foundry-user-preferences

Foundry User Preferences & Memory Guidelines

User-Specific Development Preferences

Based on established user memories and preferences, these guidelines ensure consistent alignment with user expectations and workflow patterns.

Function Return Philosophy

User Memory #7066344: Functions should only return what is needed, avoiding unused return values.

Implementation Guidelines

• Design response structs with only essential fields
• Use #[serde(skip_serializing_if = "Option::is_none")] for optional fields
• Avoid "kitchen sink" response objects that include unused data
• Focus responses on immediate workflow needs

Response Design Pattern

#[derive(Serialize, Debug)]
pub struct ProjectResponse {
    // Essential data
    pub project: ProjectData,
    pub validation_status: String,
    pub next_steps: Vec<String>,

    // Optional fields only when needed
    #[serde(skip_serializing_if = "Option::is_none")]
    pub workflow_hints: Option<Vec<String>>,

    // Avoid: unused metadata, verbose debugging info, redundant data
}

Examples of Preferred Returns

// ✅ Good: Returns only what's needed
pub fn load_project_summary(name: &str) -> Result<String> {
    // Returns just the summary content
}

// ❌ Avoid: Returns unnecessary data
pub fn load_project_summary(name: &str) -> Result<(String, PathBuf, Metadata, Vec<String>)> {
    // Summary, path, metadata, and file list - too much!
}

Content Creation Philosophy

User Memory #7065075: For the Foundry CLI project, the summary must be provided as an argument by the LLM when initiating a project; the CLI should not perform any summarization itself.

Implementation Requirements

• Never implement automatic content generation or summarization
• Always require LLMs to provide complete content as arguments
• Never derive content from other content (e.g., auto-generate summary from vision)
• Always validate content format and length, but not content quality