Files
AutoGPT/autogpt_platform/backend/backend/api/features/chat/model.py
Zamil Majdy a78145505b fix(copilot): merge split assistant messages to prevent Anthropic API errors (#12062)
## Summary
- When the copilot model responds with both text content AND a
long-running tool call (e.g., `create_agent`), the streaming code
created two separate consecutive assistant messages — one with text, one
with `tool_calls`. This caused Anthropic's API to reject with
`"unexpected tool_use_id found in tool_result blocks"` because the
`tool_result` couldn't find a matching `tool_use` in the immediately
preceding assistant message.
- Added a defensive merge of consecutive assistant messages in
`to_openai_messages()` (fixes existing corrupt sessions too)
- Fixed `_yield_tool_call` to add tool_calls to the existing
current-turn assistant message instead of creating a new one
- Changed `accumulated_tool_calls` assignment to use `extend` to prevent
overwriting tool_calls added by long-running tool flow

## Test plan
- [x] All 23 chat feature tests pass (`backend/api/features/chat/`)
- [x] All 44 prompt utility tests pass (`backend/util/prompt_test.py`)
- [x] All pre-commit hooks pass (ruff, isort, black, pyright)
- [ ] Manual test: create an agent via copilot, then ask a follow-up
question — should no longer get 400 error

<!-- greptile_comment -->

<h2>Greptile Overview</h2>

<details><summary><h3>Greptile Summary</h3></summary>

Fixes a critical bug where long-running tool calls (like `create_agent`)
caused Anthropic API 400 errors due to split assistant messages. The fix
ensures tool calls are added to the existing assistant message instead
of creating new ones, and adds a defensive merge function to repair any
existing corrupt sessions.

**Key changes:**
- Added `_merge_consecutive_assistant_messages()` to defensively merge
split assistant messages in `to_openai_messages()`
- Modified `_yield_tool_call()` to append tool calls to the current-turn
assistant message instead of creating a new one
- Changed `accumulated_tool_calls` from assignment to `extend` to
preserve tool calls already added by long-running tool flow

**Impact:** Resolves the issue where users received 400 errors after
creating agents via copilot and asking follow-up questions.
</details>


<details><summary><h3>Confidence Score: 4/5</h3></summary>

- Safe to merge with minor verification recommended
- The changes are well-targeted and solve a real API compatibility
issue. The logic is sound: searching backwards for the current assistant
message is correct, and using `extend` instead of assignment prevents
overwriting. The defensive merge in `to_openai_messages()` also fixes
existing corrupt sessions. All existing tests pass according to the PR
description.
- No files require special attention - changes are localized and
defensive
</details>


<details><summary><h3>Sequence Diagram</h3></summary>

```mermaid
sequenceDiagram
    participant User
    participant StreamAPI as stream_chat_completion
    participant Chunks as _stream_chat_chunks
    participant ToolCall as _yield_tool_call
    participant Session as ChatSession
    
    User->>StreamAPI: Send message
    StreamAPI->>Chunks: Stream chat chunks
    
    alt Text + Long-running tool call
        Chunks->>StreamAPI: Text delta (content)
        StreamAPI->>Session: Append assistant message with content
        Chunks->>ToolCall: Tool call detected
        
        Note over ToolCall: OLD: Created new assistant message<br/>NEW: Appends to existing assistant
        
        ToolCall->>Session: Search backwards for current assistant
        ToolCall->>Session: Append tool_call to existing message
        ToolCall->>Session: Add pending tool result
    end
    
    StreamAPI->>StreamAPI: Merge accumulated_tool_calls
    Note over StreamAPI: Use extend (not assign)<br/>to preserve existing tool_calls
    
    StreamAPI->>Session: to_openai_messages()
    Session->>Session: _merge_consecutive_assistant_messages()
    Note over Session: Defensive: Merges any split<br/>assistant messages
    Session-->>StreamAPI: Merged messages
    
    StreamAPI->>User: Return response
```
</details>


