release(v0.2.0): streaming, MCP protocol, Browser Hand, security enhancements
## Major Features ### Streaming Response System - Implement LlmDriver trait with `stream()` method returning async Stream - Add SSE parsing for Anthropic and OpenAI API streaming - Integrate Tauri event system for frontend streaming (`stream:chunk` events) - Add StreamChunk types: Delta, ToolStart, ToolEnd, Complete, Error ### MCP Protocol Implementation - Add MCP JSON-RPC 2.0 types (mcp_types.rs) - Implement stdio-based MCP transport (mcp_transport.rs) - Support tool discovery, execution, and resource operations ### Browser Hand Implementation - Complete browser automation with Playwright-style actions - Support Navigate, Click, Type, Scrape, Screenshot, Wait actions - Add educational Hands: Whiteboard, Slideshow, Speech, Quiz ### Security Enhancements - Implement command whitelist/blacklist for shell_exec tool - Add SSRF protection with private IP blocking - Create security.toml configuration file ## Test Improvements - Fix test import paths (security-utils, setup) - Fix vi.mock hoisting issues with vi.hoisted() - Update test expectations for validateUrl and sanitizeFilename - Add getUnsupportedLocalGatewayStatus mock ## Documentation Updates - Update architecture documentation - Improve configuration reference - Add quick-start guide updates Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
iven
@@ -330,16 +330,160 @@ fn filter_by_proactivity(alerts: &[HeartbeatAlert], level: &ProactivityLevel) ->
|
||||
|
||||
// === Built-in Checks ===
|
||||
|
||||
/// Check for pending task memories (placeholder - would connect to memory store)
|
||||
fn check_pending_tasks(_agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
// In full implementation, this would query the memory store
|
||||
// For now, return None (no tasks)
|
||||
/// Pattern detection counters (shared state for personality detection)
|
||||
use std::collections::HashMap as StdHashMap;
|
||||
use std::sync::RwLock;
|
||||
use std::sync::OnceLock;
|
||||
|
||||
/// Global correction counters
|
||||
static CORRECTION_COUNTERS: OnceLock<RwLock<StdHashMap<String, usize>>> = OnceLock::new();
|
||||
|
||||
/// Global memory stats cache (updated by frontend via Tauri command)
|
||||
/// Key: agent_id, Value: (task_count, total_memories, storage_bytes)
|
||||
static MEMORY_STATS_CACHE: OnceLock<RwLock<StdHashMap<String, MemoryStatsCache>>> = OnceLock::new();
|
||||
|
||||
/// Cached memory stats for an agent
|
||||
#[derive(Clone, Debug, Default)]
|
||||
pub struct MemoryStatsCache {
|
||||
pub task_count: usize,
|
||||
pub total_entries: usize,
|
||||
pub storage_size_bytes: usize,
|
||||
pub last_updated: Option<String>,
|
||||
}
|
||||
|
||||
fn get_correction_counters() -> &'static RwLock<StdHashMap<String, usize>> {
|
||||
CORRECTION_COUNTERS.get_or_init(|| RwLock::new(StdHashMap::new()))
|
||||
}
|
||||
|
||||
fn get_memory_stats_cache() -> &'static RwLock<StdHashMap<String, MemoryStatsCache>> {
|
||||
MEMORY_STATS_CACHE.get_or_init(|| RwLock::new(StdHashMap::new()))
|
||||
}
|
||||
|
||||
/// Update memory stats cache for an agent
|
||||
/// Call this from frontend via Tauri command after fetching memory stats
|
||||
pub fn update_memory_stats_cache(agent_id: &str, task_count: usize, total_entries: usize, storage_size_bytes: usize) {
|
||||
let cache = get_memory_stats_cache();
|
||||
if let Ok(mut cache) = cache.write() {
|
||||
cache.insert(agent_id.to_string(), MemoryStatsCache {
|
||||
task_count,
|
||||
total_entries,
|
||||
storage_size_bytes,
|
||||
last_updated: Some(chrono::Utc::now().to_rfc3339()),
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
/// Get memory stats for an agent
|
||||
fn get_cached_memory_stats(agent_id: &str) -> Option<MemoryStatsCache> {
|
||||
let cache = get_memory_stats_cache();
|
||||
if let Ok(cache) = cache.read() {
|
||||
cache.get(agent_id).cloned()
|
||||
} else {
|
||||
None
|
||||
}
|
||||
}
|
||||
|
||||
/// Record a user correction for pattern detection
