openclaw - 💡(How to fix) Fix [Feature Request] Skill Graph - 按需加载依赖链，减少 Token 消耗 [1 comments, 2 participants]

TL;DR

Implement a lazy loading mechanism for skill dependencies to reduce token consumption by only loading necessary skills on demand.

Guidance

• Construct a dependency graph between skills to determine the required skills for a given task
• Implement a dynamic loading mechanism that loads skills only when their dependencies are needed
• Introduce a configuration option (e.g., skills.maxDepth) to control the maximum depth of skill loading
• Consider visualizing the skill dependency graph to better understand the relationships between skills

Example

# Simplified example of a skill dependency graph
skill_graph = {
    'L1': ['L2-1', 'L2-2'],
    'L2-1': ['L3-1'],
    'L2-2': ['L3-2']
}

def load_skills(skill_name, max_depth):
    # Load the skill and its dependencies up to the specified depth
    loaded_skills = []
    to_load = [skill_name]
    for _ in range(max_depth):
        next_to_load = []
        for skill in to_load:
            loaded_skills.append(skill)
            if skill in skill_graph:
                next_to_load.extend(skill_graph[skill])
        to_load = next_to_load
    return loaded_skills

# Load skills up to a maximum depth of 2
loaded_skills = load_skills('L1', 2)
print(loaded_skills)  # Output: ['L1', 'L2-1', 'L2-2']

Notes

The proposed solution assumes that the skill dependencies can be represented as a graph and that the dependencies are deterministic. The actual implementation may require additional considerations, such as handling cyclic dependencies or skills with multiple dependencies.

Recommendation

Apply a workaround by implementing a lazy loading mechanism for skill dependencies, as this approach is likely to reduce token consumption and improve system performance.