Merge branch 'hackathon-copilot-search' of github.com:Significant-Gravitas/AutoGPT into feat/backfill_block_and_docs

- Created pluggable ContentHandler architecture for different content types
- Implemented StoreAgentHandler, BlockHandler, and DocumentationHandler
- Added backfill support for all content types with explicit processing order (blocks → agents → docs)
- Updated scheduler to process all content types automatically
- Fixed pgvector type resolution by adding set_public_search_path parameter
- Added comprehensive integration tests
- Updated stats aggregation to cover all content types

autogpt_platform/backend/backend/api/features/store/content_handlers.py

@@ -0,0 +1,417 @@
+"""
+Content Type Handlers for Unified Embeddings
+
+Pluggable system for different content sources (store agents, blocks, docs).
+Each handler knows how to fetch and process its content type for embedding.
+"""
+
+import logging
+from abc import ABC, abstractmethod
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+from prisma.enums import ContentType
+
+from backend.data.db import query_raw_with_schema
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class ContentItem:
+    """Represents a piece of content to be embedded."""
+
+    content_id: str  # Unique identifier (DB ID or file path)
+    content_type: ContentType
+    searchable_text: str  # Combined text for embedding
+    metadata: dict[str, Any]  # Content-specific metadata
+    user_id: str | None = None  # For user-scoped content
+
+
+class ContentHandler(ABC):
+    """Base handler for fetching and processing content for embeddings."""
+
+    @property
+    @abstractmethod
+    def content_type(self) -> ContentType:
+        """The ContentType this handler manages."""
+        pass
+
+    @abstractmethod
+    async def get_missing_items(self, batch_size: int) -> list[ContentItem]:
+        """
+        Fetch items that don't have embeddings yet.
+
+        Args:
+            batch_size: Maximum number of items to return
+
+        Returns:
+            List of ContentItem objects ready for embedding
+        """
+        pass
+
+    @abstractmethod
+    async def get_stats(self) -> dict[str, int]:
+        """
+        Get statistics about embedding coverage.
+
+        Returns:
+            Dict with keys: total, with_embeddings, without_embeddings
+        """
+        pass
+
+
+class StoreAgentHandler(ContentHandler):
+    """Handler for marketplace store agent listings."""
+
+    @property
+    def content_type(self) -> ContentType:
+        return ContentType.STORE_AGENT
+
+    async def get_missing_items(self, batch_size: int) -> list[ContentItem]:
+        """Fetch approved store listings without embeddings."""
+        from backend.api.features.store.embeddings import build_searchable_text
+
+        missing = await query_raw_with_schema(
+            """
+            SELECT
+                slv.id,
+                slv.name,
+                slv.description,
+                slv."subHeading",
+                slv.categories
+            FROM {schema_prefix}"StoreListingVersion" slv
+            LEFT JOIN {schema_prefix}"UnifiedContentEmbedding" uce
+                ON slv.id = uce."contentId" AND uce."contentType" = 'STORE_AGENT'::{schema_prefix}"ContentType"
+            WHERE slv."submissionStatus" = 'APPROVED'
+            AND slv."isDeleted" = false
+            AND uce."contentId" IS NULL
+            LIMIT $1
+            """,
+            batch_size,
+        )
+
+        return [
+            ContentItem(
+                content_id=row["id"],
+                content_type=ContentType.STORE_AGENT,
+                searchable_text=build_searchable_text(
+                    name=row["name"],
+                    description=row["description"],
+                    sub_heading=row["subHeading"],
+                    categories=row["categories"] or [],
+                ),
+                metadata={
+                    "name": row["name"],
+                    "categories": row["categories"] or [],
+                },
+                user_id=None,  # Store agents are public
+            )
+            for row in missing
+        ]
+
+    async def get_stats(self) -> dict[str, int]:
+        """Get statistics about store agent embedding coverage."""
+        # Count approved versions
+        approved_result = await query_raw_with_schema(
+            """
+            SELECT COUNT(*) as count
+            FROM {schema_prefix}"StoreListingVersion"
+            WHERE "submissionStatus" = 'APPROVED'
+            AND "isDeleted" = false
+            """
+        )
+        total_approved = approved_result[0]["count"] if approved_result else 0
+
+        # Count versions with embeddings
+        embedded_result = await query_raw_with_schema(
+            """
+            SELECT COUNT(*) as count
+            FROM {schema_prefix}"StoreListingVersion" slv
+            JOIN {schema_prefix}"UnifiedContentEmbedding" uce ON slv.id = uce."contentId" AND uce."contentType" = 'STORE_AGENT'::{schema_prefix}"ContentType"
+            WHERE slv."submissionStatus" = 'APPROVED'
+            AND slv."isDeleted" = false
+            """
+        )
