Home
Softono
ai_digest_finance

ai_digest_finance

Open source Python
Website GitHub
18
Stars
0
Forks
0
Issues
1
Watchers
4 weeks
Last Commit
AI Agents Data Pipelines & ETL Data Visualisation

About ai_digest_finance

aidigestfinance is an automated research summarization tool that generates monthly PDF reports on scientific articles in finance and economics. It pulls recent articles from the OpenAlex database via API, stores metadata in a PostgreSQL database, and uses Google's Gemini Pro model with its large context window to analyze the content. The system produces structured PDF documents containing general insights, key findings, emerging themes, commonalities, future research directions, and a list of source articles. The tool is designed for researchers, analysts, and professionals who want concise monthly overviews of trends and developments in economic and financial research without manually reviewing dozens of papers. Configuration is done through a few simple variables specifying the report title, month, year, and topic. The modular architecture separates article extraction, database operations, AI analysis, and PDF generation into distinct scripts for easy customization and extension.

Platforms

Web Self-hosted

Languages

Python

Links

Website Source Code
n
Published by
namesjoe
Visit View Profile

README.md

View on GitHub

ai_digest_finance

Description

Project for creation PDF Summary with Monthly review about what studies and scientific articles came last month and quick overview what they have in common and what trends do they research

It generates fully automated PDF FILE with sections

  • General Insights
  • Key findings
  • Main themes or trends emerging from these articles
  • Commonalities or connections
  • Future directions
  • Sources Articles list

Example: AI_DIGEST_July_2024.pdf

We use OpenALex https://openalex.org free database of scientific articles as our main source of data via their API We also use Gemini Pro model with its vast context window to qustion it with out questions (with prompts)

Architecture and Scripts Description

./questions.py

queries_to_extract_source_articles: list of queries to extract articles from Openalex Database

questions: List of questions we ask Gemini Pro model to make sections in result PDF document

./articles.py

Here we extract articles from OpenALex and store it in Postgres DB table

CREATE TABLE public.science_articles ( id serial4 NOT NULL, doi varchar(255) NULL, author text NULL, title text NULL, published_date date NULL, abstract text NULL, CONSTRAINT science_articles_doi_key UNIQUE (doi), CONSTRAINT science_articles_pkey PRIMARY KEY (id), CONSTRAINT science_articles_title_published_date_key UNIQUE (title, published_date) );

If You gather data about multiples fields of study (FOS), there is need to adjust schema to have FOS field to identify right which articles should be used for the report

./gemini_report.py

Here we extract articles from Postges DB we gathered articles into using following variables

title = "AI DIGEST" -- Title of the document

month = "July" -- Month which be used to extract articles

year = 2024 -- year which be used to extract articles

topic = "Overview of Scientific Articles about Economy, Finance and related fields." -- Description of the PDF document

Then we use API key for GEMINI PRO MODEL with its outstanding context window over 1 million tokens and transfer extracted articles to the model, after that we asking questions (questions variable) to makes sections it our PDF Document later ANd after getting answers from Gemini we conduct PDF document

./pgres_utils.py

utils to create connection and insert\extract data

./AI_DIGEST_July_2024.pdf

<<< Example of the condicted document