Update API format and support multiple string/pre-extracted inputs

src/engram/__init__.py

@@ -1,15 +1,16 @@
 from ._models import (
     CommittedOperation,
     CommittedOperations,
-    ConversationContent,
+    ConversationInput,
     Memory,
-    MessageContent,
-    PreExtractedContent,
+    MessageInput,
+    PreExtractedInput,
+    PreExtractedItem,
     RetrievalConfig,
     Run,
     RunStatus,
     SearchResults,
-    StringContent,
+    StringInput,
     ToolCallCustomInput,
     ToolCallFuncInput,
     ToolCallInput,
@@ -33,18 +34,19 @@
     "CommittedOperation",
     "CommittedOperations",
     "ConnectionError",
-    "ConversationContent",
+    "ConversationInput",
     "EngramClient",
     "EngramError",
     "EngramTimeoutError",
     "Memory",
-    "MessageContent",
-    "PreExtractedContent",
+    "MessageInput",
+    "PreExtractedInput",
+    "PreExtractedItem",
     "RetrievalConfig",
     "Run",
     "RunStatus",
     "SearchResults",
-    "StringContent",
+    "StringInput",
     "ToolCallCustomInput",
     "ToolCallFuncInput",
     "ToolCallInput",

src/engram/_models/__init__.py

@@ -1,31 +1,33 @@
 from .memory import (
-    AddContent,
-    ConversationContent,
+    AddInput,
+    ConversationInput,
     Memory,
-    MessageContent,
-    PreExtractedContent,
+    MessageInput,
+    PreExtractedInput,
+    PreExtractedItem,
     RetrievalConfig,
     SearchResults,
-    StringContent,
+    StringInput,
     ToolCallCustomInput,
     ToolCallFuncInput,
     ToolCallInput,
 )
 from .run import CommittedOperation, CommittedOperations, Run, RunStatus
 
 __all__ = [
-    "AddContent",
+    "AddInput",
     "CommittedOperation",
     "CommittedOperations",
-    "ConversationContent",
+    "ConversationInput",
     "Memory",
-    "MessageContent",
-    "PreExtractedContent",
+    "MessageInput",
+    "PreExtractedInput",
+    "PreExtractedItem",
     "RetrievalConfig",
     "Run",
     "RunStatus",
     "SearchResults",
-    "StringContent",
+    "StringInput",
     "ToolCallCustomInput",
     "ToolCallFuncInput",
     "ToolCallInput",

src/engram/_models/memory.py

@@ -6,18 +6,27 @@
 
 
 @dataclass(slots=True)
-class PreExtractedContent:
-    """Pre-extracted content that bypasses the extraction pipeline."""
+class PreExtractedInput:
+    """Pre-extracted input that skips the extraction step continues through the pipeline as-is.
+    Each individual item represents a separate memory.
+    """
+
+    items: list[PreExtractedItem]
+
+
+@dataclass(slots=True)
+class PreExtractedItem:
+    """A single pre-extracted memory."""
 
     content: str
     topic: str
 
 
 @dataclass(slots=True)
-class StringContent:
-    """String content that bypasses the extraction pipeline."""
+class StringInput:
+    """String input to extract memories from."""
 
-    content: str
+    content: str | list[str]
 
 
 @dataclass(slots=True)
@@ -50,7 +59,7 @@ class ToolCallInput:
 
 
 @dataclass(slots=True)
-class MessageContent:
+class MessageInput:
     """A message in a conversation using the OpenAI Chat Completions format.
 
     - 'tool' role (tool results) is mapped to 'user' by the server.
@@ -66,18 +75,18 @@ class MessageContent:
 
 
 @dataclass(slots=True)
-class ConversationContent:
-    """Conversation content that bypasses the extraction pipeline."""
+class ConversationInput:
+    """Conversation input that bypasses the extraction pipeline."""
 