+        with_embeddings = embedded_result[0]["count"] if embedded_result else 0
+
+        return {
+            "total": total_approved,
+            "with_embeddings": with_embeddings,
+            "without_embeddings": total_approved - with_embeddings,
+        }
+
+
+class BlockHandler(ContentHandler):
+    """Handler for block definitions (Python classes)."""
+
+    @property
+    def content_type(self) -> ContentType:
+        return ContentType.BLOCK
+
+    async def get_missing_items(self, batch_size: int) -> list[ContentItem]:
+        """Fetch blocks without embeddings."""
+        from backend.data.block import get_blocks
+
+        # Get all available blocks
+        all_blocks = get_blocks()
+
+        # Check which ones have embeddings
+        if not all_blocks:
+            return []
+
+        block_ids = list(all_blocks.keys())
+
+        # Query for existing embeddings
+        placeholders = ",".join([f"${i+1}" for i in range(len(block_ids))])
+        existing_result = await query_raw_with_schema(
+            f"""
+            SELECT "contentId"
+            FROM {{schema_prefix}}"UnifiedContentEmbedding"
+            WHERE "contentType" = 'BLOCK'::{{schema_prefix}}"ContentType"
+            AND "contentId" = ANY(ARRAY[{placeholders}])
+            """,
+            *block_ids,
+        )
+
+        existing_ids = {row["contentId"] for row in existing_result}
+        missing_blocks = [
+            (block_id, block_cls)
+            for block_id, block_cls in all_blocks.items()
+            if block_id not in existing_ids
+        ]
+
+        # Convert to ContentItem
+        items = []
+        for block_id, block_cls in missing_blocks[:batch_size]:
+            try:
+                block_instance = block_cls()
+
+                # Build searchable text from block metadata
+                parts = []
+                if hasattr(block_instance, "name") and block_instance.name:
+                    parts.append(block_instance.name)
+                if (
+                    hasattr(block_instance, "description")
+                    and block_instance.description
+                ):
+                    parts.append(block_instance.description)
+                if hasattr(block_instance, "categories") and block_instance.categories:
+                    # Convert BlockCategory enum to strings
+                    parts.append(
+                        " ".join(str(cat.value) for cat in block_instance.categories)
+                    )
+
+                # Add input/output schema info
+                if hasattr(block_instance, "input_schema"):
+                    schema = block_instance.input_schema
+                    if hasattr(schema, "model_json_schema"):
+                        schema_dict = schema.model_json_schema()
+                        if "properties" in schema_dict:
+                            for prop_name, prop_info in schema_dict[
+                                "properties"
+                            ].items():
+                                if "description" in prop_info:
+                                    parts.append(
+                                        f"{prop_name}: {prop_info['description']}"
+                                    )
+
+                searchable_text = " ".join(parts)
+
+                items.append(
+                    ContentItem(
+                        content_id=block_id,
+                        content_type=ContentType.BLOCK,
+                        searchable_text=searchable_text,
+                        metadata={
+                            "name": getattr(block_instance, "name", ""),
+                            "categories": getattr(block_instance, "categories", []),
+                        },
+                        user_id=None,  # Blocks are public
+                    )
+                )
+            except Exception as e:
+                logger.warning(f"Failed to process block {block_id}: {e}")
+                continue
+
+        return items
+
+    async def get_stats(self) -> dict[str, int]:
+        """Get statistics about block embedding coverage."""
+        from backend.data.block import get_blocks
+
+        all_blocks = get_blocks()
+        total_blocks = len(all_blocks)
+
+        if total_blocks == 0:
+            return {"total": 0, "with_embeddings": 0, "without_embeddings": 0}
+
+        block_ids = list(all_blocks.keys())
+        placeholders = ",".join([f"${i+1}" for i in range(len(block_ids))])
+
+        embedded_result = await query_raw_with_schema(
+            f"""
+            SELECT COUNT(*) as count
+            FROM {{schema_prefix}}"UnifiedContentEmbedding"
+            WHERE "contentType" = 'BLOCK'::{{schema_prefix}}"ContentType"
+            AND "contentId" = ANY(ARRAY[{placeholders}])
+            """,
+            *block_ids,
+        )
+
+        with_embeddings = embedded_result[0]["count"] if embedded_result else 0
+
+        return {
+            "total": total_blocks,
+            "with_embeddings": with_embeddings,
+            "without_embeddings": total_blocks - with_embeddings,
+        }
+
+
+class DocumentationHandler(ContentHandler):
+    """Handler for documentation files (.md/.mdx)."""
+
+    @property
