine interface of the resume optimization system is subdivided into the following function modules: product parameter information upload module, tender document upload module, and tender document product parameter information matching module. For the detailed core function module diagram, please refer to Figure 2 , and the specific implementation method is as follows: [[ID=1']]

[0065] Product parameter information upload module

[0066] Users can import a custom product parameter information table in the specified area. This product parameter information table mainly contains the key information that users hope to search for and match among numerous tender documents. This information may involve key data such as the technical specifications, performance indicators, and material composition of the product, which is crucial for evaluating the compliance and superiority of tender documents. Once the user completes the file upload, the system will immediately start the parsing mechanism. This mechanism can accurately identify the structure of the Excel table and extract the column names and their corresponding values. During the parsing process, the system will verify the integrity of the data and the correctness of the format to ensure that the extracted information is accurate. After parsing is completed, the system will store the extracted data in the database. The database design fully considers data scalability and query efficiency, enabling subsequent product parameter matching operations to be executed quickly and accurately. By storing this key information in the database, the system can quickly compare it with the corresponding content in the tender documents, thus providing users with efficient matching results.

[0067] Tender Document Upload Module

[0068] Users upload the corresponding tender documents in this specified area, supporting the import of PDF or Word format files. When the user sends a request through the API, the system immediately receives and starts a series of preprocessing procedures. The specific execution process of these preprocessing procedures is as follows:

[0069] 1. Delete temporary tender documents: Delete the old tender documents in the TMP_DIR directory.

[0070] 2. Save the uploaded tender documents: Save the resume uploaded by the user through the Streamlit file uploader to TMP_DIR.

[0071] 3. Analyze the file format. Next, the system will intelligently analyze the format of the uploaded file. If the file is a Word document (DOCX format), the system will directly proceed to the next step. If the file is in PDF format, a more complex processing flow is required:

[0072] (1) Execute the is_text_pdf method to determine whether the PDF file is a text PDF or an image PDF. Open the PDF file using fitz and traverse each page to obtain the text. Determine the PDF type based on the total length of the text (if the text length is less than a certain threshold, it is considered an image PDF). Return a boolean value indicating whether the PDF is a text PDF.

[0073] (2) If it is a text PDF, call the pdf2docx method. Execute pdf2docx, inputting the path of the PDF file (pdf_path).

[0074] Generate the file name of the Word document (docx_file). Create an instance of the converter (for handling PDF to Word conversion). Call the convert method of the converter to perform the conversion. Finally, close the converter.

[0075] (3) If it is an image PDF, first call the pdf2png method to convert the PDF to images, and then call the img2docx method to convert the images to a Word document. Execute the pdf2png method to convert each page of the PDF file to a PNG image and save it in the specified directory. Input the directory (pdf_directory) and file name (file_name) where the PDF file is located. Generate the path (imagePath) where the images will be saved. Check and create the directory for saving the images. Open the PDF file using fitz and get the total number of pages. Traverse each page, use the get_pixmap method to convert the page to an image, and save it in PNG format. Return the path where the images are saved. Then execute the img2docx method to convert the images (usually PNG images converted from the PDF) to a Word document. Input the directory (img_path) and file name (file_name) where the images are located. Read all the image files in the directory. Use the table_engine of paddleocr to process the images and extract the text. Add the extracted text to a new Word document. Finally, save the Word document.

[0076] 4. Split the file: Use the Loader in the langchain library to load and split the word version of the tender document. This system supports splitting in two ways: by character count and by paragraph, which greatly improves the retrieval accuracy.

[0077] 5. Update metadata: Add a document number to the metadata of each tender document.

[0078] 6. Create a vector database: Use the selected embedding model bge to create a FAISS vector database for the tender document.

[0079] Tender Document Product Parameter Information Matching Module

[0080] The product parameter information matching module of the tender document is the core component in the tender document processing system, which realizes the efficient matching of product parameters and the content of tender documents. After the user completes the upload of relevant data through the product parameter information upload module and the tender document upload module, this module provides powerful matching and retrieval functions. Specifically, the user can freely select any one column or multiple columns in the product parameter information table within the specified area of this module as the matching basis, and select the tender documents that need to be analyzed. This design gives the user extremely high flexibility and autonomy, and can customize the matching process according to actual needs. After selecting the matching parameters and tender documents, the system will start the matching mechanism. Using Hairo's advanced bge-es-reranker retrieval technology, the system will efficiently retrieve all the content of the selected columns in the product parameter information table in a cyclic manner with the corresponding tender document vector library. During this process, the system will sequentially return the tender document fragments with the highest similarity to the product parameters. To further improve the accuracy and depth of the matching, Hairo's large model is innovatively introduced for evaluation. Specifically, the system will merge the product parameter information, such as control points, with the retrieved tender document fragments and the specially designed refined guiding text. This guiding text plays a crucial role in the system. It can not only guide the model to deeply analyze and understand the selected column information in the product parameter information table and the retrieved tender document fragments, but also guide the model to accurately judge the similarity relationship between the two. Under the guidance of the prompt sequence, the model will carefully compare and evaluate each row of product parameter information and the corresponding tender document fragments, and give the corresponding similarity score. This scoring mechanism provides an intuitive display of the matching results for the user, facilitating the user to quickly understand the degree of association between the product parameters and the tender documents. To ensure the reliability and orderliness of the matching results, the system will store the matching information of each row in an orderly manner in the database in the form of a dictionary. This storage mechanism not only ensures the integrity and security of the data, but also provides convenience for subsequent data analysis and processing.

