Best Practices for Using Data in Alfa Agents

Including a data source in your prompt

It is important when prompting to specify the data that you want Alfa to use to ensure your analysts know where you want them to find their information.

For example, if you want to search the news for what’s happening with Apple, you would want to specifically state that you want Alfa to read the news.

Example prompts:

• Read all news for Apple, and summarize for key developments.

• Read 10Ks and 10Qs for SPY from last quarter. Brainstorm macroeconomic impacts.

• Search the web for mentions of Coronavirus. Filter for mentions of new cases.

Note: If you do not specify the data source, Alfa may pull information that is not relevant to your intended query.

Examples of data sources you can reference in your prompts:

• Web

• News

• Exchange filings (SEC, SEDAR+ etc.)

• Earnings reports/transcripts

• Custom document uploads

• Provisioned research (sell-side reports)

• Time-series data (fundamentals, economic, technical)

• Corporate presentations

Click here to see Alfa's full list of data sources and data providers.

How the data you select can impact the cost of an agent

Typically, the more data a prompt requires Alfa to read through, the higher the token usage for that agent.

To help keep your costs down, we recommend starting small and then scaling up once you get an output that you’re happy with.

• For example, instead of asking Alfa to perform an analysis on the entire S&P 500, start by asking it to perform the same analysis on a much smaller group of stocks.

• Once you’re happy with the output you receive, then you can ask it to perform the same analysis on the entire S&P 500.

Alfa can pull data from many sources, and ensuring you reference the correct data source is key to returning beneficial results.