+    def content_type(self) -> ContentType:
+        return ContentType.DOCUMENTATION
+
+    def _get_docs_root(self) -> Path:
+        """Get the documentation root directory."""
+        # Assuming docs are in /docs relative to project root
+        backend_root = Path(__file__).parent.parent.parent.parent
+        docs_root = backend_root.parent.parent / "docs"
+        return docs_root
+
+    def _extract_title_and_content(self, file_path: Path) -> tuple[str, str]:
+        """Extract title and content from markdown file."""
+        try:
+            content = file_path.read_text(encoding="utf-8")
+
+            # Try to extract title from first # heading
+            lines = content.split("\n")
+            title = ""
+            body_lines = []
+
+            for line in lines:
+                if line.startswith("# ") and not title:
+                    title = line[2:].strip()
+                else:
+                    body_lines.append(line)
+
+            # If no title found, use filename
+            if not title:
+                title = file_path.stem.replace("-", " ").replace("_", " ").title()
+
+            body = "\n".join(body_lines)
+
+            return title, body
+        except Exception as e:
+            logger.warning(f"Failed to read {file_path}: {e}")
+            return file_path.stem, ""
+
+    async def get_missing_items(self, batch_size: int) -> list[ContentItem]:
+        """Fetch documentation files without embeddings."""
+        docs_root = self._get_docs_root()
+
+        if not docs_root.exists():
+            logger.warning(f"Documentation root not found: {docs_root}")
+            return []
+
+        # Find all .md and .mdx files
+        all_docs = list(docs_root.rglob("*.md")) + list(docs_root.rglob("*.mdx"))
+
+        # Get relative paths for content IDs
+        doc_paths = [str(doc.relative_to(docs_root)) for doc in all_docs]
+
+        if not doc_paths:
+            return []
+
+        # Check which ones have embeddings
+        placeholders = ",".join([f"${i+1}" for i in range(len(doc_paths))])
+        existing_result = await query_raw_with_schema(
+            f"""
+            SELECT "contentId"
+            FROM {{schema_prefix}}"UnifiedContentEmbedding"
+            WHERE "contentType" = 'DOCUMENTATION'::{{schema_prefix}}"ContentType"
+            AND "contentId" = ANY(ARRAY[{placeholders}])
+            """,
+            *doc_paths,
+        )
+
+        existing_ids = {row["contentId"] for row in existing_result}
+        missing_docs = [
+            (doc_path, doc_file)
+            for doc_path, doc_file in zip(doc_paths, all_docs)
+            if doc_path not in existing_ids
+        ]
+
+        # Convert to ContentItem
+        items = []
+        for doc_path, doc_file in missing_docs[:batch_size]:
+            try:
+                title, content = self._extract_title_and_content(doc_file)
+
+                # Build searchable text
+                searchable_text = f"{title} {content}"
+
+                items.append(
+                    ContentItem(
+                        content_id=doc_path,
+                        content_type=ContentType.DOCUMENTATION,
+                        searchable_text=searchable_text,
+                        metadata={
+                            "title": title,
+                            "path": doc_path,
+                        },
+                        user_id=None,  # Documentation is public
+                    )
+                )
+            except Exception as e:
+                logger.warning(f"Failed to process doc {doc_path}: {e}")
+                continue
+
+        return items
+
+    async def get_stats(self) -> dict[str, int]:
+        """Get statistics about documentation embedding coverage."""
+        docs_root = self._get_docs_root()
+
+        if not docs_root.exists():
+            return {"total": 0, "with_embeddings": 0, "without_embeddings": 0}
+
+        # Count all .md and .mdx files
+        all_docs = list(docs_root.rglob("*.md")) + list(docs_root.rglob("*.mdx"))
+        total_docs = len(all_docs)
+
+        if total_docs == 0:
+            return {"total": 0, "with_embeddings": 0, "without_embeddings": 0}
+
+        doc_paths = [str(doc.relative_to(docs_root)) for doc in all_docs]
+        placeholders = ",".join([f"${i+1}" for i in range(len(doc_paths))])
+
+        embedded_result = await query_raw_with_schema(
+            f"""
+            SELECT COUNT(*) as count
+            FROM {{schema_prefix}}"UnifiedContentEmbedding"
+            WHERE "contentType" = 'DOCUMENTATION'::{{schema_prefix}}"ContentType"
+            AND "contentId" = ANY(ARRAY[{placeholders}])
+            """,
+            *doc_paths,
+        )
+
+        with_embeddings = embedded_result[0]["count"] if embedded_result else 0
+
+        return {
+            "total": total_docs,
+            "with_embeddings": with_embeddings,
+            "without_embeddings": total_docs - with_embeddings,
+        }
+
+
+# Content handler registry
+CONTENT_HANDLERS: dict[ContentType, ContentHandler] = {
+    ContentType.STORE_AGENT: StoreAgentHandler(),
+    ContentType.BLOCK: BlockHandler(),
+    ContentType.DOCUMENTATION: DocumentationHandler(),
+}

