feat(indexer): build edges extraction + CI diagnostics ingestion

- Add _build_edges_from_detection(): materializes CMake/Make/Bazel
  parse results into token_build_edges (target→source, target→dep,
  target→flag edges). Integrated into build_build_doc().
- Add scripts/ingest_ci_diagnostics.py: fetches GitHub Actions logs,
  parses GCC/Clang/MSVC/CMake/Ninja diagnostics, outputs structured
  JSONL for token_diagnostic_edges pipeline.
- Add scripts/cleanup_verified_intermediates.py: purges source/extract
  cache for repos with verified parquet output.

Author: apstenku123

scripts/cleanup_verified_intermediates.py

@@ -0,0 +1,92 @@
+#!/usr/bin/env python3
+"""Purge source/intermediate files for repos whose parquet output is verified.
+
+Streaming principle: download -> process -> verify parquet -> delete source+intermediate.
+Parquet (zstd-max compressed) is the ONLY persistent artifact.
+
+Usage:
+    python scripts/cleanup_verified_intermediates.py [--dry-run] [--extract-cache] [--source-cache]
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import shutil
+from pathlib import Path
+
+import pyarrow.parquet as pq
+
+OUTPUT_BASE = Path("outputs")
+EXTRACT_CACHE = OUTPUT_BASE / "extract_cache_case5_shared"
+SOURCE_CACHE = OUTPUT_BASE / "source_cache" / "code"
+CODE_PARQUET = OUTPUT_BASE / "reindexed_case5_v7_20260715_130725_code"
+DONE_MANIFESTS = [
+    OUTPUT_BASE / "conveyor_case5_v11_resume_20260719_120500" / "_done.json",
+    OUTPUT_BASE / "conveyor_case5_v10_resume_20260718_150406" / "_done.json",
+]
+
+
+def load_done_repos() -> set[str]:
+    done = set()
+    for mf in DONE_MANIFESTS:
+        if not mf.exists():
+            continue
+        data = json.loads(mf.read_text())
+        for key, val in data.get("done", {}).items():
+            if isinstance(val, dict) and val.get("source") == "code":
+                done.add(key.split("::")[0])
+    return done
+
+
+def verify_parquet_exists(repo: str) -> bool:
+    for length_dir in CODE_PARQUET.iterdir():
+        if not length_dir.is_dir():
+            continue
+        if (length_dir / f"{repo}.parquet").exists():
+            schema = pq.read_schema(length_dir / f"{repo}.parquet")
+            if "trained_token_count" in schema.names:
+                return True
+    return False
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--dry-run", action="store_true")
+    parser.add_argument("--extract-cache", action="store_true", help="Clean extract cache")
+    parser.add_argument("--source-cache", action="store_true", help="Clean source cache")
+    args = parser.parse_args()
+
+    done_repos = load_done_repos()
+    print(f"Repos with verified parquet: {len(done_repos)}")
+
+    freed = 0
+
+    if args.extract_cache and EXTRACT_CACHE.exists():
+        for entry in EXTRACT_CACHE.iterdir():
+            repo_name = entry.name
+            if repo_name in done_repos and verify_parquet_exists(repo_name):
+                size = sum(f.stat().st_size for f in entry.rglob("*") if f.is_file())
+                if args.dry_run:
+                    print(f"  [dry-run] would remove extract_cache/{repo_name} ({size/1e9:.2f} GB)")
+                else:
+                    shutil.rmtree(entry, ignore_errors=True)
+                    print(f"  removed extract_cache/{repo_name} ({size/1e9:.2f} GB)")
+                freed += size
+
+    if args.source_cache and SOURCE_CACHE.exists():
+        for entry in SOURCE_CACHE.iterdir():
+            repo_name = entry.name
+            if repo_name in done_repos and verify_parquet_exists(repo_name):
+                size = sum(f.stat().st_size for f in entry.rglob("*") if f.is_file())
+                if args.dry_run:
+                    print(f"  [dry-run] would remove source_cache/{repo_name} ({size/1e9:.2f} GB)")
+                else:
+                    shutil.rmtree(entry, ignore_errors=True)
+                    print(f"  removed source_cache/{repo_name} ({size/1e9:.2f} GB)")
+                freed += size
+
+    print(f"\nTotal freed: {freed/1e9:.2f} GB")
+
+
+if __name__ == "__main__":
+    main()