<!-- greptile_other_comments_section -->

<!-- /greptile_comment -->
2026-02-12 01:52:17 +00:00

679 lines
25 KiB
Python

import asyncio
import logging
import uuid
from datetime import UTC, datetime
from typing import Any, cast
from weakref import WeakValueDictionary
from openai.types.chat import (
ChatCompletionAssistantMessageParam,
ChatCompletionDeveloperMessageParam,
ChatCompletionFunctionMessageParam,
ChatCompletionMessageParam,
ChatCompletionSystemMessageParam,
ChatCompletionToolMessageParam,
ChatCompletionUserMessageParam,
)
from openai.types.chat.chat_completion_assistant_message_param import FunctionCall
from openai.types.chat.chat_completion_message_tool_call_param import (
ChatCompletionMessageToolCallParam,
Function,
)
from prisma.models import ChatMessage as PrismaChatMessage
from prisma.models import ChatSession as PrismaChatSession
from pydantic import BaseModel
from backend.data.redis_client import get_redis_async
from backend.util import json
from backend.util.exceptions import DatabaseError, RedisError
from . import db as chat_db
from .config import ChatConfig
logger = logging.getLogger(__name__)
config = ChatConfig()
def _parse_json_field(value: str | dict | list | None, default: Any = None) -> Any:
"""Parse a JSON field that may be stored as string or already parsed."""
if value is None:
return default
if isinstance(value, str):
return json.loads(value)
return value
# Redis cache key prefix for chat sessions
CHAT_SESSION_CACHE_PREFIX = "chat:session:"
def _get_session_cache_key(session_id: str) -> str:
"""Get the Redis cache key for a chat session."""
return f"{CHAT_SESSION_CACHE_PREFIX}{session_id}"
# Session-level locks to prevent race conditions during concurrent upserts.
# Uses WeakValueDictionary to automatically garbage collect locks when no longer referenced,
# preventing unbounded memory growth while maintaining lock semantics for active sessions.
# Invalidation: Locks are auto-removed by GC when no coroutine holds a reference (after
# async with lock: completes). Explicit cleanup also occurs in delete_chat_session().
_session_locks: WeakValueDictionary[str, asyncio.Lock] = WeakValueDictionary()
_session_locks_mutex = asyncio.Lock()
async def _get_session_lock(session_id: str) -> asyncio.Lock:
"""Get or create a lock for a specific session to prevent concurrent upserts.
Uses WeakValueDictionary for automatic cleanup: locks are garbage collected
when no coroutine holds a reference to them, preventing memory leaks from
unbounded growth of session locks.
"""
async with _session_locks_mutex:
lock = _session_locks.get(session_id)
if lock is None:
lock = asyncio.Lock()
_session_locks[session_id] = lock
return lock
class ChatMessage(BaseModel):
role: str
content: str | None = None
name: str | None = None
tool_call_id: str | None = None
refusal: str | None = None
tool_calls: list[dict] | None = None
function_call: dict | None = None
class Usage(BaseModel):
prompt_tokens: int
completion_tokens: int
total_tokens: int
class ChatSession(BaseModel):
session_id: str
user_id: str
title: str | None = None
messages: list[ChatMessage]
usage: list[Usage]
credentials: dict[str, dict] = {} # Map of provider -> credential metadata
started_at: datetime
updated_at: datetime
successful_agent_runs: dict[str, int] = {}
successful_agent_schedules: dict[str, int] = {}
def add_tool_call_to_current_turn(self, tool_call: dict) -> None:
"""Attach a tool_call to the current turn's assistant message.
Searches backwards for the most recent assistant message (stopping at
any user message boundary). If found, appends the tool_call to it.
Otherwise creates a new assistant message with the tool_call.
"""
for msg in reversed(self.messages):
if msg.role == "user":
break
if msg.role == "assistant":
if not msg.tool_calls:
msg.tool_calls = []
msg.tool_calls.append(tool_call)
return
self.messages.append(
ChatMessage(role="assistant", content="", tool_calls=[tool_call])
)
@staticmethod