autogpt_platform/backend/backend/api/features/store/content_handlers_integration_test.py

@@ -0,0 +1,214 @@
+"""
+Integration tests for content handlers using real DB.
+
+Run with: poetry run pytest backend/api/features/store/content_handlers_integration_test.py -xvs
+
+These tests use the real database but mock OpenAI calls.
+"""
+
+from unittest.mock import patch
+
+import pytest
+
+from backend.api.features.store.content_handlers import (
+    CONTENT_HANDLERS,
+    BlockHandler,
+    DocumentationHandler,
+    StoreAgentHandler,
+)
+from backend.api.features.store.embeddings import (
+    backfill_all_content_types,
+    ensure_content_embedding,
+    get_embedding_stats,
+)
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_store_agent_handler_real_db():
+    """Test StoreAgentHandler with real database queries."""
+    handler = StoreAgentHandler()
+
+    # Get stats from real DB
+    stats = await handler.get_stats()
+
+    # Stats should have correct structure
+    assert "total" in stats
+    assert "with_embeddings" in stats
+    assert "without_embeddings" in stats
+    assert stats["total"] >= 0
+    assert stats["with_embeddings"] >= 0
+    assert stats["without_embeddings"] >= 0
+
+    # Get missing items (max 1 to keep test fast)
+    items = await handler.get_missing_items(batch_size=1)
+
+    # Items should be list (may be empty if all have embeddings)
+    assert isinstance(items, list)
+
+    if items:
+        item = items[0]
+        assert item.content_id is not None
+        assert item.content_type.value == "STORE_AGENT"
+        assert item.searchable_text != ""
+        assert item.user_id is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_block_handler_real_db():
+    """Test BlockHandler with real database queries."""
+    handler = BlockHandler()
+
+    # Get stats from real DB
+    stats = await handler.get_stats()
+
+    # Stats should have correct structure
+    assert "total" in stats
+    assert "with_embeddings" in stats
+    assert "without_embeddings" in stats
+    assert stats["total"] >= 0  # Should have at least some blocks
+    assert stats["with_embeddings"] >= 0
+    assert stats["without_embeddings"] >= 0
+
+    # Get missing items (max 1 to keep test fast)
+    items = await handler.get_missing_items(batch_size=1)
+
+    # Items should be list
+    assert isinstance(items, list)
+
+    if items:
+        item = items[0]
+        assert item.content_id is not None  # Should be block UUID
+        assert item.content_type.value == "BLOCK"
+        assert item.searchable_text != ""
+        assert item.user_id is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_documentation_handler_real_fs():
+    """Test DocumentationHandler with real filesystem."""
+    handler = DocumentationHandler()
+
+    # Get stats from real filesystem
+    stats = await handler.get_stats()
+
+    # Stats should have correct structure
+    assert "total" in stats
+    assert "with_embeddings" in stats
+    assert "without_embeddings" in stats
+    assert stats["total"] >= 0
+    assert stats["with_embeddings"] >= 0
+    assert stats["without_embeddings"] >= 0
+
+    # Get missing items (max 1 to keep test fast)
+    items = await handler.get_missing_items(batch_size=1)
+
+    # Items should be list
+    assert isinstance(items, list)
+
+    if items:
+        item = items[0]
+        assert item.content_id is not None  # Should be relative path
+        assert item.content_type.value == "DOCUMENTATION"
+        assert item.searchable_text != ""
+        assert item.user_id is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_get_embedding_stats_all_types():
+    """Test get_embedding_stats aggregates all content types."""
+    stats = await get_embedding_stats()
+
+    # Should have structure with by_type and totals
+    assert "by_type" in stats
+    assert "totals" in stats
+
+    # Check each content type is present
+    by_type = stats["by_type"]
+    assert "STORE_AGENT" in by_type
+    assert "BLOCK" in by_type
+    assert "DOCUMENTATION" in by_type
+
+    # Check totals are aggregated
+    totals = stats["totals"]
+    assert totals["total"] >= 0
+    assert totals["with_embeddings"] >= 0
+    assert totals["without_embeddings"] >= 0
+    assert "coverage_percent" in totals
+
+
+@pytest.mark.asyncio(loop_scope="session")
+@patch("backend.api.features.store.embeddings.generate_embedding")
+async def test_ensure_content_embedding_blocks(mock_generate):
+    """Test creating embeddings for blocks (mocked OpenAI)."""
+    # Mock OpenAI to return fake embedding
+    mock_generate.return_value = [0.1] * 1536
+
+    # Get one block without embedding
+    handler = BlockHandler()
+    items = await handler.get_missing_items(batch_size=1)
+
+    if not items:
+        pytest.skip("No blocks without embeddings")
+
+    item = items[0]
+
+    # Try to create embedding (OpenAI mocked)
+    result = await ensure_content_embedding(
+        content_type=item.content_type,
+        content_id=item.content_id,
+        searchable_text=item.searchable_text,
+        metadata=item.metadata,
+        user_id=item.user_id,
+    )
+
+    # Should succeed with mocked OpenAI
+    assert result is True
+    mock_generate.assert_called_once()
+
+
+@pytest.mark.asyncio(loop_scope="session")
+@patch("backend.api.features.store.embeddings.generate_embedding")
+async def test_backfill_all_content_types_dry_run(mock_generate):
+    """Test backfill_all_content_types processes all handlers in order."""
+    # Mock OpenAI to return fake embedding
+    mock_generate.return_value = [0.1] * 1536
+
+    # Run backfill with batch_size=1 to process max 1 per type
+    result = await backfill_all_content_types(batch_size=1)
+
+    # Should have results for all content types
+    assert "by_type" in result
+    assert "totals" in result
+
+    by_type = result["by_type"]
+    assert "BLOCK" in by_type
+    assert "STORE_AGENT" in by_type
+    assert "DOCUMENTATION" in by_type
+
+    # Each type should have correct structure
+    for content_type, type_result in by_type.items():
+        assert "processed" in type_result
+        assert "success" in type_result
+        assert "failed" in type_result
+
+    # Totals should aggregate
+    totals = result["totals"]
+    assert totals["processed"] >= 0
+    assert totals["success"] >= 0
+    assert totals["failed"] >= 0
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_content_handler_registry():
+    """Test all handlers are registered in correct order."""
+    from prisma.enums import ContentType
+
+    # All three types should be registered
+    assert ContentType.STORE_AGENT in CONTENT_HANDLERS
+    assert ContentType.BLOCK in CONTENT_HANDLERS
+    assert ContentType.DOCUMENTATION in CONTENT_HANDLERS
+
+    # Check handler types
+    assert isinstance(CONTENT_HANDLERS[ContentType.STORE_AGENT], StoreAgentHandler)
+    assert isinstance(CONTENT_HANDLERS[ContentType.BLOCK], BlockHandler)
+    assert isinstance(CONTENT_HANDLERS[ContentType.DOCUMENTATION], DocumentationHandler)