|
||||
/// Call this when user corrects agent behavior
|
||||
pub fn record_user_correction(agent_id: &str, correction_type: &str) {
|
||||
let key = format!("{}:{}", agent_id, correction_type);
|
||||
let counters = get_correction_counters();
|
||||
if let Ok(mut counters) = counters.write() {
|
||||
*counters.entry(key).or_insert(0) += 1;
|
||||
}
|
||||
}
|
||||
|
||||
/// Get and reset correction count
|
||||
fn get_correction_count(agent_id: &str, correction_type: &str) -> usize {
|
||||
let key = format!("{}:{}", agent_id, correction_type);
|
||||
let counters = get_correction_counters();
|
||||
if let Ok(mut counters) = counters.write() {
|
||||
counters.remove(&key).unwrap_or(0)
|
||||
} else {
|
||||
0
|
||||
}
|
||||
}
|
||||
|
||||
/// Check all correction patterns for an agent
|
||||
fn check_correction_patterns(agent_id: &str) -> Vec<HeartbeatAlert> {
|
||||
let patterns = [
|
||||
("communication_style", "简洁", "用户偏好简洁回复，建议减少冗长解释"),
|
||||
("tone", "轻松", "用户偏好轻松语气，建议减少正式用语"),
|
||||
("detail_level", "概要", "用户偏好概要性回答，建议先给结论再展开"),
|
||||
("language", "中文", "用户语言偏好，建议优先使用中文"),
|
||||
("code_first", "代码优先", "用户偏好代码优先，建议先展示代码再解释"),
|
||||
];
|
||||
|
||||
let mut alerts = Vec::new();
|
||||
for (pattern_type, _keyword, suggestion) in patterns {
|
||||
let count = get_correction_count(agent_id, pattern_type);
|
||||
if count >= 3 {
|
||||
alerts.push(HeartbeatAlert {
|
||||
title: "人格改进建议".to_string(),
|
||||
content: format!("{} (检测到 {} 次相关纠正)", suggestion, count),
|
||||
urgency: Urgency::Medium,
|
||||
source: "personality-improvement".to_string(),
|
||||
timestamp: chrono::Utc::now().to_rfc3339(),
|
||||
});
|
||||
}
|
||||
}
|
||||
alerts
|
||||
}
|
||||
|
||||
/// Check for pending task memories
|
||||
/// Uses cached memory stats to detect task backlog
|
||||
fn check_pending_tasks(agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
if let Some(stats) = get_cached_memory_stats(agent_id) {
|
||||
// Alert if there are 5+ pending tasks
|
||||
if stats.task_count >= 5 {
|
||||
return Some(HeartbeatAlert {
|
||||
title: "待办任务积压".to_string(),
|
||||
content: format!("当前有 {} 个待办任务未完成，建议处理或重新评估优先级", stats.task_count),
|
||||
urgency: if stats.task_count >= 10 {
|
||||
Urgency::High
|
||||
} else {
|
||||
Urgency::Medium
|
||||
},
|
||||
source: "pending-tasks".to_string(),
|
||||
timestamp: chrono::Utc::now().to_rfc3339(),
|
||||
});
|
||||
}
|
||||
}
|
||||
None
|
||||
}
|
||||
|
||||
/// Check memory storage health (placeholder)
|
||||
fn check_memory_health(_agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
// In full implementation, this would check memory stats
|
||||
/// Check memory storage health
|
||||
/// Uses cached memory stats to detect storage issues
|
||||
fn check_memory_health(agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
if let Some(stats) = get_cached_memory_stats(agent_id) {
|
||||
// Alert if storage is very large (> 50MB)
|
||||
if stats.storage_size_bytes > 50 * 1024 * 1024 {
|
||||
return Some(HeartbeatAlert {
|
||||
title: "记忆存储过大".to_string(),
|
||||
content: format!(
|
||||
"记忆存储已达 {:.1}MB，建议清理低重要性记忆或归档旧记忆",
|
||||
stats.storage_size_bytes as f64 / (1024.0 * 1024.0)
|
||||
),
|
||||
urgency: Urgency::Medium,
|
||||
source: "memory-health".to_string(),
|
||||
timestamp: chrono::Utc::now().to_rfc3339(),
|
||||
});
|
||||
}
|
||||
|
||||
// Alert if too many memories (> 1000)
|
||||
if stats.total_entries > 1000 {
|
||||
return Some(HeartbeatAlert {
|
||||
title: "记忆条目过多".to_string(),
|
||||
content: format!(
|
||||
"当前有 {} 条记忆，可能影响检索效率，建议清理或归档",
|
||||
stats.total_entries
|
||||
),
|
||||
urgency: Urgency::Low,
|
||||
source: "memory-health".to_string(),
|
||||
timestamp: chrono::Utc::now().to_rfc3339(),
|
||||
});
|
||||
}
|
||||
}
|
||||
None
|
||||
}
|
||||
|
||||
@@ -358,38 +502,43 @@ fn check_idle_greeting(_agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
///
|
||||
/// When threshold is reached, proposes a personality change via the identity system.