def new(user_id: str) -> "ChatSession":
return ChatSession(
session_id=str(uuid.uuid4()),
user_id=user_id,
title=None,
messages=[],
usage=[],
credentials={},
started_at=datetime.now(UTC),
updated_at=datetime.now(UTC),
)
@staticmethod
def from_db(
prisma_session: PrismaChatSession,
prisma_messages: list[PrismaChatMessage] | None = None,
) -> "ChatSession":
"""Convert Prisma models to Pydantic ChatSession."""
messages = []
if prisma_messages:
for msg in prisma_messages:
messages.append(
ChatMessage(
role=msg.role,
content=msg.content,
name=msg.name,
tool_call_id=msg.toolCallId,
refusal=msg.refusal,
tool_calls=_parse_json_field(msg.toolCalls),
function_call=_parse_json_field(msg.functionCall),
)
)
# Parse JSON fields from Prisma
credentials = _parse_json_field(prisma_session.credentials, default={})
successful_agent_runs = _parse_json_field(
prisma_session.successfulAgentRuns, default={}
)
successful_agent_schedules = _parse_json_field(
prisma_session.successfulAgentSchedules, default={}
)
# Calculate usage from token counts
usage = []
if prisma_session.totalPromptTokens or prisma_session.totalCompletionTokens:
usage.append(
Usage(
prompt_tokens=prisma_session.totalPromptTokens or 0,
completion_tokens=prisma_session.totalCompletionTokens or 0,
total_tokens=(prisma_session.totalPromptTokens or 0)
+ (prisma_session.totalCompletionTokens or 0),
)
)
return ChatSession(
session_id=prisma_session.id,
user_id=prisma_session.userId,
title=prisma_session.title,
messages=messages,
usage=usage,
credentials=credentials,
started_at=prisma_session.createdAt,
updated_at=prisma_session.updatedAt,
successful_agent_runs=successful_agent_runs,
successful_agent_schedules=successful_agent_schedules,
)
@staticmethod
def _merge_consecutive_assistant_messages(
messages: list[ChatCompletionMessageParam],
) -> list[ChatCompletionMessageParam]:
"""Merge consecutive assistant messages into single messages.
Long-running tool flows can create split assistant messages: one with
text content and another with tool_calls. Anthropic's API requires
tool_result blocks to reference a tool_use in the immediately preceding
assistant message, so these splits cause 400 errors via OpenRouter.
"""
if len(messages) < 2:
return messages
result: list[ChatCompletionMessageParam] = [messages[0]]
for msg in messages[1:]:
prev = result[-1]
if prev.get("role") != "assistant" or msg.get("role") != "assistant":
result.append(msg)
continue
prev = cast(ChatCompletionAssistantMessageParam, prev)
curr = cast(ChatCompletionAssistantMessageParam, msg)
curr_content = curr.get("content") or ""
if curr_content:
prev_content = prev.get("content") or ""
prev["content"] = (
f"{prev_content}\n{curr_content}" if prev_content else curr_content
)
curr_tool_calls = curr.get("tool_calls")
if curr_tool_calls:
prev_tool_calls = prev.get("tool_calls")
prev["tool_calls"] = (
list(prev_tool_calls) + list(curr_tool_calls)
if prev_tool_calls
else list(curr_tool_calls)
)
return result
def to_openai_messages(self) -> list[ChatCompletionMessageParam]:
messages = []
for message in self.messages:
if message.role == "developer":
m = ChatCompletionDeveloperMessageParam(
role="developer",
content=message.content or "",
)
if message.name:
m["name"] = message.name
messages.append(m)
elif message.role == "system":
m = ChatCompletionSystemMessageParam(
role="system",
content=message.content or "",
)
if message.name:
m["name"] = message.name
messages.append(m)
elif message.role == "user":
m = ChatCompletionUserMessageParam(
role="user",
content=message.content or "",
)
if message.name:
m["name"] = message.name
messages.append(m)
elif message.role == "assistant":
m = ChatCompletionAssistantMessageParam(
role="assistant",
content=message.content or "",
)
if message.function_call:
m["function_call"] = FunctionCall(
arguments=message.function_call["arguments"],
name=message.function_call["name"],
)
if message.refusal:
m["refusal"] = message.refusal
if message.tool_calls:
t: list[ChatCompletionMessageToolCallParam] = []