autogpt_platform/backend/backend/api/features/store/content_handlers_test.py

@@ -0,0 +1,324 @@
+"""
+E2E tests for content handlers (blocks, store agents, documentation).
+
+Tests the full flow: discovering content → generating embeddings → storing.
+"""
+
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+from prisma.enums import ContentType
+
+from backend.api.features.store.content_handlers import (
+    CONTENT_HANDLERS,
+    BlockHandler,
+    DocumentationHandler,
+    StoreAgentHandler,
+)
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_store_agent_handler_get_missing_items(mocker):
+    """Test StoreAgentHandler fetches approved agents without embeddings."""
+    handler = StoreAgentHandler()
+
+    # Mock database query
+    mock_missing = [
+        {
+            "id": "agent-1",
+            "name": "Test Agent",
+            "description": "A test agent",
+            "subHeading": "Test heading",
+            "categories": ["AI", "Testing"],
+        }
+    ]
+
+    with patch(
+        "backend.api.features.store.content_handlers.query_raw_with_schema",
+        return_value=mock_missing,
+    ):
+        items = await handler.get_missing_items(batch_size=10)
+
+        assert len(items) == 1
+        assert items[0].content_id == "agent-1"
+        assert items[0].content_type == ContentType.STORE_AGENT
+        assert "Test Agent" in items[0].searchable_text
+        assert "A test agent" in items[0].searchable_text
+        assert items[0].metadata["name"] == "Test Agent"
+        assert items[0].user_id is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_store_agent_handler_get_stats(mocker):
+    """Test StoreAgentHandler returns correct stats."""
+    handler = StoreAgentHandler()
+
+    # Mock approved count query
+    mock_approved = [{"count": 50}]
+    # Mock embedded count query
+    mock_embedded = [{"count": 30}]
+
+    with patch(
+        "backend.api.features.store.content_handlers.query_raw_with_schema",
+        side_effect=[mock_approved, mock_embedded],
+    ):
+        stats = await handler.get_stats()
+
+        assert stats["total"] == 50
+        assert stats["with_embeddings"] == 30
+        assert stats["without_embeddings"] == 20
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_block_handler_get_missing_items(mocker):
+    """Test BlockHandler discovers blocks without embeddings."""
+    handler = BlockHandler()
+
+    # Mock get_blocks to return test blocks
+    mock_block_class = MagicMock()
+    mock_block_instance = MagicMock()
+    mock_block_instance.name = "Calculator Block"
+    mock_block_instance.description = "Performs calculations"
+    mock_block_instance.categories = [MagicMock(value="MATH")]
+    mock_block_instance.input_schema.model_json_schema.return_value = {
+        "properties": {"expression": {"description": "Math expression to evaluate"}}
+    }
+    mock_block_class.return_value = mock_block_instance
+
+    mock_blocks = {"block-uuid-1": mock_block_class}
+
+    # Mock existing embeddings query (no embeddings exist)
+    mock_existing = []
+
+    with patch(
+        "backend.api.features.store.content_handlers.get_blocks",
+        return_value=mock_blocks,
+    ):
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=mock_existing,
+        ):
+            items = await handler.get_missing_items(batch_size=10)
+
+            assert len(items) == 1
+            assert items[0].content_id == "block-uuid-1"
+            assert items[0].content_type == ContentType.BLOCK
+            assert "Calculator Block" in items[0].searchable_text
+            assert "Performs calculations" in items[0].searchable_text
+            assert "MATH" in items[0].searchable_text
+            assert "expression: Math expression" in items[0].searchable_text
+            assert items[0].user_id is None
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_block_handler_get_stats(mocker):
+    """Test BlockHandler returns correct stats."""
+    handler = BlockHandler()
+
+    # Mock get_blocks
+    mock_blocks = {
+        "block-1": MagicMock(),
+        "block-2": MagicMock(),
+        "block-3": MagicMock(),
+    }
+
+    # Mock embedded count query (2 blocks have embeddings)
+    mock_embedded = [{"count": 2}]
+
+    with patch(
+        "backend.api.features.store.content_handlers.get_blocks",
+        return_value=mock_blocks,
+    ):
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=mock_embedded,
+        ):
+            stats = await handler.get_stats()
+
+            assert stats["total"] == 3
+            assert stats["with_embeddings"] == 2
+            assert stats["without_embeddings"] == 1
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_documentation_handler_get_missing_items(tmp_path, mocker):
+    """Test DocumentationHandler discovers docs without embeddings."""
+    handler = DocumentationHandler()
+
+    # Create temporary docs directory with test files
+    docs_root = tmp_path / "docs"
+    docs_root.mkdir()
+
+    (docs_root / "guide.md").write_text("# Getting Started\n\nThis is a guide.")
+    (docs_root / "api.mdx").write_text("# API Reference\n\nAPI documentation.")
+
+    # Mock _get_docs_root to return temp dir
+    with patch.object(handler, "_get_docs_root", return_value=docs_root):
+        # Mock existing embeddings query (no embeddings exist)
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=[],
+        ):
+            items = await handler.get_missing_items(batch_size=10)
+
+            assert len(items) == 2
+
+            # Check guide.md
+            guide_item = next(
+                (item for item in items if item.content_id == "guide.md"), None
+            )
+            assert guide_item is not None
