Fix engram_compile: fetch pinned nodes directly when vector search empty

Replace scan-by-offset fallback with engram_get_node_json calls for the
known high-salience identity nodes (family, origin). Offset-based scanning
is order-dependent and unreliable; direct ID fetch is stable regardless of
snapshot position. Ensures biographical context (Fox, Bobby, etc.) is
always in the system prompt when vector search returns nothing.

chat.el

@@ -21,8 +21,28 @@ fn engram_compile(intent: String) -> String {
 
     let act_part: String = if act_ok { activate_json } else { "" }
     let srch_part: String = if srch_ok { search_json } else { "" }
-    let sep: String = if !str_eq(act_part, "") && !str_eq(srch_part, "") { "\n" } else { "" }
-    let ctx: String = act_part + sep + srch_part
+
+    // Fallback: when vector search returns nothing (no embeddings), fetch pinned
+    // high-salience nodes by their known IDs. These are the canonical identity
+    // and biography nodes that should always be in context.
+    // engram_get_node_json(id) returns a single node as JSON or "" if missing.
+    let scan_part: String = if !act_ok && !srch_ok {
+        let family_node: String = engram_get_node_json("knw-35940684-abc4-42f0-b942-818f66b1f69a")
+        let origin_node: String = engram_get_node_json("knw-729fc901-8335-44c4-9f3a-b150b4aa0915")
+        let fam_ok: Bool = !str_eq(family_node, "") && !str_eq(family_node, "null")
+        let orig_ok: Bool = !str_eq(origin_node, "") && !str_eq(origin_node, "null")
+        let fam_str: String = if fam_ok { family_node } else { "" }
+        let orig_str: String = if orig_ok { origin_node } else { "" }
+        let sep: String = if fam_ok && orig_ok { "\n" } else { "" }
+        let combined: String = fam_str + sep + orig_str
+        if str_eq(combined, "") { "" } else { combined }
+    } else {
+        ""
+    }
+
+    let sep1: String = if !str_eq(act_part, "") && !str_eq(srch_part, "") { "\n" } else { "" }
+    let sep2: String = if (!str_eq(act_part, "") || !str_eq(srch_part, "")) && !str_eq(scan_part, "") { "\n" } else { "" }
+    let ctx: String = act_part + sep1 + srch_part + sep2 + scan_part
 
     if str_eq(ctx, "") { return "" }
 
@@ -80,6 +100,21 @@ fn hist_trim(hist: String) -> String {
     return hist
 }
 
+// clean_llm_response — strips GPT-2 BPE byte-to-unicode artifacts that vLLM
+// emits when the tokenizer hasn't decoded back to raw bytes.
+//
+//   Ġ (U+0120) = leading space on a BPE token → plain space
+//   Ċ (U+010A) = newline byte encoded as BPE token → \n
+//   ĉ (U+0109) = tab byte                        → tab (rare)
+//
+// Applied to every LLM response before it reaches callers.
+fn clean_llm_response(s: String) -> String {
+    let s1: String = str_replace(s, "Ġ", " ")
+    let s2: String = str_replace(s1, "Ċ", "\n")
+    let s3: String = str_replace(s2, "ĉ", "\t")
+    return s3
+}
+
 fn handle_chat(body: String) -> String {
     let message: String = json_get(body, "message")
     if str_eq(message, "") {
@@ -109,7 +144,8 @@ fn handle_chat(body: String) -> String {
         return "{\"error\":\"llm unavailable\",\"response\":\"\"}"
     }
 
-    let safe_response: String = json_safe(raw_response)
+    let clean_response: String = clean_llm_response(raw_response)
+    let safe_response: String = json_safe(clean_response)
 
     let updated_hist: String = hist_append(stored_hist, "user", message)
     let updated_hist2: String = hist_append(updated_hist, "assistant", raw_response)
@@ -246,7 +282,8 @@ fn handle_chat_as_soul(body: String) -> String {
         return "{\"error\":\"llm unavailable\",\"response\":\"\",\"speaker_slug\":\"" + speaker + "\",\"model\":\"" + model + "\"}"
     }
 
-    let safe_response: String = json_safe(raw_response)
+    let clean_response: String = clean_llm_response(raw_response)
+    let safe_response: String = json_safe(clean_response)
     return "{\"response\":\"" + safe_response + "\",\"model\":\"" + model + "\",\"speaker_slug\":\"" + speaker + "\"}"
 }