-    messages: list[MessageContent]
+    messages: list[MessageInput]
     metadata: dict[str, Any] | None = None
     created_at: str | None = None
     updated_at: str | None = None
 
 
-# Type alias for the content argument to memories.add()
-AddContent: TypeAlias = (
-    str | list[dict[str, str]] | PreExtractedContent | ConversationContent | StringContent
+# Type alias for the input_data argument to memories.add()
+AddInput: TypeAlias = (
+    str | list[dict[str, str]] | PreExtractedInput | ConversationInput | StringInput
 )
 
 

src/engram/_resources/memories.py

@@ -3,7 +3,7 @@
 from uuid import UUID
 
 from .._http import AsyncHttpTransport, HttpTransport
-from .._models import AddContent, Memory, RetrievalConfig, Run, SearchResults
+from .._models import AddInput, Memory, RetrievalConfig, Run, SearchResults
 from .._serialization import (
     build_add_body,
     build_memory_params,
@@ -29,14 +29,14 @@ def __init__(self, transport: HttpTransport) -> None:
 
     def add(
         self,
-        content: AddContent,
+        input_data: AddInput,
         *,
         user_id: str | None = None,
         conversation_id: str | None = None,
         group: str | None = None,
     ) -> Run:
         body = build_add_body(
-            content,
+            input_data,
             user_id=user_id,
             conversation_id=conversation_id,
             group=group,
@@ -101,14 +101,14 @@ def __init__(self, transport: AsyncHttpTransport) -> None:
 
     async def add(
         self,
-        content: AddContent,
+        input_data: AddInput,
         *,
         user_id: str | None = None,
         conversation_id: str | None = None,
         group: str | None = None,
     ) -> Run:
         body = build_add_body(
-            content,
+            input_data,
             user_id=user_id,
             conversation_id=conversation_id,
             group=group,

src/engram/_serialization/_builders.py

@@ -3,11 +3,11 @@
 from typing import Any
 
 from .._models import (
-    AddContent,
-    ConversationContent,
-    PreExtractedContent,
+    AddInput,
+    ConversationInput,
+    PreExtractedInput,
     RetrievalConfig,
-    StringContent,
+    StringInput,
     ToolCallInput,
 )
 
@@ -21,29 +21,28 @@ def _serialize_tool_call(tc: ToolCallInput) -> dict[str, Any]:
     return out
 
 
-def _serialize_content(content: AddContent) -> dict[str, Any]:
-    """Build the content envelope with the type discriminator."""
-    if isinstance(content, str):
-        return {"type": "string", "content": content}
-    if isinstance(content, StringContent):
-        return {"type": "string", "content": content.content}
-    if isinstance(content, PreExtractedContent):
+def _serialize_input(input_data: AddInput) -> dict[str, Any]:
+    """Build the input envelope with the type discriminator."""
+    if isinstance(input_data, str):
+        return {"string": {"content": [input_data]}}
+    if isinstance(input_data, StringInput):
+        if isinstance(input_data.content, list):
+            return {"string": {"content": input_data.content}}
+        else:
+            return {"string": {"content": [input_data.content]}}
+    if isinstance(input_data, PreExtractedInput):
+        items = [{"content": item.content, "topic": item.topic} for item in input_data.items]
+        return {"pre_extracted": {"items": items}}
+    if isinstance(input_data, list):
         return {
-            "type": "pre_extracted",
-            "content": content.content,
-            "topic": content.topic,
+            "conversation": {"messages": input_data},
         }
-    if isinstance(content, list):
-        return {
-            "type": "conversation",
-            "conversation": {"messages": content},
-        }
-    if isinstance(content, ConversationContent):
-        return _serialize_conversation_content(content)
-    raise TypeError(f"Unsupported content type: {type(content)}")  # pragma: no cover
+    if isinstance(input_data, ConversationInput):
+        return _serialize_conversation_content(input_data)
+    raise TypeError(f"Unsupported input type: {type(input_data)}")  # pragma: no cover
 
 
-def _serialize_conversation_content(content: ConversationContent) -> dict[str, Any]:
+def _serialize_conversation_content(content: ConversationInput) -> dict[str, Any]:
     messages = []
     for msg in content.messages:
         m: dict[str, Any] = {"role": msg.role, "content": msg.content}
@@ -63,17 +62,17 @@ def _serialize_conversation_content(content: ConversationContent) -> dict[str, A
         conversation["created_at"] = content.created_at
     if content.updated_at is not None:
         conversation["updated_at"] = content.updated_at
-    return {"type": "conversation", "conversation": conversation}
+    return {"conversation": conversation}
 
 
 def build_add_body(
-    content: AddContent,
+    input_data: AddInput,
     *,
     user_id: str | None,
     conversation_id: str | None,
     group: str | None,
 ) -> dict[str, Any]:
-    body: dict[str, Any] = {"content": _serialize_content(content)}
+    body: dict[str, Any] = {"input": _serialize_input(input_data)}
     if user_id is not None:
         body["user_id"] = user_id
     if conversation_id is not None: