~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.
Python academiaanthropicarxivbravedeep-researchencryptionhome-automationhomeserverlocallocal-deep-researchlocal-llmmistralollamaopenaipubmedresearchresearch-toolretrieval-augmented-generationsearxngself-hosted
fix(security): validate file size and count in RAG upload endpoint (#3938)
LearningCircuit89613f9f4016
diff --git a/changelog.d/+rag-upload-validator.security.md b/changelog.d/+rag-upload-validator.security.mdnew file mode 100644index 0000000000..aa1ee14828--- /dev/null+++ b/changelog.d/+rag-upload-validator.security.md@@ -0,0 +1 @@+RAG collection upload endpoint now validates per-file size (50MB limit) and per-request file count (200 limit), matching the research upload endpoint. Previously only the request-level `MAX_CONTENT_LENGTH` (10GB) was enforced, allowing a single oversized file or a request with thousands of zero-byte files to reach the per-file processing loop. Pre-flight `Content-Length` check rejects oversized files before reading bytes into memory.diff --git a/src/local_deep_research/research_library/routes/rag_routes.py b/src/local_deep_research/research_library/routes/rag_routes.pyindex 9a5ebeb3f0..4ff6f0f569 100644--- a/src/local_deep_research/research_library/routes/rag_routes.py+++ b/src/local_deep_research/research_library/routes/rag_routes.py@@ -39,6 +39,7 @@ from ...web.utils.request_helpers import parse_bool_argfrom ...utilities.db_utils import get_settings_managerfrom ..services.library_rag_service import LibraryRAGServicefrom ...settings.manager import SettingsManager+from ...security.file_upload_validator import FileUploadValidatorfrom ...security.path_validator import PathValidatorfrom ...security.rate_limiter import (upload_rate_limit_ip,@@ -1340,6 +1341,16 @@ def upload_to_collection(collection_id):{"success": False, "error": "No files selected"}), 400+ # Bound the per-request file count BEFORE doing any work. The+ # request-level MAX_CONTENT_LENGTH gate covers total bytes, but+ # not file *count*; a request with 10000 zero-byte files would+ # otherwise reach the loop below.+ is_valid, error_msg = FileUploadValidator.validate_file_count(+ len(files)+ )+ if not is_valid:+ return jsonify({"success": False, "error": error_msg}), 400+username = session["username"]with get_user_db_session(username) as db_session:# Verify collection exists in this user's database@@ -1401,10 +1412,39 @@ def upload_to_collection(collection_id):continuetry:+ # Pre-flight size check on Content-Length BEFORE reading+ # bytes into memory. Cheap rejection for oversized files;+ # avoids loading 50MB+ into memory just to discard it.+ is_valid, error_msg = (+ FileUploadValidator.validate_file_size(+ content_length=file.content_length,+ file_content=None,+ )+ )+ if not is_valid:+ errors.append(+ {"filename": filename, "error": error_msg}+ )+ continue+# Read file contentfile_content = file.read()file.seek(0) # Reset for potential re-reading+ # Post-read size check (Content-Length can be missing or+ # spoofed; the actual byte count is authoritative).+ is_valid, error_msg = (+ FileUploadValidator.validate_file_size(+ content_length=None,+ file_content=file_content,+ )+ )+ if not is_valid:+ errors.append(+ {"filename": filename, "error": error_msg}+ )+ continue+# Calculate file hash for deduplicationfile_hash = hashlib.sha256(file_content).hexdigest()diff --git a/tests/research_library/routes/test_rag_routes_upload_coverage.py b/tests/research_library/routes/test_rag_routes_upload_coverage.pyindex 583a973673..0e9ed69017 100644--- a/tests/research_library/routes/test_rag_routes_upload_coverage.py+++ b/tests/research_library/routes/test_rag_routes_upload_coverage.py@@ -207,6 +207,75 @@ class TestUploadToCollection:assert data["success"] is Falseassert "No files provided" in data["error"]+ def test_upload_rejects_too_many_files(self, app):+ """File count over MAX_FILES_PER_REQUEST is rejected with 400."""+ from local_deep_research.security.file_upload_validator import (+ FileUploadValidator,+ )++ # Patch the limit low so the test doesn't have to ship 201 files.+ with patch.object(FileUploadValidator, "MAX_FILES_PER_REQUEST", 3):+ with _auth_client(app) as (client, ctx):+ # Send 4 files (one over the patched limit).+ resp = client.post(+ "/library/api/collections/coll-1/upload",+ data={+ "files": [+ (BytesIO(b"a"), f"f{i}.txt") for i in range(4)+ ]+ },+ content_type="multipart/form-data",+ )+ assert resp.status_code == 400+ data = resp.get_json()+ assert data["success"] is False+ assert "Too many files" in data["error"]++ def test_upload_rejects_oversized_file(self, app):+ """Per-file size over MAX_FILE_SIZE is rejected; other files still process."""+ from local_deep_research.security.file_upload_validator import (+ FileUploadValidator,+ )++ mock_coll = Mock()+ mock_coll.id = "coll-1"++ db_session = _make_db_session()+ q = _build_mock_query(first_result=mock_coll)+ db_session.query = Mock(return_value=q)++ mock_password_store = Mock()+ mock_password_store.get_session_password.return_value = None++ # 100-byte limit; send a 200-byte file → rejected.+ with patch.object(FileUploadValidator, "MAX_FILE_SIZE", 100):+ with _auth_client(+ app,+ mock_db_session=db_session,+ extra_patches=[+ patch(+ f"{_DB_PASS}.session_password_store",+ mock_password_store,+ ),+ ],+ ) as (client, ctx):+ resp = client.post(+ "/library/api/collections/coll-1/upload",+ data={+ "files": (BytesIO(b"x" * 200), "big.txt"),+ },+ content_type="multipart/form-data",+ )++ assert resp.status_code == 200+ data = resp.get_json()+ # Oversized file appears in the per-file errors list, NOT 400 — so a+ # batch with one oversized + others can still succeed for the rest.+ assert data["success"] is True+ assert any(+ "File too large" in e.get("error", "") for e in data["errors"]+ )+def test_upload_empty_files_list(self, app):"""File with empty filename is silently skipped; response is success with 0 uploaded."""mock_coll = Mock()