cai/src/cai/sdk/agents/models/openai_chatcompletions.py

4247 lines
203 KiB
Python

from __future__ import annotations
import asyncio
import dataclasses
import hashlib
import inspect
import json
import os
import re
import time
from collections.abc import AsyncIterator, Iterable
from dataclasses import dataclass, field
from typing import TYPE_CHECKING, Any, Literal, cast, overload
import litellm
import tiktoken
from openai import NOT_GIVEN, AsyncOpenAI, AsyncStream, NotGiven
# Create custom InputTokensDetails class since it's not available in current OpenAI version
from openai._models import BaseModel
from openai.types import ChatModel
from openai.types.chat import (
ChatCompletion,
ChatCompletionAssistantMessageParam,
ChatCompletionChunk,
ChatCompletionContentPartImageParam,
ChatCompletionContentPartParam,
ChatCompletionContentPartTextParam,
ChatCompletionDeveloperMessageParam,
ChatCompletionMessage,
ChatCompletionMessageParam,
ChatCompletionMessageToolCallParam,
ChatCompletionSystemMessageParam,
ChatCompletionToolChoiceOptionParam,
ChatCompletionToolMessageParam,
ChatCompletionUserMessageParam,
)
from openai.types.chat.chat_completion_tool_param import ChatCompletionToolParam
from openai.types.chat.completion_create_params import ResponseFormat
from openai.types.completion_usage import CompletionUsage
from openai.types.responses import (
EasyInputMessageParam,
Response,
ResponseCompletedEvent,
ResponseContentPartAddedEvent,
ResponseContentPartDoneEvent,
ResponseCreatedEvent,
ResponseFileSearchToolCallParam,
ResponseFunctionCallArgumentsDeltaEvent,
ResponseFunctionToolCall,
ResponseFunctionToolCallParam,
ResponseInputContentParam,
ResponseInputImageParam,
ResponseInputTextParam,
ResponseOutputItem,
ResponseOutputItemAddedEvent,
ResponseOutputItemDoneEvent,
ResponseOutputMessage,
ResponseOutputMessageParam,
ResponseOutputRefusal,
ResponseOutputText,
ResponseRefusalDeltaEvent,
ResponseTextDeltaEvent,
ResponseUsage,
)
from openai.types.responses.response_input_param import FunctionCallOutput, ItemReference, Message
from openai.types.responses.response_usage import OutputTokensDetails
from wasabi import color
from cai.sdk.agents.simple_agent_manager import SimpleAgentManager, AGENT_MANAGER
from cai.sdk.agents.parallel_isolation import PARALLEL_ISOLATION
from cai.sdk.agents.run_to_jsonl import get_session_recorder
from cai.sdk.agents.global_usage_tracker import GLOBAL_USAGE_TRACKER
from cai.util import (
_LIVE_STREAMING_PANELS,
COST_TRACKER,
calculate_model_cost,
cli_print_agent_messages,
cli_print_tool_output,
create_agent_streaming_context,
finish_agent_streaming,
get_ollama_api_base,
start_active_timer,
start_claude_thinking_if_applicable,
start_idle_timer,
stop_active_timer,
stop_idle_timer,
update_agent_streaming_content,
)
class InputTokensDetails(BaseModel):
prompt_tokens: int
"""The number of prompt tokens."""
cached_tokens: int = 0
"""The number of cached tokens."""
# Custom ResponseUsage that makes prompt_tokens/input_tokens and completion_tokens/output_tokens compatible
class CustomResponseUsage(ResponseUsage):
"""
Custom ResponseUsage class that provides compatibility between different field naming conventions.
Works with both input_tokens/output_tokens and prompt_tokens/completion_tokens.
