Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

fix(release): preserve complete oversized notes (#6927)

LearningCircuitf30213eebb7d
diff --git a/.github/scripts/shorten_release_notes.py b/.github/scripts/shorten_release_notes.py
new file mode 100644
index 0000000000..affa1fac2b
--- /dev/null
+++ b/.github/scripts/shorten_release_notes.py
@@ -0,0 +1,79 @@
+"""Keep oversized GitHub Release bodies readable and lossless.
+
+The complete composed body is uploaded as a Release asset when GitHub's body
+limit would otherwise cut the hand-written changelog or generated PR list.
+"""
+
+from __future__ import annotations
+
+import argparse
+from pathlib import Path
+from urllib.parse import quote
+
+
+MAX_BODY_CHARS = 124_400 # Leave room below GitHub's 125,000-char limit.
+
+
+def shorten_release_notes(
+ body_file: Path,
+ full_asset_file: Path,
+ release_tag: str,
+ repository: str,
+) -> bool:
+ """Shorten an oversized body and save its exact original as an asset.
+
+ Returns whether an asset was written. A body within the limit is untouched.
+ """
+ body = body_file.read_text(encoding="utf-8")
+ if len(body) <= MAX_BODY_CHARS:
+ return False
+ if body_file.resolve() == full_asset_file.resolve():
+ raise ValueError("The full-notes asset must differ from the body file")
+
+ asset_url = (
+ f"https://github.com/{repository}/releases/download/"
+ f"{quote(release_tag, safe='')}/{quote(full_asset_file.name, safe='')}"
+ )
+ header = f"**Complete release notes:** [Download the full text]({asset_url}).\n\n"
+ marker = (
+ "\n\n_Release body shortened to fit GitHub's size limit. "
+ f"[Download all notes and the PR list]({asset_url})._\n"
+ )
+ excerpt_limit = MAX_BODY_CHARS - len(header) - len(marker)
+ if excerpt_limit <= 0:
+ raise ValueError("Release-note asset link exceeds the body budget")
+
+ # Prefer a paragraph boundary so the published body does not end midway
+ # through a changelog item. An unusually long single paragraph still has
+ # to be cut, but its complete form remains in the attached asset.
+ boundary = body.rfind("\n\n", 0, excerpt_limit)
+ if boundary < excerpt_limit // 2:
+ boundary = excerpt_limit
+ excerpt = body[:boundary].rstrip()
+ shortened = header + excerpt + marker
+ if len(shortened) > MAX_BODY_CHARS:
+ raise ValueError("Shortened release body still exceeds its limit")
+
+ full_asset_file.write_text(body, encoding="utf-8")
+ body_file.write_text(shortened, encoding="utf-8")
+ return True
+
+
+def main() -> None:
+ parser = argparse.ArgumentParser(description=__doc__)
+ parser.add_argument("body_file", type=Path)
+ parser.add_argument("full_asset_file", type=Path)
+ parser.add_argument("release_tag")
+ parser.add_argument("repository")
+ args = parser.parse_args()
+ if shorten_release_notes(
+ args.body_file,
+ args.full_asset_file,
+ args.release_tag,
+ args.repository,
+ ):
+ print(f"Attached complete release text as {args.full_asset_file.name}")
+
+
+if __name__ == "__main__":
+ main()
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
index 152de21b64..301aed9799 100644
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -701,6 +701,7 @@ jobs:
docs/release_notes/
changelog.d/
pyproject.toml
+ .github/scripts/shorten_release_notes.py
sparse-checkout-cone-mode: false
fetch-depth: 1
persist-credentials: false
@@ -1298,22 +1299,19 @@ jobs:
printf '%s\n' "$AUTO_BODY"
} > "$NOTES_FILE"
- # GitHub release body limit is 125,000 chars — the API rejects
- # longer bodies with HTTP 422 and gh CLI does not pre-validate.
- # Truncate from the tail (the auto PR list is the largest variable
- # section in big releases) with a visible marker. Character-aware
- # truncation (Python) — see diff section above.
- python3 - "$NOTES_FILE" 124400 "[notes truncated to fit GitHub's 125,000-char body limit]" <<'PYEOF'
- import sys
- path, max_chars, marker = sys.argv[1], int(sys.argv[2]), sys.argv[3]
- with open(path, encoding='utf-8', errors='replace') as f:
- content = f.read()
- if len(content) > max_chars:
- content = content[:max_chars] + f"\n\n_{marker}_\n"
- with open(path, 'w', encoding='utf-8') as f:
- f.write(content)
- print(f"Release body truncated to {len(content)} chars", file=sys.stderr)
- PYEOF
+ # The v2.0.0 hand-written changelog alone exceeds GitHub's
+ # 125,000-character body limit. Preserve the exact composed body
+ # (including the generated PR list) as a release asset and put a
+ # prominent link in the shortened body. Smaller releases are
+ # unchanged and do not receive the extra asset.
