Configuration#
The default parameters of the app can be set up in the config.yaml
Some of the most important parameters are explained here:
- Embedding model: An embedding model can be set for the news data and the SEC filing separately.
- Similarity Top K: This setting determines how many of similar chunks of data are sent to the LLM to solve the task.
- LLM Provider: Mainly choose
groq
oropenai
for providing LLM model. HuggingFace support is not recommended because larger models are generally not available. - LLM Model: The LLM to be used for the agents in the crew and the query engines. The default is
llama3-8b-8192
fromgroq
. - LLM Temperature: This controls the randomness of an LLM. 0 is more deterministic, while 1 is more random.
- LLM Max tokens: Limit for the number of tokens to be generated by the LLM.
- tickers_market_cap_exclude: The market cap categories that are excluded from the
financedatabase
table for picking ticker.