"""
@property
def prompt_tokens(self) -> int:
"""Alias for input_tokens to maintain compatibility"""
return self.input_tokens
@property
def completion_tokens(self) -> int:
"""Alias for output_tokens to maintain compatibility"""
return self.output_tokens
from cai.internal.components.metrics import process_intermediate_logs
from .. import _debug
from ..agent_output import AgentOutputSchema
from ..exceptions import AgentsException, UserError
from ..handoffs import Handoff
from ..items import ModelResponse, TResponseInputItem, TResponseOutputItem, TResponseStreamEvent
from ..logger import logger
from ..tool import FunctionTool, Tool
from ..tracing import generation_span
from ..tracing.span_data import GenerationSpanData
from ..tracing.spans import Span
from ..usage import Usage
from ..version import __version__
from .fake_id import FAKE_RESPONSES_ID
from .interface import Model, ModelTracing
if TYPE_CHECKING:
from ..model_settings import ModelSettings
# Suppress debug info from litellm
litellm.suppress_debug_info = True
if os.getenv("CAI_MODEL") == "o3-mini" or os.getenv("CAI_MODEL") == "gemini-1.5-pro":
litellm.drop_params = True
_USER_AGENT = f"Agents/Python {__version__}"
_HEADERS = {"User-Agent": _USER_AGENT}
# Global registry to track active model instances
# This allows us to access instance-based histories for commands like /history
import weakref
import contextvars
# DEPRECATED: Use AGENT_REGISTRY instead
ACTIVE_MODEL_INSTANCES = {}
# Persistent message history store for agents without active instances
# This allows /load and /flush commands to work even when agents aren't running
PERSISTENT_MESSAGE_HISTORIES = {}
# Context variable to track the current active model per async context
_current_model_context = contextvars.ContextVar('current_model', default=None)
def set_current_active_model(model):
"""Set the current active model for tool execution context."""
_current_model_context.set(weakref.ref(model) if model else None)
def get_current_active_model():
"""Get the current active model."""
model_ref = _current_model_context.get()
if model_ref:
return model_ref()
return None
def get_agent_message_history(agent_name: str) -> list:
"""Get message history for a specific agent.
With SimpleAgentManager, this is much simpler - we only have one active agent.
"""
# Remove any ID suffix if present (e.g., "[P1]")
if "[" in agent_name and agent_name.endswith("]"):
base_name = agent_name.rsplit("[", 1)[0].strip()
else:
base_name = agent_name
# Get history from SimpleAgentManager
return AGENT_MANAGER.get_message_history(base_name)
def get_all_agent_histories() -> dict:
"""Get all agent message histories.
With SimpleAgentManager, we only track the active agent's history.
"""
return AGENT_MANAGER.get_all_histories()
def clear_agent_history(agent_name: str):
"""Clear history for a specific agent.
With SimpleAgentManager, this is much simpler.
"""
# Remove any ID suffix if present
if "[" in agent_name and agent_name.endswith("]"):
base_name = agent_name.rsplit("[", 1)[0].strip()
else:
base_name = agent_name
# Clear from SimpleAgentManager
AGENT_MANAGER.clear_history(base_name)
# Also clear the current instance if it matches
active_agent = AGENT_MANAGER.get_active_agent()
if active_agent and hasattr(active_agent, 'message_history'):
if hasattr(active_agent, 'agent_name') and active_agent.agent_name == base_name:
active_agent.message_history.clear()
# Reset context usage for this agent
os.environ['CAI_CONTEXT_USAGE'] = '0.0'
def clear_all_histories():
"""Clear all agent histories."""
# Clear from SimpleAgentManager
AGENT_MANAGER.clear_all_histories()
# Clear active agent's history if present
active_agent = AGENT_MANAGER.get_active_agent()
if active_agent and hasattr(active_agent, 'message_history'):
active_agent.message_history.clear()
# Clear all persistent histories
PERSISTENT_MESSAGE_HISTORIES.clear()
# Reset context usage since all histories are cleared
os.environ['CAI_CONTEXT_USAGE'] = '0.0'
@dataclass
class _StreamingState:
started: bool = False
text_content_index_and_output: tuple[int, ResponseOutputText] | None = None
refusal_content_index_and_output: tuple[int, ResponseOutputRefusal] | None = None
function_calls: dict[int, ResponseFunctionToolCall] = field(default_factory=dict)
# Add a new function for consistent token counting using tiktoken
def _check_reasoning_compatibility(messages):
"""
Check if message history is compatible with Claude reasoning/thinking.