for tool_call in message.tool_calls:
# Tool calls are stored with nested structure: {id, type, function: {name, arguments}}
function_data = tool_call.get("function", {})
# Skip tool calls that are missing required fields
if "id" not in tool_call or "name" not in function_data:
logger.warning(
f"Skipping invalid tool call: missing required fields. "
f"Got: {tool_call.keys()}, function keys: {function_data.keys()}"
)
continue
# Arguments are stored as a JSON string
arguments_str = function_data.get("arguments", "{}")
t.append(
ChatCompletionMessageToolCallParam(
id=tool_call["id"],
type="function",
function=Function(
arguments=arguments_str,
name=function_data["name"],
),
)
)
m["tool_calls"] = t
if message.name:
m["name"] = message.name
messages.append(m)
elif message.role == "tool":
messages.append(
ChatCompletionToolMessageParam(
role="tool",
content=message.content or "",
tool_call_id=message.tool_call_id or "",
)
)
elif message.role == "function":
messages.append(
ChatCompletionFunctionMessageParam(
role="function",
content=message.content,
name=message.name or "",
)
)
return self._merge_consecutive_assistant_messages(messages)
async def _get_session_from_cache(session_id: str) -> ChatSession | None:
"""Get a chat session from Redis cache."""
redis_key = _get_session_cache_key(session_id)
async_redis = await get_redis_async()
raw_session: bytes | None = await async_redis.get(redis_key)
if raw_session is None:
return None
try:
session = ChatSession.model_validate_json(raw_session)
logger.info(
f"Loading session {session_id} from cache: "
f"message_count={len(session.messages)}, "
f"roles={[m.role for m in session.messages]}"
)
return session
except Exception as e:
logger.error(f"Failed to deserialize session {session_id}: {e}", exc_info=True)
raise RedisError(f"Corrupted session data for {session_id}") from e
async def _cache_session(session: ChatSession) -> None:
"""Cache a chat session in Redis."""
redis_key = _get_session_cache_key(session.session_id)
async_redis = await get_redis_async()
await async_redis.setex(redis_key, config.session_ttl, session.model_dump_json())
async def cache_chat_session(session: ChatSession) -> None:
"""Cache a chat session without persisting to the database."""
await _cache_session(session)
async def invalidate_session_cache(session_id: str) -> None:
"""Invalidate a chat session from Redis cache.
Used by background tasks to ensure fresh data is loaded on next access.
This is best-effort - Redis failures are logged but don't fail the operation.
"""
try:
redis_key = _get_session_cache_key(session_id)
async_redis = await get_redis_async()
await async_redis.delete(redis_key)
except Exception as e:
# Best-effort: log but don't fail - cache will expire naturally
logger.warning(f"Failed to invalidate session cache for {session_id}: {e}")
async def _get_session_from_db(session_id: str) -> ChatSession | None:
"""Get a chat session from the database."""
prisma_session = await chat_db.get_chat_session(session_id)
if not prisma_session:
return None
messages = prisma_session.Messages
logger.info(
f"Loading session {session_id} from DB: "
f"has_messages={messages is not None}, "
f"message_count={len(messages) if messages else 0}, "
f"roles={[m.role for m in messages] if messages else []}"
)
return ChatSession.from_db(prisma_session, messages)
async def _save_session_to_db(
session: ChatSession, existing_message_count: int
) -> None:
"""Save or update a chat session in the database."""
# Check if session exists in DB
existing = await chat_db.get_chat_session(session.session_id)
if not existing:
# Create new session
await chat_db.create_chat_session(
session_id=session.session_id,
user_id=session.user_id,
)
existing_message_count = 0
# Calculate total tokens from usage
total_prompt = sum(u.prompt_tokens for u in session.usage)
total_completion = sum(u.completion_tokens for u in session.usage)
# Update session metadata
await chat_db.update_chat_session(
