feat: Modify the Summary BaseModel to use public_id of message rather than internal ID (#224)

* feat: Modify the Summary BaseModel to use public_id of message rather than internal ID

* fix: fallback invocation

* fix: strict validation

* fix: rm ID from Message schema

* fix: make SDK changes

sdks/python/src/honcho/session_context.py

@@ -13,7 +13,7 @@ class Summary(BaseModel):
     """Represents a summary of a session's conversation."""
 
     content: str = Field(..., description="The summary text")
-    message_id: int = Field(
+    message_id: str = Field(
         ..., description="The ID of the message that this summary covers up to"
     )
     summary_type: str = Field(..., description="The type of summary (short or long)")

sdks/typescript/src/session_context.ts

@@ -3,7 +3,7 @@ import type { Peer } from './peer'
 
 export interface SummaryData {
   content: string
-  message_id: number
+  message_id: string
   summary_type: string
   created_at: string
   token_count: number
@@ -21,7 +21,7 @@ export class Summary {
   /**
    * The ID of the message that this summary covers up to.
    */
-  readonly messageId: number
+  readonly messageId: string
 
   /**
    * The type of summary (short or long).
@@ -138,9 +138,9 @@ export class SessionContext {
     const assistantId = typeof assistant === 'string' ? assistant : assistant.id
     const summaryMessage = this.summary
       ? {
-          role: 'system',
-          content: `<summary>${this.summary.content}</summary>`,
-        }
+        role: 'system',
+        content: `<summary>${this.summary.content}</summary>`,
+      }
       : null
     const messages = this.messages.map((message) => ({
       role: message.peer_id === assistantId ? 'assistant' : 'user',
@@ -172,20 +172,20 @@ export class SessionContext {
     const assistantId = typeof assistant === 'string' ? assistant : assistant.id
     const summaryMessage = this.summary
       ? {
-          role: 'user',
-          content: `<summary>${this.summary.content}</summary>`,
-        }
+        role: 'user',
+        content: `<summary>${this.summary.content}</summary>`,
+      }
       : null
     const messages = this.messages.map((message) =>
       message.peer_id === assistantId
         ? {
-            role: 'assistant',
-            content: message.content,
-          }
+          role: 'assistant',
+          content: message.content,
+        }
         : {
-            role: 'user',
-            content: `${message.peer_id}: ${message.content}`,
-          }
+          role: 'user',
+          content: `${message.peer_id}: ${message.content}`,
+        }
     )
     return summaryMessage ? [summaryMessage, ...messages] : messages
   }

src/deriver/consumer.py

@@ -5,7 +5,9 @@ import sentry_sdk
 from langfuse import get_client
 from pydantic import ValidationError
 from rich.console import Console
+from sqlalchemy import select
 
+from src import models
 from src.config import settings
 from src.dependencies import tracked_db
 from src.deriver.deriver import process_representation_tasks_batch
@@ -52,6 +54,30 @@ async def process_item(task_type: str, queue_payload: dict[str, Any]) -> None:
                 queue_payload,
             )
             raise ValueError(f"Invalid payload structure: {str(e)}") from e
+
+        message_public_id = validated.message_public_id
+        if not message_public_id:
+            logger.info(
+                "Fetching message public ID for message %s", validated.message_id
+            )
+            async with tracked_db(operation_name="summary_fallback") as db:
+                stmt = (
+                    select(models.Message)
+                    .where(models.Message.workspace_name == validated.workspace_name)
+                    .where(models.Message.session_name == validated.session_name)
+                    .where(models.Message.id == validated.message_id)
+                )
+                result = await db.execute(stmt)
+
+                message = result.scalar_one_or_none()
+                if message is None:
+                    logger.error(
+                        "Failed to fetch message with ID %s for process_summary_task",
+                        validated.message_id,
+                    )
+                    return
+                message_public_id = message.public_id
+
         with sentry_sdk.start_transaction(name="process_summary_task", op="deriver"):
             if settings.LANGFUSE_PUBLIC_KEY:
                 with lf.start_as_current_span(
@@ -70,6 +96,7 @@ async def process_item(task_type: str, queue_payload: dict[str, Any]) -> None:
                         validated.session_name,
                         validated.message_id,
                         validated.message_seq_in_session,
+                        message_public_id,
                     )
                     log_performance_metrics(
                         f"summary_{validated.workspace_name}_{validated.message_id}"
@@ -80,6 +107,7 @@ async def process_item(task_type: str, queue_payload: dict[str, Any]) -> None:
                     validated.session_name,
                     validated.message_id,
                     validated.message_seq_in_session,
+                    message_public_id,
                 )
                 log_performance_metrics(
                     f"summary_{validated.workspace_name}_{validated.message_id}"

src/deriver/queue_payload.py

@@ -37,6 +37,8 @@ class SummaryPayload(BasePayload):
     session_name: str
     message_id: int
     message_seq_in_session: int
+    # Optional for backward compatibility with older queue items
+    message_public_id: str | None = None
 
 
 class WebhookPayload(BasePayload):
@@ -124,12 +126,20 @@ def create_payload(
         elif task_type == "summary":
             if message_seq_in_session is None:
                 raise ValueError("message_seq_in_session is required for summary tasks")
+            message_public_id = message.get("message_public_id")
+            if message_public_id is not None and (
+                not isinstance(message_public_id, str) or not message_public_id.strip()
+            ):
+                raise ValueError(
+                    "message_public_id must be a non-empty string if provided"
+                )
 
             validated_payload = SummaryPayload(
                 workspace_name=workspace_name,
                 session_name=session_name,
                 message_id=message_id,
                 message_seq_in_session=message_seq_in_session,
+                message_public_id=message_public_id,
             )
 