According to Claude 4 docs, when reasoning is enabled, the final assistant
message must start with a thinking block. If there are assistant messages
with regular text content, reasoning should be disabled.
Args:
messages: List of message dictionaries
Returns:
bool: True if compatible with reasoning, False otherwise
"""
if not messages:
return True # Empty messages are compatible
# Find the last assistant message
last_assistant_msg = None
for msg in reversed(messages):
if msg.get("role") == "assistant":
last_assistant_msg = msg
break
if not last_assistant_msg:
return True # No assistant messages, compatible
# Check if the last assistant message has regular text content
content = last_assistant_msg.get("content")
if content:
# If it's a string with text content, not compatible
if isinstance(content, str) and content.strip():
return False
# If it's a list, check for text content blocks
elif isinstance(content, list):
for block in content:
if isinstance(block, dict):
if block.get("type") == "text" and block.get("text", "").strip():
return False
# Check if message has tool_calls (these are compatible)
if last_assistant_msg.get("tool_calls"):
return True
# If no content or only thinking blocks, it's compatible
return True
def count_tokens_with_tiktoken(text_or_messages):
"""
Count tokens consistently using tiktoken library.
Works with both strings and message lists.
Returns a tuple of (input_tokens, reasoning_tokens).
"""
if not text_or_messages:
return 0, 0
try:
# Try to use cl100k_base encoding (used by GPT-4 and GPT-3.5-turbo)
encoding = tiktoken.get_encoding("cl100k_base")
except:
# Fall back to GPT-2 encoding if cl100k is not available
try:
encoding = tiktoken.get_encoding("gpt2")
except:
# If tiktoken fails, fall back to character estimate
if isinstance(text_or_messages, str):
return len(text_or_messages) // 4, 0
elif isinstance(text_or_messages, list):
total_len = 0
for msg in text_or_messages:
if isinstance(msg, dict) and "content" in msg:
if isinstance(msg["content"], str):
total_len += len(msg["content"])
return total_len // 4, 0
else:
return 0, 0
# Process different input types
if isinstance(text_or_messages, str):
token_count = len(encoding.encode(text_or_messages))
return token_count, 0
elif isinstance(text_or_messages, list):
total_tokens = 0
reasoning_tokens = 0
# Add tokens for the messages format (ChatML format overhead)
# Each message has a base overhead (usually ~4 tokens)
total_tokens += len(text_or_messages) * 4
for msg in text_or_messages:
if isinstance(msg, dict):
# Add tokens for role
if "role" in msg:
total_tokens += len(encoding.encode(msg["role"]))
# Count content tokens
if "content" in msg and msg["content"]:
if isinstance(msg["content"], str):
content_tokens = len(encoding.encode(msg["content"]))
total_tokens += content_tokens
# Count tokens in assistant messages as reasoning tokens
if msg.get("role") == "assistant":
reasoning_tokens += content_tokens
elif isinstance(msg["content"], list):
for content_part in msg["content"]:
if isinstance(content_part, dict) and "text" in content_part:
part_tokens = len(encoding.encode(content_part["text"]))
total_tokens += part_tokens
if msg.get("role") == "assistant":
reasoning_tokens += part_tokens
return total_tokens, reasoning_tokens
else:
return 0, 0
class OpenAIChatCompletionsModel(Model):
"""OpenAI Chat Completions Model"""
