feat(compiler): backfill type+description on long-doc summaries during recompile

Author: KylinMountain

openkb/agent/compiler.py

@@ -2052,6 +2052,18 @@ async def compile_long_doc(
     schema_md = get_agents_md(wiki_dir)
     summary_content = summary_path.read_text(encoding="utf-8")
 
+    # Backfill OKF fields on the indexer-written summary (idempotent).
+    if summary_content.startswith("---"):
+        fm_end = summary_content.find("---", 3)
+        if fm_end != -1:
+            fm = summary_content[:fm_end + 3]
+            body = summary_content[fm_end + 3:]
+            fm = _set_fm_line(fm, "type", "Summary")
+            if doc_description:
+                fm = _set_fm_line(fm, "description", doc_description)
+            summary_content = fm + body
+            summary_path.write_text(summary_content, encoding="utf-8")
+
     # Base context A. cache_control marker on the doc message creates a
     # cache breakpoint covering (system + doc) for every concept call.
     system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(

tests/test_recompile.py

@@ -18,12 +18,14 @@
 
 from __future__ import annotations
 
+import asyncio
 import json
 from pathlib import Path
 from unittest.mock import AsyncMock, patch
 
 from click.testing import CliRunner
 
+from openkb.agent import compiler
 from openkb.cli import cli
 from openkb.schema import AGENTS_MD
 
@@ -312,3 +314,32 @@ def test_recompile_refresh_schema_noop_when_agents_missing(kb_dir):
     assert result.exit_code == 0, result.output
     assert not agents.exists()  # not materialized
     assert not (kb_dir / "wiki" / "AGENTS.md.bak").exists()
+
+
+# ---------------------------------------------------------------------------
+# compile_long_doc backfills type + description on recompile
+# ---------------------------------------------------------------------------
+
+
+def test_compile_long_doc_backfills_summary_frontmatter(tmp_path):
+    wiki = tmp_path / "wiki"
+    (wiki / "summaries").mkdir(parents=True)
+    (wiki / "concepts").mkdir(parents=True)
+    (tmp_path / ".openkb").mkdir()
+    (tmp_path / ".openkb" / "config.yaml").write_text(
+        "model: gpt-4o-mini\nlanguage: en\n", encoding="utf-8")
+    summary_path = wiki / "summaries" / "long.md"
+    summary_path.write_text(
+        "---\ndoc_type: pageindex\nfull_text: sources/long.json\n---\n\n# Long\n",
+        encoding="utf-8",
+    )
+    with patch.object(compiler, "_llm_call", return_value="overview"), \
+         patch.object(compiler, "_compile_concepts", new=AsyncMock()), \
+         patch.object(compiler, "_close_async_llm_clients", new=AsyncMock()):
+        asyncio.run(compiler.compile_long_doc(
+            "long", summary_path, "doc-1", tmp_path, "gpt-4o-mini",
+            doc_description="A long report.",
+        ))
+    text = summary_path.read_text(encoding="utf-8")
+    assert 'type: "Summary"' in text
+    assert 'description: "A long report."' in text