session_id=session.session_id,
credentials=session.credentials,
successful_agent_runs=session.successful_agent_runs,
successful_agent_schedules=session.successful_agent_schedules,
total_prompt_tokens=total_prompt,
total_completion_tokens=total_completion,
)
# Add new messages (only those after existing count)
new_messages = session.messages[existing_message_count:]
if new_messages:
messages_data = []
for msg in new_messages:
messages_data.append(
{
"role": msg.role,
"content": msg.content,
"name": msg.name,
"tool_call_id": msg.tool_call_id,
"refusal": msg.refusal,
"tool_calls": msg.tool_calls,
"function_call": msg.function_call,
}
)
logger.info(
f"Saving {len(new_messages)} new messages to DB for session {session.session_id}: "
f"roles={[m['role'] for m in messages_data]}, "
f"start_sequence={existing_message_count}"
)
await chat_db.add_chat_messages_batch(
session_id=session.session_id,
messages=messages_data,
start_sequence=existing_message_count,
)
async def get_chat_session(
session_id: str,
user_id: str | None = None,
) -> ChatSession | None:
"""Get a chat session by ID.
Checks Redis cache first, falls back to database if not found.
Caches database results back to Redis.
Args:
session_id: The session ID to fetch.
user_id: If provided, validates that the session belongs to this user.
If None, ownership is not validated (admin/system access).
"""
# Try cache first
try:
session = await _get_session_from_cache(session_id)
if session:
# Verify user ownership if user_id was provided for validation
if user_id is not None and session.user_id != user_id:
logger.warning(
f"Session {session_id} user id mismatch: {session.user_id} != {user_id}"
)
return None
return session
except RedisError:
logger.warning(f"Cache error for session {session_id}, trying database")
except Exception as e:
logger.warning(f"Unexpected cache error for session {session_id}: {e}")
# Fall back to database
logger.info(f"Session {session_id} not in cache, checking database")
session = await _get_session_from_db(session_id)
if session is None:
logger.warning(f"Session {session_id} not found in cache or database")
return None
# Verify user ownership if user_id was provided for validation
if user_id is not None and session.user_id != user_id:
logger.warning(
f"Session {session_id} user id mismatch: {session.user_id} != {user_id}"
)
return None
# Cache the session from DB
try:
await _cache_session(session)
logger.info(f"Cached session {session_id} from database")
except Exception as e:
logger.warning(f"Failed to cache session {session_id}: {e}")
return session
async def upsert_chat_session(
session: ChatSession,
) -> ChatSession:
"""Update a chat session in both cache and database.
Uses session-level locking to prevent race conditions when concurrent
operations (e.g., background title update and main stream handler)
attempt to upsert the same session simultaneously.
Raises:
DatabaseError: If the database write fails. The cache is still updated
as a best-effort optimization, but the error is propagated to ensure
callers are aware of the persistence failure.
RedisError: If the cache write fails (after successful DB write).
"""
# Acquire session-specific lock to prevent concurrent upserts
lock = await _get_session_lock(session.session_id)
async with lock:
# Get existing message count from DB for incremental saves
existing_message_count = await chat_db.get_chat_session_message_count(
session.session_id
)
db_error: Exception | None = None
# Save to database (primary storage)
try:
await _save_session_to_db(session, existing_message_count)
except Exception as e:
logger.error(
f"Failed to save session {session.session_id} to database: {e}"
)
db_error = e
# Save to cache (best-effort, even if DB failed)
try:
await _cache_session(session)
except Exception as e:
# If DB succeeded but cache failed, raise cache error
if db_error is None:
raise RedisError(
f"Failed to persist chat session {session.session_id} to Redis: {e}"