INTERMEDIATE_LOG_INTERVAL = 5
def __init__(
self,
model: str | ChatModel,
openai_client: AsyncOpenAI,
agent_name: str = "CTF agent", # Default to CTF agent instead of generic "Agent"
agent_id: str | None = None,
agent_type: str | None = None, # The type of agent (e.g., "red_teamer")
) -> None:
self.model = model
self._client = openai_client
# Check if we're using OLLAMA models
self.is_ollama = os.getenv("OLLAMA") is not None and os.getenv("OLLAMA").lower() != "false"
self.empty_content_error_shown = False
# Track interaction counter and token totals for cli display
self.interaction_counter = 0
self.total_input_tokens = 0
self.total_output_tokens = 0
self.total_reasoning_tokens = 0
self.total_cost = 0.0
self.agent_name = agent_name
self.agent_type = agent_type or agent_name.lower().replace(" ", "_") # For registry tracking
self.uses_unified_context = False # Flag to indicate if using shared message history
# For SimpleAgentManager, we don't auto-register
# The agent will be registered when explicitly created by cli.py
self.agent_id = agent_id or AGENT_MANAGER.get_agent_id()
self._display_name = self.agent_name
# Instance-based message history
# Check if we have an isolated history for this agent (parallel mode)
if agent_id and PARALLEL_ISOLATION.is_parallel_mode():
isolated_history = PARALLEL_ISOLATION.get_isolated_history(agent_id)
if isolated_history is not None:
self.message_history = isolated_history
else:
self.message_history = []
else:
# Start with empty history - do not inherit from previous agents
self.message_history = []
# Register with SimpleAgentManager only when explicitly created
# This prevents phantom instances during module imports
if agent_id is not None: # Only register when explicitly given an ID
# For parallel agents, register as parallel
if agent_id.startswith("P") and int(os.getenv("CAI_PARALLEL", "1")) > 1:
AGENT_MANAGER.set_parallel_agent(agent_id, self, self.agent_name)
else:
AGENT_MANAGER.set_active_agent(self, self.agent_name, agent_id)
# Instance-based converter
self._converter = _Converter()
# Flags for CLI integration
self.disable_rich_streaming = False # Prevents creating a rich panel in the model
self.suppress_final_output = False # Prevents duplicate output at end of streaming
# Initialize the session logger
self.logger = get_session_recorder()
# DEPRECATED: Still maintain backward compatibility with ACTIVE_MODEL_INSTANCES
# TODO: Remove this after updating all dependent code
ACTIVE_MODEL_INSTANCES[(self._display_name, self.agent_id)] = weakref.ref(self)
def get_full_display_name(self) -> str:
"""Get the full display name including ID."""
return f"{self._display_name} [{self.agent_id}]"
def __del__(self):
"""Clean up when the model instance is destroyed."""
try:
# DEPRECATED: Remove from old registry for backward compatibility
if hasattr(self, '_display_name') and hasattr(self, 'agent_id'):
key = (self._display_name, self.agent_id)
if key in ACTIVE_MODEL_INSTANCES:
del ACTIVE_MODEL_INSTANCES[key]
# SimpleAgentManager handles history persistence
# No need to save to PERSISTENT_MESSAGE_HISTORIES
except Exception:
# Ignore any errors during cleanup
pass
def add_to_message_history(self, msg):
"""Add a message to this instance's history if it's not a duplicate.
Now only adds to the instance's local history, no global registry.
