📈 Global Markets Intelligence Pipeline (AI Powered)

This Apify Actor processes global markets news from diverse RSS feeds and specialized financial APIs. It transforms raw headlines into structured, actionable market intelligence using a reliable, two-stage Gemini (Google) AI pipeline that leverages real-time web search for maximum accuracy.

This pipeline replaces manual ticker hunting and unreliable content extraction with a sophisticated, cost-optimized system designed for high-quality data and downstream content generation.

✨ Value Proposition

The actor delivers highly structured, accurate financial data by resolving two major LLM pain points:

1. Factual Accuracy (Pay Point 1): Uses Gemini with Grounding with Google Search to analyze raw headlines by pulling real-time web context. This reduces hallucinations and ensures the sentiment, category, and entities are based on the latest facts.
2. Structured Output (Mandatory): The results are normalized into a clean, easy-to-use JSON format suitable for direct ingestion by market intelligence tools and content engines.

⚙️ How the Pipeline Works

The actor operates in three primary stages:

1. Parallel Data Fetching (Dual Source)

The actor executes two data fetches concurrently, combining the results for maximum coverage:

• Traditional RSS Feeds: Collects articles from over 50 top financial and general news sources (FT, Bloomberg, Investing.com, etc.).
• Alpha Vantage News: Fetches highly relevant, pre-filtered market news directly from the Alpha Vantage NEWS_SENTIMENT API based on the user-selected topic.

2. Gemini Grounding & Analysis (Pay Point 1)

For every collected article, the pipeline executes a two-step Gemini process to generate structured analysis:

• Step A: RAG (Grounding): The article's title is sent to the gemini-2.5-flash model with the googleSearch tool enabled. This forces the model to perform a real-time web search and generate a grounded, factual text analysis.
• Step B: Extraction: The grounded analysis text from Step A is then sent to a second Gemini call, configured to return a structured JSON object with the required fields: sentiment, category, key_entities, and numeric_score.

3. Optional Summarization (Pay Point 2)

If enabled, the model generates a concise, market-focused summary for the summary field, optimizing the output for content creators.

---

📥 Input Configuration (INPUT_SCHEMA.json)

Parameter	Type	Default	Description
source	select (Required)	all	The news content filtering strategy. Used to filter RSS feeds and set the topic query for the Alpha Vantage API. Selecting 'Alpha Vantage News' limits the actor to only the AV source.
customFeedUrl	string (Optional)	None	Provides a custom RSS feed URL when source is set to custom.
maxArticles	integer	10	The maximum number of articles fetched per source (RSS and Alpha Vantage). Total output can be up to twice this value.
useSummarization	boolean	true	Enables the second LLM call (Pay Point 2) to generate a new, concise summary for each article. Recommended option for high-quality downstream content.
runTestMode	boolean	false	Enables zero-cost local testing by skipping all external API calls (Gemini and Alpha Vantage).

---

📊 Output Dataset Schema (DatasetRecord)

The actor pushes one record per analyzed article to the default dataset.

Field Name	Type	Description
source	string	The original source name (e.g., 'Financial Times (FT)', 'Alpha Vantage').
title	string	The headline of the article.
url	HttpUrl	The URL of the original article.
published	string (Optional)	The publication date/time.
summary	string (Optional)	The LLM-generated summary (Pay Point 2) or the original RSS summary.
sentiment	string (Optional)	LLM Analysis result: Positive, Neutral, or Negative.
category	string (Optional)	LLM Analysis result: Thematic category (e.g., Monetary Policy, Technology/FinTech).
key_entities	list<string> (Optional)	Key companies, people, or macroeconomic terms identified in the analysis.
gdelt_tone	float (Optional)	Repurposed field for the LLM's numerical impact score (-10.0 to +10.0).

---

💰 Cost Structure

The pipeline operates with two primary variable cost points (LLM Pay Points) powered by the Gemini API:

Pay Point	Service	Purpose
1. LLM Analysis	Gemini (gemini-2.5-flash)	MANDATORY. Performs RAG with Google Search Grounding and extracts structured sentiment/category/entities. This is the main expense for accuracy.
2. LLM Summarization	Gemini (gemini-2.5-flash)	OPTIONAL. Generates the final, concise summary for the output record. Enabled via useSummarization=true.