+ FULL_NOTES_ASSET="release-notes-full.md"
+ python3 .github/scripts/shorten_release_notes.py \
+ "$NOTES_FILE" "$FULL_NOTES_ASSET" \
+ "$RELEASE_TAG" "$GITHUB_REPOSITORY"
+ FULL_NOTES_ASSETS=()
+ if [[ -f "$FULL_NOTES_ASSET" ]]; then
+ FULL_NOTES_ASSETS+=("$FULL_NOTES_ASSET")
+ fi
# Includes filesystem SBOM (Trivy scan of source tree) AND
# container-image SBOM (Syft scan of built image layers, per-arch),
@@ -1345,6 +1343,7 @@ jobs:
sbom-cyclonedx.json \
sbom-cyclonedx.json.bundle \
provenance.intoto.jsonl \
+ "${FULL_NOTES_ASSETS[@]}" \
"${CONTAINER_SBOMS[@]}"
env:
# PAT_TOKEN is required here because GITHUB_TOKEN cannot trigger
diff --git a/docs/RELEASE_GUIDE.md b/docs/RELEASE_GUIDE.md
index 1be801446d..31e8f6c6ee 100644
--- a/docs/RELEASE_GUIDE.md
+++ b/docs/RELEASE_GUIDE.md
@@ -39,6 +39,10 @@ and short-circuits everything downstream).
merging the bump.
3. **Auto-generated PR list** from GitHub's generate-notes API,
label-categorized.
+- **Large release notes**: GitHub accepts at most 125,000 characters in a
+ release body. If the composed notes exceed that limit, the workflow attaches
+ the exact complete text as `release-notes-full.md` and links it prominently
+ from the shortened body. Smaller release bodies are left unchanged.
- **No duplicates**: If a release for `v<version>` already exists, the
`version-check` job sets `should_release=false` and every downstream
job (security gate, CI gate, build, publish) is skipped.
diff --git a/tests/ci/test_shorten_release_notes.py b/tests/ci/test_shorten_release_notes.py
new file mode 100644
index 0000000000..66f8ef47c5
--- /dev/null
+++ b/tests/ci/test_shorten_release_notes.py
@@ -0,0 +1,68 @@
+"""Verify that an oversized release retains its complete published text."""
+
+import subprocess
+import sys
+from pathlib import Path
+
+
+SCRIPT = (
+ Path(__file__).resolve().parents[2]
+ / ".github/scripts/shorten_release_notes.py"
+)
+ASSET_URL = (
+ "https://github.com/LearningCircuit/local-deep-research/"
+ "releases/download/v2.0.0/release-notes-full.md"
+)
+
+
+def _prepare(body: Path, asset: Path) -> None:
+ subprocess.run(
+ [
+ sys.executable,
+ str(SCRIPT),
+ str(body),
+ str(asset),
+ "v2.0.0",
+ "LearningCircuit/local-deep-research",
+ ],
+ check=True,
+ capture_output=True,
+ text=True,
+ )
+
+
+def test_small_release_body_is_unchanged(tmp_path):
+ body = tmp_path / "body.md"
+ asset = tmp_path / "release-notes-full.md"
+ content = "# Breaking Changes\n\n- A concise migration note.\n"
+ body.write_text(content, encoding="utf-8")
+
+ _prepare(body, asset)
+
+ assert body.read_text(encoding="utf-8") == content
+ assert not asset.exists()
+
+
+def test_oversized_release_links_exact_complete_body(tmp_path):
+ body = tmp_path / "body.md"
+ asset = tmp_path / "release-notes-full.md"
+ content = (
+ "# 💥 Breaking Changes\n\n"
+ + "- A complete migration paragraph with accented text: café.\n\n"
+ * 3_000
+ + "# What's Changed\n\n- The final PR still appears in the asset.\n"
+ )
+ body.write_text(content, encoding="utf-8")
+
+ _prepare(body, asset)
+
+ published = body.read_text(encoding="utf-8")
+ assert asset.read_text(encoding="utf-8") == content
+ assert len(published) <= 124_400
+ assert published.startswith(
+ f"**Complete release notes:** [Download the full text]({ASSET_URL})."
+ )
+ assert "[Download all notes and the PR list]" in published
+ assert published.count(ASSET_URL) == 2
+ assert "The final PR still appears in the asset" not in published
+ assert published.endswith(f"({ASSET_URL})._\n")