"""
is_duplicate = False
if self.message_history:
if msg.get("role") in ["system", "user"]:
is_duplicate = any(
existing.get("role") == msg.get("role")
and existing.get("content") == msg.get("content")
for existing in self.message_history
)
elif msg.get("role") == "assistant" and msg.get("tool_calls"):
# For tool calls, remove any existing message with the same tool call ID
# This handles the case where streaming might create duplicate entries
tool_call_id = msg["tool_calls"][0].get("id")
# Remove duplicates in-place to preserve list reference (important for swarm patterns)
indices_to_remove = []
for i, existing in enumerate(self.message_history):
if (existing.get("role") == "assistant"
and existing.get("tool_calls")
and existing["tool_calls"][0].get("id") == tool_call_id):
indices_to_remove.append(i)
# Remove in reverse order to avoid index shifting
for i in reversed(indices_to_remove):
self.message_history.pop(i)
is_duplicate = False # Always add after removing duplicates
elif msg.get("role") == "tool":
is_duplicate = any(
existing.get("role") == "tool"
and existing.get("tool_call_id") == msg.get("tool_call_id")
for existing in self.message_history
)
if not is_duplicate:
self.message_history.append(msg)
# Also update SimpleAgentManager
AGENT_MANAGER.add_to_history(self.agent_name, msg)
# Update isolated history if in parallel mode
if PARALLEL_ISOLATION.is_parallel_mode() and self.agent_id:
PARALLEL_ISOLATION.update_isolated_history(self.agent_id, msg)
def set_agent_name(self, name: str) -> None:
"""Set the agent name for CLI display purposes."""
self.agent_name = name
def _non_null_or_not_given(self, value: Any) -> Any:
return value if value is not None else NOT_GIVEN
async def get_response(
self,
system_instructions: str | None,
input: str | list[TResponseInputItem],
model_settings: ModelSettings,
tools: list[Tool],
output_schema: AgentOutputSchema | None,
handoffs: list[Handoff],
tracing: ModelTracing,
) -> ModelResponse:
# Increment the interaction counter for CLI display
self.interaction_counter += 1
self._intermediate_logs()
# Set this as the current active model for tool execution context
set_current_active_model(self)
# Stop idle timer and start active timer to track LLM processing time
stop_idle_timer()
start_active_timer()
with generation_span(
model=str(self.model),
model_config=dataclasses.asdict(model_settings)
| {"base_url": str(self._get_client().base_url)},
disabled=tracing.is_disabled(),
) as span_generation:
# Prepare the messages for consistent token counting
# IMPORTANT: Include existing message history for context
converted_messages = []
# First, add all existing messages from history
if self.message_history:
for msg in self.message_history:
msg_copy = msg.copy() # Use copy to avoid modifying original
# Remove any existing cache_control to avoid exceeding the 4-block limit
if "cache_control" in msg_copy:
del msg_copy["cache_control"]
converted_messages.append(msg_copy)
# Then convert and add the new input
new_messages = self._converter.items_to_messages(input, model_instance=self)
converted_messages.extend(new_messages)
if system_instructions:
# Check if we already have a system message
has_system = any(msg.get("role") == "system" for msg in converted_messages)
if not has_system:
converted_messages.insert(
0,
{
"content": system_instructions,
"role": "system",
},
)
# Add support for prompt caching for claude (not automatically applied)
# Gemini supports it too
# https://www.anthropic.com/news/token-saving-updates
# Maximize cache efficiency by using up to 4 cache_control blocks
if (str(self.model).startswith("claude") or "gemini" in str(self.model)) and len(
converted_messages
) > 0:
# Strategy: Cache the most valuable messages for maximum savings
# 1. System message (always first priority)
# 2. Long user messages (high token count)
# 3. Assistant messages with tool calls (complex context)
# 4. Recent context (last message)
cache_candidates = []
# Always cache system message if present
for i, msg in enumerate(converted_messages):
if msg.get("role") == "system":
cache_candidates.append((i, len(str(msg.get("content", ""))), "system"))
break
# Find long user messages and assistant messages with tool calls
for i, msg in enumerate(converted_messages):
content_len = len(str(msg.get("content", "")))
role = msg.get("role")
if role == "user" and content_len > 500: # Long user messages
cache_candidates.append((i, content_len, "user"))
elif role == "assistant" and msg.get("tool_calls"): # Tool calls
cache_candidates.append(
(i, content_len + 200, "assistant_tools")
) # Bonus for tool calls
# Always consider the last message for recent context
if len(converted_messages) > 1:
last_idx = len(converted_messages) - 1
last_msg = converted_messages[last_idx]
last_content_len = len(str(last_msg.get("content", "")))
cache_candidates.append((last_idx, last_content_len, "recent"))
# Sort by value (content length) and select top 4 unique indices
cache_candidates.sort(key=lambda x: x[1], reverse=True)
selected_indices = []
for idx, _, msg_type in cache_candidates:
if idx not in selected_indices:
selected_indices.append(idx)
if len(selected_indices) >= 4: # Max 4 cache blocks
break
# Apply cache_control to selected messages
for idx in selected_indices:
msg_copy = converted_messages[idx].copy()
msg_copy["cache_control"] = {"type": "ephemeral"}
converted_messages[idx] = msg_copy
# # --- Add to message_history: user, system, and assistant tool call messages ---
# # Add system prompt to message_history
# if system_instructions:
# sys_msg = {
# "role": "system",
# "content": system_instructions
# }
# self.add_to_message_history(sys_msg)
# Add user prompt(s) to message_history
if isinstance(input, str):
user_msg = {"role": "user", "content": input}
self.add_to_message_history(user_msg)
# Log the user message
self.logger.log_user_message(input)
elif isinstance(input, list):
for item in input:
# Try to extract user messages
if isinstance(item, dict):
if item.get("role") == "user":
user_msg = {"role": "user", "content": item.get("content", "")}
self.add_to_message_history(user_msg)
# Log the user message
if item.get("content"):
self.logger.log_user_message(item.get("content"))
# IMPORTANT: Ensure the message list has valid tool call/result pairs
# This needs to happen before the API call to prevent errors
try:
from cai.util import fix_message_list
converted_messages = fix_message_list(converted_messages)
except Exception:
pass
# Get token count estimate before API call for consistent counting
estimated_input_tokens, _ = count_tokens_with_tiktoken(converted_messages)
# Calculate and set context usage for toolbar
max_tokens = self._get_model_max_tokens(str(self.model))
context_usage = estimated_input_tokens / max_tokens if max_tokens > 0 else 0.0
os.environ['CAI_CONTEXT_USAGE'] = str(context_usage)
# Check if auto-compaction is needed
input, system_instructions, compacted = await self._auto_compact_if_needed(estimated_input_tokens, input, system_instructions)
# If compaction occurred, recalculate tokens with new input
if compacted:
converted_messages = self._converter.items_to_messages(input, model_instance=self)
if system_instructions:
converted_messages.insert(0, {"role": "system", "content": system_instructions})
estimated_input_tokens, _ = count_tokens_with_tiktoken(converted_messages)
# Pre-check price limit using estimated input tokens and a conservative estimate for output
# This prevents starting a request that would immediately exceed the price limit
if hasattr(COST_TRACKER, "check_price_limit"):
# Use a conservative estimate for output tokens (roughly equal to input)
estimated_cost = calculate_model_cost(
str(self.model), estimated_input_tokens, estimated_input_tokens
) # Conservative estimate
try:
COST_TRACKER.check_price_limit(estimated_cost)
except Exception:
# Stop active timer and start idle timer before re-raising the exception
stop_active_timer()
start_idle_timer()
raise
try:
response = await self._fetch_response(
system_instructions,
input,
model_settings,
tools,
output_schema,
handoffs,
span_generation,
tracing,
stream=False,
)
except KeyboardInterrupt:
# Handle KeyboardInterrupt during API call
# Clean up any pending tool calls that weren't executed
if hasattr(self, "_pending_tool_calls"):
# Clear all pending tool calls to prevent incomplete history
self._pending_tool_calls.clear()
# Let the interrupt propagate up to end the current operation
stop_active_timer()
start_idle_timer()
raise
if _debug.DONT_LOG_MODEL_DATA:
logger.debug("Received model response")
else:
import json
logger.debug(
f"LLM resp:\n{json.dumps(response.choices[0].message.model_dump(), indent=2)}\n"
)
# Ensure we have reasonable token counts
if response.usage:
input_tokens = response.usage.prompt_tokens
output_tokens = response.usage.completion_tokens
total_tokens = response.usage.total_tokens
# Use estimated tokens if API returns zeroes or implausible values
if input_tokens == 0 or input_tokens < (len(str(input)) // 10): # Sanity check
input_tokens = estimated_input_tokens
total_tokens = input_tokens + output_tokens
# # Debug information
# print(f"\nDEBUG CONSISTENT TOKEN COUNTS - API tokens: input={input_tokens}, output={output_tokens}, total={total_tokens}")
# print(f"Estimated tokens were: input={estimated_input_tokens}")
else:
# If no usage info, use our estimates
input_tokens = estimated_input_tokens
output_tokens = 0
total_tokens = input_tokens
# print(f"\nDEBUG CONSISTENT TOKEN COUNTS - No API tokens, using estimates: input={input_tokens}, output={output_tokens}")
# Update token totals for CLI display
self.total_input_tokens += input_tokens
self.total_output_tokens += output_tokens
reasoning_tokens = 0
if (
response.usage
and hasattr(response.usage, "completion_tokens_details")
and response.usage.completion_tokens_details
and hasattr(response.usage.completion_tokens_details, "reasoning_tokens")
):
reasoning_tokens = response.usage.completion_tokens_details.reasoning_tokens
self.total_reasoning_tokens += reasoning_tokens
# Process costs for non-streaming mode
model_name = str(self.model)
interaction_cost = calculate_model_cost(model_name, input_tokens, output_tokens)
# Process the costs through COST_TRACKER only once
if interaction_cost > 0.0:
# Check price limit before processing
if hasattr(COST_TRACKER, "check_price_limit"):
COST_TRACKER.check_price_limit(interaction_cost)
# Process interaction cost
COST_TRACKER.process_interaction_cost(
model_name,
input_tokens,
output_tokens,
reasoning_tokens,
interaction_cost
)
# Process total cost
total_cost = COST_TRACKER.process_total_cost(
model_name,
self.total_input_tokens,
self.total_output_tokens,
self.total_reasoning_tokens,
None
)
# Track usage globally
GLOBAL_USAGE_TRACKER.track_usage(
model_name=model_name,
input_tokens=input_tokens,
output_tokens=output_tokens,
cost=interaction_cost,
agent_name=self.agent_name
)
else:
# For free models
total_cost = COST_TRACKER.session_total_cost
# Still track token usage even for free models
GLOBAL_USAGE_TRACKER.track_usage(
model_name=model_name,
input_tokens=input_tokens,
output_tokens=output_tokens,
cost=0.0,
agent_name=self.agent_name
)
# Check if this message contains tool calls
tool_output = None
should_display_message = True
if (
hasattr(response.choices[0].message, "tool_calls")
and response.choices[0].message.tool_calls
):
# For each tool call in the message, get corresponding output if available
for tool_call in response.choices[0].message.tool_calls:
call_id = tool_call.id
# Check if this tool call has already been displayed
if (
hasattr(_Converter, "tool_outputs")
and call_id in self._converter.tool_outputs
):
tool_output_content = self._converter.tool_outputs[call_id]
# Check if this is a command sent to an existing async session
is_async_session_input = False
has_auto_output = False
is_regular_command = False
try:
import json
# Handle empty arguments before trying to parse JSON
tool_args = tool_call.function.arguments
if tool_args is None or (isinstance(tool_args, str) and tool_args.strip() == ""):
tool_args = "{}"
args = json.loads(tool_args)
# Check if this is a regular command (not a session command)
if (
isinstance(args, dict)
and args.get("command")
and not args.get("session_id")
and not args.get("async_mode")
):
is_regular_command = True
# Only consider it an async session input if it has session_id AND it's not creating a new session
elif (
isinstance(args, dict)