[0081] After all the retrieval and judgment processes are completed, the system will extract the execution results of this time from the database and display them in the original format of the product parameter information table. This design aims to provide the user with a clear and intuitive experience of viewing the matching results, facilitating the user to perform subsequent operations and make decisions based on the matching results.

[0082] This embodiment provides a large model-based tender document product parameter information matching system for the service industry, including a product parameter information upload module, a tender document upload module, and a tender document product parameter information matching module; relying on AI technology, this system can quickly and accurately achieve the efficient and precise matching and retrieval between user-defined product parameter information and the content of tender documents, providing strong support for the user's procurement or bidding decisions.

[0083] Embodiment 2. Based on Embodiment 1, a matching method for product parameter information of a tender document product based on a large model in the service industry is proposed. The method includes the following steps:

[0084] Receive the product parameter information table and tender document uploaded by the user, where the product parameter information table contains key information for matching, and the tender document is in Word or PDF format;

[0085] Use the parsing mechanism to parse the product parameter information table, extract the column names and their corresponding values, verify the integrity of the data and the correctness of the format, and then store the extracted data in the database;

[0086] Preprocess the tender document, including format analysis, PDF file conversion, file segmentation, metadata update, and create a FAISS vector database using the selected large model; specifically include: analyze the format of the tender document, if it is in PDF format, further determine whether it is a text PDF or a picture PDF; for a text PDF, directly perform conversion processing; for a picture PDF, first convert it to a PNG picture, and then convert the picture to a Word document; use the Loader in the langchain library to load and segment the Word version of the tender document, supporting segmentation in two ways: by character count and by paragraph; add a document number to the metadata of each tender document; use the selected large model to create a FAISS vector database for the tender document for efficient retrieval later.

[0087] Provide a user interface that allows the user to select any one or more columns in the product parameter information table as the matching basis and select the tender documents to be analyzed;

[0088] Start the matching mechanism, use the bge-es-reranker retrieval technology of the large model to perform efficient retrieval on the selected product parameter information and the tender document vector library, introduce the large model to deeply evaluate the retrieval results, give a similarity score, and finally store the matching information in the database and display the matching results in the original format; specifically include: extract the corresponding content as the matching basis according to the columns in the product parameter information table selected by the user; use the bge-es-reranker retrieval technology to perform circular retrieval on the extracted product parameter information and the tender document vector library, and return the tender document fragment with the highest similarity to the product parameter; merge the product parameter information with the retrieved tender document fragment and the refined guiding text, introduce the large model for in-depth analysis and similarity judgment; according to the evaluation results of the large model, give the similarity score of each row of product parameter information and the corresponding tender document fragment.

[0089] The method further includes: storing the matching information of each line in an orderly manner in the database in the form of a dictionary to ensure the integrity and security of the data; after all the retrieval and judgment processes are completed, extracting the execution results of this time from the database and displaying them in the original format of the product parameter information table to provide users with a clear and intuitive experience of viewing the matching results.

[0090] Implement a strict permission management mechanism to ensure that each user account can only view the materials uploaded by itself, protecting the security and privacy of the tender documents; support the invocation of multiple leading large models, such as Inspur Yunhai Ruo large model, Alibaba Qianwen and GPT, to ensure the accuracy and efficiency of the analysis results; the database supports mongodb and redis types for storing the parsed product parameter information, vector database and matching results, and the design considers the scalability and query efficiency of the data.

[0091] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A parameter information matching system for tender documents of the service industry based on large models, characterized in that: The system includes: A human - machine interface module, which enables business users and administrators to access the system through a browser and enter system pages with different functions and layouts according to the user type; A product parameter information upload module, which allows users to import a custom product parameter information table in a specified area, and parses, validates, and stores this information in the database; A tender document upload module, which supports users to upload tender documents in PDF or Word format, performs format analysis, pre - processing, splitting, and metadata update, and creates a vector database; A tender document product parameter information matching module, which uses large - model technology to enable users to select columns in the product parameter information table as the matching basis, performs efficient retrieval with the tender document vector library, introduces a large model to evaluate similarity, stores the matching information in the database, and finally displays the matching results in the original format.