         # Convert back to dict for compatibility with JSON serialization

src/routers/messages.py

@@ -66,6 +66,7 @@ async def create_messages_for_session(
                 "content": message.content,
                 "peer_name": message.peer_name,
                 "created_at": message.created_at,
+                "message_public_id": message.public_id,
             }
             for message in created_messages
         ]
@@ -127,6 +128,7 @@ async def create_messages_with_file(
             "content": message.content,
             "peer_name": message.peer_name,
             "created_at": message.created_at,
+            "message_public_id": message.public_id,
         }
         for message in created_messages
     ]

src/schemas.py

@@ -235,7 +235,12 @@ class Session(SessionBase):
 class Summary(BaseModel):
     content: str = Field(description="The summary text")
     message_id: int = Field(
-        description="The ID of the message that this summary covers up to"
+        description="The internal ID of the message that this summary covers up to",
+        exclude=True,
+    )
+    message_public_id: str = Field(
+        description="The public ID of the message that this summary covers up to",
+        serialization_alias="message_id",
     )
     summary_type: str = Field(description="The type of summary (short or long)")
     created_at: str = Field(

src/utils/summarizer.py

@@ -39,6 +39,7 @@ class Summary(TypedDict):
     summary_type: str
     created_at: str
     token_count: int
+    message_public_id: str
 
 
 def to_schema_summary(s: Summary) -> schemas.Summary:
@@ -48,6 +49,7 @@ def to_schema_summary(s: Summary) -> schemas.Summary:
         summary_type=s["summary_type"],
         created_at=s["created_at"],
         token_count=s["token_count"],
+        message_public_id=s["message_public_id"],
     )
 
 
@@ -180,6 +182,7 @@ async def summarize_if_needed(
     session_name: str,
     message_id: int,
     message_seq_in_session: int,
+    message_public_id: str,
 ) -> None:
     """
     Create short/long summaries if thresholds met.
@@ -209,6 +212,7 @@ async def summarize_if_needed(
                     session_name,
                     message_id,
                     SummaryType.LONG,
+                    message_public_id,
                 )
                 logger.info(
                     "Saved long summary for session %s covering up to message %s (%s in session)",
@@ -225,6 +229,7 @@ async def summarize_if_needed(
                     session_name,
                     message_id,
                     SummaryType.SHORT,
+                    message_public_id,
                 )
                 logger.info(
                     "Saved short summary for session %s covering up to message %s (%s in session)",
@@ -248,6 +253,7 @@ async def summarize_if_needed(
                     session_name,
                     message_id,
                     SummaryType.LONG,
+                    message_public_id,
                 )
                 logger.info(
                     "Saved long summary for session %s covering up to message %s (%s in session)",
@@ -262,6 +268,7 @@ async def summarize_if_needed(
                     session_name,
                     message_id,
                     SummaryType.SHORT,
+                    message_public_id,
                 )
                 logger.info(
                     "Saved short summary for session %s covering up to message %s (%s in session)",
@@ -277,6 +284,7 @@ async def _create_and_save_summary(
     session_name: str,
     message_id: int,
     summary_type: SummaryType,
+    message_public_id: str,
 ) -> None:
     """
     Create a new summary and save it to the database.
@@ -311,6 +319,7 @@ async def _create_and_save_summary(
         previous_summary_text=previous_summary_text,
         summary_type=summary_type,
         input_tokens=input_tokens,
+        message_public_id=message_public_id,
     )
 
     await _save_summary(
@@ -334,6 +343,7 @@ async def _create_summary(
     previous_summary_text: str | None,
     summary_type: SummaryType,
     input_tokens: int,
+    message_public_id: str,
 ) -> Summary:
     """
     Generate a summary of the provided messages using an LLM.
@@ -390,6 +400,7 @@ async def _create_summary(
         summary_type=summary_type.value,
         created_at=utc_now_iso(),
         token_count=summary_tokens,
+        message_public_id=message_public_id,
     )
 
 
@@ -610,6 +621,7 @@ async def get_session_context(
                 summary_type=latest_long_summary["summary_type"],
                 created_at=latest_long_summary["created_at"],
                 token_count=latest_long_summary["token_count"],
+                message_public_id=latest_long_summary["message_public_id"],
             )
             messages_tokens = token_limit - latest_long_summary["token_count"]
             messages_start_id = latest_long_summary["message_id"]
@@ -622,6 +634,7 @@ async def get_session_context(
                 summary_type=latest_short_summary["summary_type"],
                 created_at=latest_short_summary["created_at"],
                 token_count=latest_short_summary["token_count"],
+                message_public_id=latest_short_summary["message_public_id"],
             )
             messages_tokens = token_limit - latest_short_summary["token_count"]
             messages_start_id = latest_short_summary["message_id"]

tests/deriver/conftest.py

@@ -135,6 +135,7 @@ def create_queue_payload() -> Callable[..., Any]:
             "message_id": message.id,
             "content": message.content,
             "created_at": message.created_at or datetime.now(timezone.utc),
+            "message_public_id": message.public_id,
         }
 
         return create_payload(

tests/deriver/test_queue_processing.py

@@ -3,6 +3,7 @@ from typing import Any
 from unittest.mock import patch
 
 import pytest
+from nanoid import generate as generate_nanoid
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
@@ -706,6 +707,7 @@ class TestQueueProcessing:
                 workspace_name=session.workspace_name,
                 peer_name=peer.name,
                 content="First summary message",
+                public_id=generate_nanoid(),
             ),
             models.Message(
                 id=1000,
@@ -713,6 +715,7 @@ class TestQueueProcessing:
                 workspace_name=session.workspace_name,
                 peer_name=peer.name,
                 content="Second summary message",
+                public_id=generate_nanoid(),
             ),
         ]