+            assert guide_item.content_type == ContentType.DOCUMENTATION
+            assert "Getting Started" in guide_item.searchable_text
+            assert "This is a guide" in guide_item.searchable_text
+            assert guide_item.metadata["title"] == "Getting Started"
+            assert guide_item.user_id is None
+
+            # Check api.mdx
+            api_item = next(
+                (item for item in items if item.content_id == "api.mdx"), None
+            )
+            assert api_item is not None
+            assert "API Reference" in api_item.searchable_text
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_documentation_handler_get_stats(tmp_path, mocker):
+    """Test DocumentationHandler returns correct stats."""
+    handler = DocumentationHandler()
+
+    # Create temporary docs directory
+    docs_root = tmp_path / "docs"
+    docs_root.mkdir()
+    (docs_root / "doc1.md").write_text("# Doc 1")
+    (docs_root / "doc2.md").write_text("# Doc 2")
+    (docs_root / "doc3.mdx").write_text("# Doc 3")
+
+    # Mock embedded count query (1 doc has embedding)
+    mock_embedded = [{"count": 1}]
+
+    with patch.object(handler, "_get_docs_root", return_value=docs_root):
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=mock_embedded,
+        ):
+            stats = await handler.get_stats()
+
+            assert stats["total"] == 3
+            assert stats["with_embeddings"] == 1
+            assert stats["without_embeddings"] == 2
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_documentation_handler_title_extraction(tmp_path):
+    """Test DocumentationHandler extracts title from markdown heading."""
+    handler = DocumentationHandler()
+
+    # Test with heading
+    doc_with_heading = tmp_path / "with_heading.md"
+    doc_with_heading.write_text("# My Title\n\nContent here")
+    title, content = handler._extract_title_and_content(doc_with_heading)
+    assert title == "My Title"
+    assert "# My Title" not in content
+    assert "Content here" in content
+
+    # Test without heading
+    doc_without_heading = tmp_path / "no-heading.md"
+    doc_without_heading.write_text("Just content, no heading")
+    title, content = handler._extract_title_and_content(doc_without_heading)
+    assert title == "No Heading"  # Uses filename
+    assert "Just content" in content
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_content_handlers_registry():
+    """Test all content types are registered."""
+    assert ContentType.STORE_AGENT in CONTENT_HANDLERS
+    assert ContentType.BLOCK in CONTENT_HANDLERS
+    assert ContentType.DOCUMENTATION in CONTENT_HANDLERS
+
+    assert isinstance(CONTENT_HANDLERS[ContentType.STORE_AGENT], StoreAgentHandler)
+    assert isinstance(CONTENT_HANDLERS[ContentType.BLOCK], BlockHandler)
+    assert isinstance(CONTENT_HANDLERS[ContentType.DOCUMENTATION], DocumentationHandler)
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_block_handler_handles_missing_attributes():
+    """Test BlockHandler gracefully handles blocks with missing attributes."""
+    handler = BlockHandler()
+
+    # Mock block with minimal attributes
+    mock_block_class = MagicMock()
+    mock_block_instance = MagicMock()
+    mock_block_instance.name = "Minimal Block"
+    # No description, categories, or schema
+    del mock_block_instance.description
+    del mock_block_instance.categories
+    del mock_block_instance.input_schema
+    mock_block_class.return_value = mock_block_instance
+
+    mock_blocks = {"block-minimal": mock_block_class}
+
+    with patch(
+        "backend.api.features.store.content_handlers.get_blocks",
+        return_value=mock_blocks,
+    ):
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=[],
+        ):
+            items = await handler.get_missing_items(batch_size=10)
+
+            assert len(items) == 1
+            assert items[0].searchable_text == "Minimal Block"
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_block_handler_skips_failed_blocks():
+    """Test BlockHandler skips blocks that fail to instantiate."""
+    handler = BlockHandler()
+
+    # Mock one good block and one bad block
+    good_block = MagicMock()
+    good_instance = MagicMock()
+    good_instance.name = "Good Block"
+    good_instance.description = "Works fine"
+    good_instance.categories = []
+    good_block.return_value = good_instance
+
+    bad_block = MagicMock()
+    bad_block.side_effect = Exception("Instantiation failed")
+
+    mock_blocks = {"good-block": good_block, "bad-block": bad_block}
+
+    with patch(
+        "backend.api.features.store.content_handlers.get_blocks",
+        return_value=mock_blocks,
+    ):
+        with patch(
+            "backend.api.features.store.content_handlers.query_raw_with_schema",
+            return_value=[],
+        ):
+            items = await handler.get_missing_items(batch_size=10)
+
+            # Should only get the good block
+            assert len(items) == 1
+            assert items[0].content_id == "good-block"
+
+
+@pytest.mark.asyncio(loop_scope="session")
+async def test_documentation_handler_missing_docs_directory():
+    """Test DocumentationHandler handles missing docs directory gracefully."""
+    handler = DocumentationHandler()
+
+    # Mock _get_docs_root to return non-existent path
+    fake_path = Path("/nonexistent/docs")
+    with patch.object(handler, "_get_docs_root", return_value=fake_path):
+        items = await handler.get_missing_items(batch_size=10)
+        assert items == []
+
+        stats = await handler.get_stats()
+        assert stats["total"] == 0
+        assert stats["with_embeddings"] == 0
+        assert stats["without_embeddings"] == 0

autogpt_platform/backend/backend/api/features/store/embeddings.py

@@ -14,6 +14,7 @@ import prisma
 from prisma.enums import ContentType
 from tiktoken import encoding_for_model
 
+from backend.api.features.store.content_handlers import CONTENT_HANDLERS
 from backend.data.db import execute_raw_with_schema, query_raw_with_schema
 from backend.util.clients import get_openai_client
 from backend.util.json import dumps
@@ -369,55 +370,69 @@ async def delete_content_embedding(
 
 async def get_embedding_stats() -> dict[str, Any]:
     """
-    Get statistics about embedding coverage.
+    Get statistics about embedding coverage for all content types.
 
-    Returns counts of:
-    - Total approved listing versions
-    - Versions with embeddings
-    - Versions without embeddings
+    Returns stats per content type and overall totals.
     """
     try:
-        # Count approved versions
-        approved_result = await query_raw_with_schema(
-            """
-            SELECT COUNT(*) as count
-            FROM {schema_prefix}"StoreListingVersion"
-            WHERE "submissionStatus" = 'APPROVED'
-            AND "isDeleted" = false
-            """
-        )
-        total_approved = approved_result[0]["count"] if approved_result else 0
+        stats_by_type = {}
+        total_items = 0
+        total_with_embeddings = 0
+        total_without_embeddings = 0
 
-        # Count versions with embeddings
-        embedded_result = await query_raw_with_schema(
-            """
-            SELECT COUNT(*) as count
-            FROM {schema_prefix}"StoreListingVersion" slv
-            JOIN {schema_prefix}"UnifiedContentEmbedding" uce ON slv.id = uce."contentId" AND uce."contentType" = 'STORE_AGENT'::{schema_prefix}"ContentType"
-            WHERE slv."submissionStatus" = 'APPROVED'
-            AND slv."isDeleted" = false
-            """
-        )
-        with_embeddings = embedded_result[0]["count"] if embedded_result else 0
+        # Aggregate stats from all handlers
+        for content_type, handler in CONTENT_HANDLERS.items():
+            try:
+                stats = await handler.get_stats()
+                stats_by_type[content_type.value] = {
+                    "total": stats["total"],
+                    "with_embeddings": stats["with_embeddings"],
+                    "without_embeddings": stats["without_embeddings"],
+                    "coverage_percent": (
+                        round(stats["with_embeddings"] / stats["total"] * 100, 1)
+                        if stats["total"] > 0
+                        else 0
+                    ),
+                }
+
+                total_items += stats["total"]
+                total_with_embeddings += stats["with_embeddings"]
+                total_without_embeddings += stats["without_embeddings"]
+
+            except Exception as e:
+                logger.error(f"Failed to get stats for {content_type.value}: {e}")
+                stats_by_type[content_type.value] = {
+                    "total": 0,
+                    "with_embeddings": 0,
+                    "without_embeddings": 0,
+                    "coverage_percent": 0,
+                    "error": str(e),
+                }
 
         return {
-            "total_approved": total_approved,
-            "with_embeddings": with_embeddings,
-            "without_embeddings": total_approved - with_embeddings,
-            "coverage_percent": (
-                round(with_embeddings / total_approved * 100, 1)
-                if total_approved > 0
-                else 0
-            ),
+            "by_type": stats_by_type,
+            "totals": {
+                "total": total_items,
+                "with_embeddings": total_with_embeddings,
+                "without_embeddings": total_without_embeddings,
+                "coverage_percent": (
+                    round(total_with_embeddings / total_items * 100, 1)
+                    if total_items > 0
+                    else 0
+                ),
+            },
         }
 
     except Exception as e:
         logger.error(f"Failed to get embedding stats: {e}")
         return {
-            "total_approved": 0,
-            "with_embeddings": 0,
-            "without_embeddings": 0,
-            "coverage_percent": 0,
+            "by_type": {},
+            "totals": {
+                "total": 0,
+                "with_embeddings": 0,
+                "without_embeddings": 0,
+                "coverage_percent": 0,
+            },
             "error": str(e),
         }
 
@@ -426,73 +441,118 @@ async def backfill_missing_embeddings(batch_size: int = 10) -> dict[str, Any]:
     """
     Generate embeddings for approved listings that don't have them.
 
+    BACKWARD COMPATIBILITY: Maintained for existing usage.
+    This now delegates to backfill_all_content_types() to process all content types.
+
     Args:
-        batch_size: Number of embeddings to generate in one call
+        batch_size: Number of embeddings to generate per content type
 
     Returns:
-        Dict with success/failure counts
+        Dict with success/failure counts aggregated across all content types
     """
-    try:
-        # Find approved versions without embeddings
-        missing = await query_raw_with_schema(
-            """
-            SELECT
-                slv.id,
-                slv.name,
-                slv.description,
-                slv."subHeading",
-                slv.categories
-            FROM {schema_prefix}"StoreListingVersion" slv
-            LEFT JOIN {schema_prefix}"UnifiedContentEmbedding" uce
-                ON slv.id = uce."contentId" AND uce."contentType" = 'STORE_AGENT'::{schema_prefix}"ContentType"
-            WHERE slv."submissionStatus" = 'APPROVED'
-            AND slv."isDeleted" = false
-            AND uce."contentId" IS NULL
-            LIMIT $1
-            """,
-            batch_size,
-        )
+    # Delegate to the new generic backfill system
+    result = await backfill_all_content_types(batch_size)
 
-        if not missing:
-            return {
+    # Return in the old format for backward compatibility
+    return result["totals"]
+
+
+async def backfill_all_content_types(batch_size: int = 10) -> dict[str, Any]:
+    """
+    Generate embeddings for all content types using registered handlers.
+
+    Processes content types in order: BLOCK → STORE_AGENT → DOCUMENTATION.
+    This ensures foundational content (blocks) are searchable first.
+
+    Args:
+        batch_size: Number of embeddings to generate per content type
+
+    Returns:
+        Dict with stats per content type and overall totals
+    """
+    results_by_type = {}
+    total_processed = 0
+    total_success = 0
+    total_failed = 0
+
+    # Process content types in explicit order
+    processing_order = [
+        ContentType.BLOCK,
+        ContentType.STORE_AGENT,
+        ContentType.DOCUMENTATION,
+    ]
+
+    for content_type in processing_order:
+        handler = CONTENT_HANDLERS.get(content_type)
+        if not handler:
+            logger.warning(f"No handler registered for {content_type.value}")
+            continue
+        try:
+            logger.info(f"Processing {content_type.value} content type...")
+
+            # Get missing items from handler
+            missing_items = await handler.get_missing_items(batch_size)
+
+            if not missing_items:
+                results_by_type[content_type.value] = {
+                    "processed": 0,
+                    "success": 0,
+                    "failed": 0,
+                    "message": "No missing embeddings",
+                }
+                continue
+
+            # Process embeddings concurrently for better performance
+            embedding_tasks = [
+                ensure_content_embedding(
+                    content_type=item.content_type,
+                    content_id=item.content_id,
+                    searchable_text=item.searchable_text,
+                    metadata=item.metadata,
+                    user_id=item.user_id,
+                )
+                for item in missing_items
+            ]
+
+            results = await asyncio.gather(*embedding_tasks, return_exceptions=True)
+
+            success = sum(1 for result in results if result is True)
+            failed = len(results) - success
+
+            results_by_type[content_type.value] = {
+                "processed": len(missing_items),
+                "success": success,
+                "failed": failed,
+                "message": f"Backfilled {success} embeddings, {failed} failed",
+            }
+
+            total_processed += len(missing_items)
+            total_success += success
+            total_failed += failed
+
+            logger.info(
+                f"{content_type.value}: processed {len(missing_items)}, "
+                f"success {success}, failed {failed}"
+            )
+
+        except Exception as e:
+            logger.error(f"Failed to process {content_type.value}: {e}")
+            results_by_type[content_type.value] = {
                 "processed": 0,
                 "success": 0,
                 "failed": 0,
-                "message": "No missing embeddings",
+                "error": str(e),
             }
 
-        # Process embeddings concurrently for better performance
-        embedding_tasks = [
-            ensure_embedding(
-                version_id=row["id"],
-                name=row["name"],
-                description=row["description"],
-                sub_heading=row["subHeading"],
-                categories=row["categories"] or [],
-            )
-            for row in missing
-        ]
-
-        results = await asyncio.gather(*embedding_tasks, return_exceptions=True)
-
-        success = sum(1 for result in results if result is True)
-        failed = len(results) - success
-
-        return {
-            "processed": len(missing),
-            "success": success,
-            "failed": failed,
-            "message": f"Backfilled {success} embeddings, {failed} failed",
-        }
-
-    except Exception as e:
-        logger.error(f"Failed to backfill embeddings: {e}")
-        return {
-            "processed": 0,
-            "success": 0,
-            "failed": 0,
-            "error": str(e),
-        }
+    return {
+        "by_type": results_by_type,
+        "totals": {
+            "processed": total_processed,
+            "success": total_success,
+            "failed": total_failed,
+            "message": f"Overall: {total_success} succeeded, {total_failed} failed",
+        },
+    }
 
 
 async def embed_query(query: str) -> list[float] | None:

autogpt_platform/backend/backend/executor/scheduler.py

@@ -257,14 +257,14 @@ def execution_accuracy_alerts():
 
 def ensure_embeddings_coverage():
     """
-    Ensure approved store agents have embeddings for hybrid search.
+    Ensure all content types (store agents, blocks, docs) have embeddings for search.
 
-    Processes ALL missing embeddings in batches of 10 until 100% coverage.
-    Missing embeddings = agents invisible in hybrid search.
+    Processes ALL missing embeddings in batches of 10 per content type until 100% coverage.
+    Missing embeddings = content invisible in hybrid search.
 
     Schedule: Runs every 6 hours (balanced between coverage and API costs).
-    - Catches agents approved between scheduled runs
-    - Batch size 10: gradual processing to avoid rate limits
+    - Catches new content added between scheduled runs
+    - Batch size 10 per content type: gradual processing to avoid rate limits
     - Manual trigger available via execute_ensure_embeddings_coverage endpoint
     """
     db_client = get_database_manager_client()
@@ -277,13 +277,27 @@ def ensure_embeddings_coverage():
         )
         return {"processed": 0, "success": 0, "failed": 0, "error": stats["error"]}
 
-    if stats["without_embeddings"] == 0:
-        logger.info("All approved agents have embeddings, skipping backfill")
+    # Extract totals from new stats structure
+    totals = stats.get("totals", {})
+    without_embeddings = totals.get("without_embeddings", 0)
+    coverage_percent = totals.get("coverage_percent", 0)
+
+    if without_embeddings == 0:
+        logger.info("All content has embeddings, skipping backfill")
         return {"processed": 0, "success": 0, "failed": 0}
 
+    # Log per-content-type stats for visibility
+    by_type = stats.get("by_type", {})
+    for content_type, type_stats in by_type.items():
+        if type_stats.get("without_embeddings", 0) > 0:
+            logger.info(
+                f"{content_type}: {type_stats['without_embeddings']} items without embeddings "
+                f"({type_stats['coverage_percent']}% coverage)"
+            )
+
     logger.info(
-        f"Found {stats['without_embeddings']} agents without embeddings "
-        f"({stats['coverage_percent']}% coverage) - processing all"
+        f"Total: {without_embeddings} items without embeddings "
+        f"({coverage_percent}% coverage) - processing all"
     )
 
     total_processed = 0