Practical overview
When it fits
Use the repository when you need a starting point for learning or planning a financial annual-report question-answering workflow, and can select and assemble the required data, model and database components.
Who it is for
- Developers planning a financial question-answering prototype based on annual reports.
- Learners studying financial-language-model data preparation and ChatGLM materials.
- Teams exploring workflows for extracting and querying company financial information.
Documented uses
- Convert annual-report PDFs into text while retaining table structure.
- Organize company, financial-statement and indicator data for database storage.
- Explore a question-to-query-to-answer workflow for financial information.
- Study model fine-tuning approaches and ChatGLM-6B course materials.
Requirements & limits
The repository is publicly visible on GitHub and describes the project as open source.
- The project describes workflows, datasets, course material and contributed solutions; the repository evidence does not establish a single deployable model or API.
- The outline names SQL, MongoDB and Elasticsearch for storage, and P-Tuning v2 and LoRA as fine-tuning strategies.
- Its annual-report workflow includes PDF-to-text conversion, data segmentation, calculations such as growth rates and industry comparisons, and database storage.
- The repository includes courses for ChatGLM-6B generations 1, 2 and 3.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Prepare financial annual-report data for a question-answering workflow
Organize annual-report content and financial data for database-backed question answering.
Before you start
- Annual-report PDF files or other financial-report content.
- A database destination; the project describes SQL, Mongo, and Elasticsearch (ES) storage.
Steps
Convert report PDFs to TXT while preserving tables and merging cells, using the repository's PDF-to-TXT tool.
Split the content into basic information, financial data, and combined information such as financial indicators.
Process the data by calculating relevant formulas, growth rates, industry averages, and rankings.
Store the prepared data in a database; the described workflow includes table creation and storage.
Expected outcome
Prepared financial-report data is organized and stored for use in a subsequent question-answering workflow.
Guide limitations
- The README outlines the stages but does not provide detailed execution instructions, database configuration, or tool prerequisites.
- The described question-answering workflow is a separate next stage; this recipe does not execute it.
Follow the documented financial question-answering stages
Trace how a user question is intended to become an answer using database results.
Before you start
- Prepared data stored in a database.
- A question-answering workflow that can prepare prompts, generate queries, access the database, and generate answers.
Steps
Enter a user question.
Prepare a prompt based on the question.
Generate a database query using a method selected according to GPU utilization, then query the database.
Generate an answer from the question and query results.
Expected outcome
The documented workflow produces an answer by combining the user's question with database query results.
Guide limitations
- The README does not specify runnable commands, prompt templates, query-generation details, GPU thresholds, or setup instructions.
- The repository evidence does not establish that the project includes a deployable model or API.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- MetaGLM/FinGLM · Official homepage title · retained in its source language