feat: per-user auto-memory with ResponseContext

Three memory channels: hidden tool (sol.memory.set/get in scripts),
pre-response injection (relevant memories loaded into system prompt),
and post-response extraction (ministral-3b extracts facts after each
response). User isolation enforced at Rust level — user_id derived
from Matrix sender, never from script arguments.

New modules: context (ResponseContext), memory (schema, store, extractor).
ResponseContext threaded through responder → tools → script runtime.
OpenSearch index sol_user_memory created on startup alongside archive.

src/config.rs

@@ -25,6 +25,8 @@ pub struct OpenSearchConfig {
     pub flush_interval_ms: u64,
     #[serde(default = "default_embedding_pipeline")]
     pub embedding_pipeline: String,
+    #[serde(default = "default_memory_index")]
+    pub memory_index: String,
 }
 
 #[derive(Debug, Clone, Deserialize)]
@@ -59,6 +61,30 @@ pub struct BehaviorConfig {
     pub backfill_on_join: bool,
     #[serde(default = "default_backfill_limit")]
     pub backfill_limit: usize,
+    #[serde(default)]
+    pub instant_responses: bool,
+    #[serde(default = "default_cooldown_after_response_ms")]
+    pub cooldown_after_response_ms: u64,
+    #[serde(default = "default_evaluation_context_window")]
+    pub evaluation_context_window: usize,
+    #[serde(default = "default_detect_sol_in_conversation")]
+    pub detect_sol_in_conversation: bool,
+    #[serde(default)]
+    pub evaluation_prompt_active: Option<String>,
+    #[serde(default)]
+    pub evaluation_prompt_passive: Option<String>,
+    #[serde(default = "default_reaction_threshold")]
+    pub reaction_threshold: f32,
+    #[serde(default = "default_reaction_enabled")]
+    pub reaction_enabled: bool,
+    #[serde(default = "default_script_timeout_secs")]
+    pub script_timeout_secs: u64,
+    #[serde(default = "default_script_max_heap_mb")]
+    pub script_max_heap_mb: usize,
+    #[serde(default)]
+    pub script_fetch_allowlist: Vec<String>,
+    #[serde(default = "default_memory_extraction_enabled")]
+    pub memory_extraction_enabled: bool,
 }
 
 fn default_batch_size() -> usize { 50 }
@@ -68,15 +94,24 @@ fn default_model() -> String { "mistral-medium-latest".into() }
 fn default_evaluation_model() -> String { "ministral-3b-latest".into() }
 fn default_research_model() -> String { "mistral-large-latest".into() }
 fn default_max_tool_iterations() -> usize { 5 }
-fn default_response_delay_min_ms() -> u64 { 2000 }
-fn default_response_delay_max_ms() -> u64 { 8000 }
+fn default_response_delay_min_ms() -> u64 { 100 }
+fn default_response_delay_max_ms() -> u64 { 2300 }
 fn default_spontaneous_delay_min_ms() -> u64 { 15000 }
 fn default_spontaneous_delay_max_ms() -> u64 { 60000 }
-fn default_spontaneous_threshold() -> f32 { 0.7 }
+fn default_spontaneous_threshold() -> f32 { 0.85 }
+fn default_cooldown_after_response_ms() -> u64 { 15000 }
+fn default_evaluation_context_window() -> usize { 25 }
+fn default_detect_sol_in_conversation() -> bool { true }
+fn default_reaction_threshold() -> f32 { 0.6 }
+fn default_reaction_enabled() -> bool { true }
 fn default_room_context_window() -> usize { 30 }
 fn default_dm_context_window() -> usize { 100 }
 fn default_backfill_on_join() -> bool { true }
 fn default_backfill_limit() -> usize { 10000 }
+fn default_script_timeout_secs() -> u64 { 5 }
+fn default_script_max_heap_mb() -> usize { 64 }
+fn default_memory_index() -> String { "sol_user_memory".into() }
+fn default_memory_extraction_enabled() -> bool { true }
 
 impl Config {
     pub fn load(path: &str) -> anyhow::Result<Self> {
@@ -155,19 +190,23 @@ backfill_limit = 5000
         assert_eq!(config.opensearch.batch_size, 50);
         assert_eq!(config.opensearch.flush_interval_ms, 2000);
         assert_eq!(config.opensearch.embedding_pipeline, "tuwunel_embedding_pipeline");
+        assert_eq!(config.opensearch.memory_index, "sol_user_memory");
         assert_eq!(config.mistral.default_model, "mistral-medium-latest");
         assert_eq!(config.mistral.evaluation_model, "ministral-3b-latest");
         assert_eq!(config.mistral.research_model, "mistral-large-latest");
         assert_eq!(config.mistral.max_tool_iterations, 5);
-        assert_eq!(config.behavior.response_delay_min_ms, 2000);
-        assert_eq!(config.behavior.response_delay_max_ms, 8000);
+        assert_eq!(config.behavior.response_delay_min_ms, 100);
+        assert_eq!(config.behavior.response_delay_max_ms, 2300);
         assert_eq!(config.behavior.spontaneous_delay_min_ms, 15000);
         assert_eq!(config.behavior.spontaneous_delay_max_ms, 60000);
-        assert!((config.behavior.spontaneous_threshold - 0.7).abs() < f32::EPSILON);
+        assert!((config.behavior.spontaneous_threshold - 0.85).abs() < f32::EPSILON);
+        assert!(!config.behavior.instant_responses);
+        assert_eq!(config.behavior.cooldown_after_response_ms, 15000);
         assert_eq!(config.behavior.room_context_window, 30);
         assert_eq!(config.behavior.dm_context_window, 100);
         assert!(config.behavior.backfill_on_join);
         assert_eq!(config.behavior.backfill_limit, 10000);
+        assert!(config.behavior.memory_extraction_enabled);
     }
 
     #[test]