How to optimize the performance of locally deployed AI financial analytics tools?

2025-08-21

Performance Optimization Solutions

The following optimizations are provided for performance issues that may be encountered when running DeepAgents locally:

Model Selection Optimization: Balance accuracy and performance by prioritizing models with moderate parameter counts (e.g., gpt-oss with 7B parameters)
Hardware Configuration Recommendations: At least 16GB of RAM, with GPU acceleration to dramatically improve responsiveness
Parallel Processing Tuning: Modify the max_workers parameter in config.yml to control the number of concurrent subintelligences (4-6 threads recommended)

Specific operations:

Select the light version of the model during ollama pull (add :7b suffix)
Start the analysis task only after closing other resource-hogging programs
Non-essential sub-intelligents can be turned off for simple analysis (modify agent_dispatcher.py configuration)

Alternative: If device performance is insufficient, consider cloud server deployment with local access via port mapping.