From 80bdaab8ee03fcf4deae026abc9afd28283313c2 Mon Sep 17 00:00:00 2001
From: ProofOfConcept <poc@bcachefs.org>
Date: Fri, 6 Mar 2026 21:54:19 -0500
Subject: [PATCH] enrich: explicitly filter for text blocks in transcript
 extraction

Only extract content blocks with "type": "text". Previously relied on
tool_use/tool_result blocks lacking a "text" field, which worked but
was fragile. Now explicitly checks block type.
---
 src/enrich.rs | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/src/enrich.rs b/src/enrich.rs
index 34fd643..f024129 100644
--- a/src/enrich.rs
+++ b/src/enrich.rs
@@ -97,8 +97,10 @@ fn extract_conversation(jsonl_path: &str) -> Result<Vec<(usize, String, String,
             Some(serde_json::Value::Array(arr)) => {
                 arr.iter()
                     .filter_map(|c| {
-                        if let Some(t) = c.get("text").and_then(|v| v.as_str()) {
-                            Some(t.to_string())
+                        // Only extract text blocks; skip tool_use, tool_result, thinking, etc.
+                        let is_text = c.get("type").and_then(|v| v.as_str()) == Some("text");
+                        if is_text {
+                            c.get("text").and_then(|v| v.as_str()).map(|s| s.to_string())
                         } else {
                             c.as_str().map(|s| s.to_string())
                         }