2. The matching system for product parameter information of tender documents based on large models in the service industry according to claim 1, characterized in that: The human - machine interface module includes: A business user interface, which provides functions such as uploading tender documents, importing product parameter information tables, and viewing matching results; An administrator interface, which is used for system maintenance, including model library update, database management, and user account management and permission allocation.

3. The matching system for product parameter information of a tender document based on a large model in the service industry according to claim 2, characterized in that: The tender document upload module further includes: A file format analysis unit, which is used to determine whether the uploaded file is in Word or PDF format, and identify the text or picture type of PDF files; A PDF processing unit, which directly converts text PDFs, and for picture PDFs, first converts them into PNG pictures and then into Word documents; A file splitting unit, which uses the Loader in the langchain library to load and split Word - format tender documents, supporting splitting in two ways: by character count and by paragraph; A vector database creation unit, which uses a selected embedding large model to create a FAISS vector database for tender documents.

4. The matching system for product parameter information of tender documents based on large models in the service industry according to claim 3, characterized in that: The tender document product parameter information matching module includes: A matching basis selection unit, which allows users to freely select any column or multiple columns in the product parameter information table as the matching basis; An efficient retrieval unit, which uses bge - es - reranker retrieval technology to perform circular retrieval of the selected product parameter information with the tender document vector library; A similarity evaluation unit, which introduces a large model to deeply analyze and judge similarity by merging product parameter information and retrieved tender document fragments through refined guiding text; A result storage and display unit, which stores the matching information of each row in the database in dictionary form, and after all processes are completed, displays the matching results in the original format of the product parameter information table.

5. The matching system for product parameter information of a tender document based on a large model in the service industry according to claim 4, characterized in that: The system further includes: A security management module, which implements a strict permission management mechanism to ensure that each user account can only view the materials uploaded by itself, protecting the security and privacy of tender documents; A database module, which supports multiple types such as mongodb and redis, and is used to store parsed product parameter information, vector databases, and matching results, with design considerations for data scalability and query efficiency.

6. A matching method for a matching system of product parameter information of a tender document based on a large model in the service industry according to claim 5, characterized in that: ​ ​ Use the parsing mechanism to parse the product parameter information table, extract the column names and their corresponding values, verify the integrity of the data and the correctness of the format, and then store the extracted data in the database; Preprocess the bidding documents, including format analysis, PDF file conversion, file segmentation, metadata update, and create a FAISS vector database using the selected large model; Provide a user interface that allows users to select any one or more columns in the product parameter information table as the matching basis and select the bidding documents to be analyzed; Start the matching mechanism. Using the bge-es-reranker retrieval technology of the large model, efficiently retrieve the selected product parameters information from the bidding document vector library, and introduce the large model to deeply evaluate the retrieval results, give a similarity score, and finally store the matching information in the database and display the matching results in the original format.

7. A matching method according to claim 6, characterized in that: The steps for preprocessing the bidding documents specifically include: Analyze the format of the bidding documents. If it is in PDF format, further determine whether it is a text PDF or a picture PDF; For text PDFs, directly perform conversion processing; for picture PDFs, first convert them to PNG pictures, and then convert the pictures to Word documents; Use the Loader in the langchain library to load and split the Word version of the bidding documents, supporting splitting in two ways: by the number of characters and by paragraphs; Add a document number to the metadata of each bidding document; Create a FAISS vector database for the bidding documents using the selected large model for efficient retrieval later.

8. A matching method according to claim 7, characterized in that: The steps for starting the matching mechanism specifically include: Extract the corresponding content as the matching basis according to the columns selected by the user in the product parameter information table; Use the bge-es-reranker retrieval technology to circularly retrieve the extracted product parameter information from the bidding document vector library, and return the bidding document fragment with the highest similarity to the product parameters; Merge the product parameter information and the retrieved bidding document fragment with the refined guiding text, and introduce the large model for in-depth analysis and similarity judgment; According to the evaluation results of the large model, give the similarity score for each row of product parameter information and the corresponding bidding document fragment.

9. A matching method according to claim 8, characterized in that: The method further includes: Store the matching information of each row in the database in the form of a dictionary in an orderly manner to ensure the integrity and security of the data; After all the retrieval and judgment processes are completed, extract the execution results of this time from the database and display them in the original format of the product parameter information table, providing users with a clear and intuitive experience of viewing the matching results.

10. A matching method according to claim 9, characterized in that: The method further includes: Implement a strict permission management mechanism to ensure that each user account can only view the materials uploaded by themselves, protecting the security and privacy of the bidding documents; Support the invocation of multiple leading large models, such as Inspur Cloud Sea Ruo Model, Alibaba Qianwen, and GPT, to ensure the accuracy and efficiency of the analysis results; The database supports mongodb and redis types, which are used to store the parsed product parameter information, vector database, and matching results. The design takes into account the scalability and query efficiency of the data.

Images