) from e
# If both failed, log cache error but raise DB error (more critical)
logger.warning(
f"Cache write also failed for session {session.session_id}: {e}"
)
# Propagate DB error after attempting cache (prevents data loss)
if db_error is not None:
raise DatabaseError(
f"Failed to persist chat session {session.session_id} to database"
) from db_error
return session
async def create_chat_session(user_id: str) -> ChatSession:
"""Create a new chat session and persist it.
Raises:
DatabaseError: If the database write fails. We fail fast to ensure
callers never receive a non-persisted session that only exists
in cache (which would be lost when the cache expires).
"""
session = ChatSession.new(user_id)
# Create in database first - fail fast if this fails
try:
await chat_db.create_chat_session(
session_id=session.session_id,
user_id=user_id,
)
except Exception as e:
logger.error(f"Failed to create session {session.session_id} in database: {e}")
raise DatabaseError(
f"Failed to create chat session {session.session_id} in database"
) from e
# Cache the session (best-effort optimization, DB is source of truth)
try:
await _cache_session(session)
except Exception as e:
logger.warning(f"Failed to cache new session {session.session_id}: {e}")
return session
async def get_user_sessions(
user_id: str,
limit: int = 50,
offset: int = 0,
) -> tuple[list[ChatSession], int]:
"""Get chat sessions for a user from the database with total count.
Returns:
A tuple of (sessions, total_count) where total_count is the overall
number of sessions for the user (not just the current page).
"""
prisma_sessions = await chat_db.get_user_chat_sessions(user_id, limit, offset)
total_count = await chat_db.get_user_session_count(user_id)
sessions = []
for prisma_session in prisma_sessions:
# Convert without messages for listing (lighter weight)
sessions.append(ChatSession.from_db(prisma_session, None))
return sessions, total_count
async def delete_chat_session(session_id: str, user_id: str | None = None) -> bool:
"""Delete a chat session from both cache and database.
Args:
session_id: The session ID to delete.
user_id: If provided, validates that the session belongs to this user
before deletion. This prevents unauthorized deletion.
Returns:
True if deleted successfully, False otherwise.
"""
# Delete from database first (with optional user_id validation)
# This confirms ownership before invalidating cache
deleted = await chat_db.delete_chat_session(session_id, user_id)
if not deleted:
return False
# Only invalidate cache and clean up lock after DB confirms deletion
try:
redis_key = _get_session_cache_key(session_id)
async_redis = await get_redis_async()
await async_redis.delete(redis_key)
except Exception as e:
logger.warning(f"Failed to delete session {session_id} from cache: {e}")
# Clean up session lock (belt-and-suspenders with WeakValueDictionary)
async with _session_locks_mutex:
_session_locks.pop(session_id, None)
return True
async def update_session_title(session_id: str, title: str) -> bool:
"""Update only the title of a chat session.
This is a lightweight operation that doesn't touch messages, avoiding
race conditions with concurrent message updates. Use this for background
title generation instead of upsert_chat_session.
Args:
session_id: The session ID to update.
title: The new title to set.
Returns:
True if updated successfully, False otherwise.
"""
try:
result = await chat_db.update_chat_session(session_id=session_id, title=title)
if result is None:
logger.warning(f"Session {session_id} not found for title update")
return False
# Invalidate cache so next fetch gets updated title
try:
redis_key = _get_session_cache_key(session_id)
async_redis = await get_redis_async()
await async_redis.delete(redis_key)
except Exception as e:
logger.warning(f"Failed to invalidate cache for session {session_id}: {e}")
return True
except Exception as e:
logger.error(f"Failed to update title for session {session_id}: {e}")
return False