|
||||
fn check_personality_improvement(agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
// Pattern detection heuristics
|
||||
// In full implementation, this would:
|
||||
// 1. Query memory for recent "correction" type interactions
|
||||
// 2. Count frequency of similar corrections
|
||||
// 3. If >= 3 similar corrections, trigger proposal
|
||||
|
||||
// Common correction patterns to detect
|
||||
let correction_patterns = [
|
||||
("啰嗦|冗长|简洁", "用户偏好简洁回复", "communication_style"),
|
||||
("正式|随意|轻松", "用户偏好轻松语气", "tone"),
|
||||
("详细|概括|摘要", "用户偏好概要性回答", "detail_level"),
|
||||
("英文|中文|语言", "用户语言偏好", "language"),
|
||||
("代码|解释|说明", "用户偏好代码优先", "code_first"),
|
||||
];
|
||||
|
||||
// Placeholder: In production, query memory store for these patterns
|
||||
// For now, return None (no pattern detected)
|
||||
let _ = (agent_id, correction_patterns);
|
||||
None
|
||||
// Check all correction patterns and return the first one that triggers
|
||||
let alerts = check_correction_patterns(agent_id);
|
||||
alerts.into_iter().next()
|
||||
}
|
||||
|
||||
/// Check for learning opportunities from recent conversations
|
||||
///
|
||||
/// Identifies opportunities to capture user preferences or behavioral patterns
|
||||
/// that could enhance agent effectiveness.
|
||||
fn check_learning_opportunities(_agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
// In full implementation, this would:
|
||||
// 1. Analyze recent conversations for explicit preferences
|
||||
// 2. Detect implicit preferences from user reactions
|
||||
// 3. Suggest memory entries or identity changes
|
||||
fn check_learning_opportunities(agent_id: &str) -> Option<HeartbeatAlert> {
|
||||
// Check if any correction patterns are approaching threshold
|
||||
let counters = get_correction_counters();
|
||||
let mut approaching_threshold: Vec<String> = Vec::new();
|
||||
|
||||
None
|
||||
if let Ok(counters) = counters.read() {
|
||||
for (key, count) in counters.iter() {
|
||||
if key.starts_with(&format!("{}:", agent_id)) && *count >= 2 && *count < 3 {
|
||||
let pattern_type = key.split(':').nth(1).unwrap_or("unknown").to_string();
|
||||
approaching_threshold.push(pattern_type);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if !approaching_threshold.is_empty() {
|
||||
Some(HeartbeatAlert {
|
||||
title: "学习机会".to_string(),
|
||||
content: format!(
|
||||
"检测到用户可能有偏好调整倾向 ({})，继续观察将触发人格改进建议",
|
||||
approaching_threshold.join(", ")
|
||||
),
|
||||
urgency: Urgency::Low,
|
||||
source: "learning-opportunities".to_string(),
|
||||
timestamp: chrono::Utc::now().to_rfc3339(),
|
||||
})
|
||||
} else {
|
||||
None
|
||||
}
|
||||
}
|
||||
|
||||
// === Tauri Commands ===
|
||||
@@ -493,6 +642,29 @@ pub async fn heartbeat_get_history(
|
||||
Ok(engine.get_history(limit.unwrap_or(20)).await)
|
||||
}
|
||||
|
||||
/// Update memory stats cache for heartbeat checks
|
||||
/// This should be called by the frontend after fetching memory stats
|
||||
#[tauri::command]
|
||||
pub async fn heartbeat_update_memory_stats(
|
||||
agent_id: String,
|
||||
task_count: usize,
|
||||
total_entries: usize,
|
||||
storage_size_bytes: usize,
|
||||
) -> Result<(), String> {
|
||||
update_memory_stats_cache(&agent_id, task_count, total_entries, storage_size_bytes);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// Record a user correction for personality improvement detection
|
||||
#[tauri::command]
|
||||
pub async fn heartbeat_record_correction(
|
||||
agent_id: String,
|
||||
correction_type: String,
|
||||
) -> Result<(), String> {
|
||||
record_user_correction(&agent_id, &correction_type);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
Reference in New Issue
Block a user