From 5c9590ada73789c16619be2222f6c85221108211 Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Wed, 8 Apr 2026 21:14:54 -0400
Subject: [PATCH] Custom Deserialize for NodeLeaf: recompute tokens on
 deserialization

token_ids are not serialized (serde skip), so deserialized nodes had
0 tokens. The custom Deserialize impl recomputes tokens from the body
text, restoring the invariant at the reconstruction boundary. No
separate recompute step needed.

Co-Authored-By: Proof of Concept <poc@bcachefs.org>
---
 src/agent/context.rs | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

diff --git a/src/agent/context.rs b/src/agent/context.rs
index e550885..79ca030 100644
--- a/src/agent/context.rs
+++ b/src/agent/context.rs
@@ -63,7 +63,8 @@ pub enum NodeBody {
 }
 
 /// A leaf node: typed content with cached token IDs.
-#[derive(Debug, Clone, Serialize, Deserialize)]
+/// Token IDs are not serialized — they're recomputed on deserialization.
+#[derive(Debug, Clone, Serialize)]
 pub struct NodeLeaf {
     body: NodeBody,
     #[serde(skip)]
@@ -71,6 +72,23 @@ pub struct NodeLeaf {
     timestamp: Option<DateTime<Utc>>,
 }
 
+impl<'de> Deserialize<'de> for NodeLeaf {
+    fn deserialize<D: serde::Deserializer<'de>>(deserializer: D) -> Result<Self, D::Error> {
+        #[derive(Deserialize)]
+        struct Raw {
+            body: NodeBody,
+            timestamp: Option<DateTime<Utc>>,
+        }
+        let raw = Raw::deserialize(deserializer)?;
+        let token_ids = if raw.body.is_prompt_visible() {
+            tokenizer::encode(&raw.body.render())
+        } else {
+            vec![]
+        };
+        Ok(NodeLeaf { body: raw.body, token_ids, timestamp: raw.timestamp })
+    }
+}
+
 /// A node in the context AST.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub enum AstNode {