Getting Started

What You Are About to Run

AIVEX Feed is a self-contained Python pipeline for financial news intelligence. It collects news from multiple Turkish and English sources, extracts article bodies, detects language, translates Turkish articles into English, scores sentiment with a financial language model, links articles to tracked companies, and writes the result to one database table and one event stream.

It can run three ways:

• Scheduled job — invoked on a schedule by your own scheduler
• Library call — the entire pipeline is a single run(db, …) function call
• Desktop application — a Windows tray app that schedules runs and reports source health

Private Repository Notice: AIVEX Feed is currently private and not packaged for distribution. To request access or a demo, contact the team.

Prerequisites

• Python 3.10 or newer
• MySQL (local or remote)
• Redis
• A .env file with the required environment variables

Optional:

• Ollama — supplies the daily narrative text. Without it, narrative fields stay empty and every number is computed exactly the same.
• Reddit credentials — without them the Reddit source is skipped and the run continues.

First Run

The first run downloads three models: a financial sentiment model, a Turkish→English translation model, and a language identifier. Expect it to be slow; later runs use the cached models.

Configuration

The pipeline is configured from environment variables, including the three confidence thresholds that gate scoring: sentiment confidence, title confidence, and language confidence. The database backend is injected — see Integration for the contract.

Where to Go Next

• Outputs — the two surfaces everything is read from
• Pipeline — the six stages every run executes
• Turkish Path — how Turkish news is scored safely
• Integration — attaching Feed to an existing stack

AIVEX Feed produces research artifacts for informational purposes only. Its outputs are not financial, investment, or trading advice.