and args.get("session_id")
and not args.get("async_mode") # Not creating a new session
and not args.get("creating_session")
): # Not marked as session creation
is_async_session_input = True
# Check if this has auto_output flag
has_auto_output = args.get("auto_output", False)
except:
pass
# For regular commands that were already shown via streaming, suppress the agent message
if (
is_regular_command
and tool_call.function.name == "generic_linux_command"
):
# Check if this was executed very recently (likely shown via streaming)
if (
hasattr(_Converter, "recent_tool_calls")
and call_id in self._converter.recent_tool_calls
):
tool_call_info = self._converter.recent_tool_calls[call_id]
if "start_time" in tool_call_info:
import time
time_since_execution = (
time.time() - tool_call_info["start_time"]
)
# If executed within last 2 seconds, it was likely shown via streaming
if time_since_execution < 2.0:
should_display_message = False
tool_output = None
elif is_async_session_input:
should_display_message = True
tool_output = None
# For async session inputs without auto_output, always show the agent message
elif is_async_session_input and not has_auto_output:
should_display_message = True
tool_output = None
# For session creation messages, also show them
elif (
"Started async session" in tool_output_content
or "session" in tool_output_content.lower()
and "async" in tool_output_content.lower()
):
should_display_message = True
tool_output = None
else:
# For other tool calls, check if we should suppress based on timing
# Only suppress if this tool was JUST executed (within last 2 seconds)
if (
hasattr(_Converter, "recent_tool_calls")
and call_id in self._converter.recent_tool_calls
):
tool_call_info = self._converter.recent_tool_calls[call_id]
if "start_time" in tool_call_info:
import time
time_since_execution = (
time.time() - tool_call_info["start_time"]
)
# Only suppress if this was executed very recently
if time_since_execution < 2.0:
should_display_message = False
else:
# For older tool calls, show the message
should_display_message = True
break
# Additional check: Always show messages that have text content
# This ensures agent explanations are not suppressed
if (
hasattr(response.choices[0].message, "content")
and response.choices[0].message.content
and str(response.choices[0].message.content).strip()
):
# If the message has actual text content, always show it
should_display_message = True
# Display the agent message (this will show the command for async sessions)
if should_display_message:
# Ensure we're in non-streaming mode for proper markdown parsing
previous_stream_setting = os.environ.get("CAI_STREAM", "false")
os.environ["CAI_STREAM"] = "false" # Force non-streaming mode for markdown parsing
# Print the agent message for CLI display
cli_print_agent_messages(
agent_name=getattr(self, "agent_name", "Agent"),
message=response.choices[0].message,
counter=getattr(self, "interaction_counter", 0),
model=str(self.model),
debug=False,
interaction_input_tokens=input_tokens,
interaction_output_tokens=output_tokens,
interaction_reasoning_tokens=reasoning_tokens,
total_input_tokens=getattr(self, "total_input_tokens", 0),
total_output_tokens=getattr(self, "total_output_tokens", 0),
total_reasoning_tokens=getattr(self, "total_reasoning_tokens", 0),
interaction_cost=interaction_cost,
total_cost=total_cost,
tool_output=tool_output, # Pass tool_output only when needed
suppress_empty=True, # Keep suppress_empty=True as requested
)
# Restore previous streaming setting
os.environ["CAI_STREAM"] = previous_stream_setting
# --- DEFERRED: Tool calls are no longer added immediately ---
# Tool calls will be added atomically with their responses
# to prevent incomplete message history on interruption
assistant_msg = response.choices[0].message
if hasattr(assistant_msg, "tool_calls") and assistant_msg.tool_calls:
# Store pending tool calls but don't add to history yet
if not hasattr(self, "_pending_tool_calls"):
self._pending_tool_calls = {}
for tool_call in assistant_msg.tool_calls:
# Handle empty arguments before storing
tool_args = tool_call.function.arguments
if tool_args is None or (isinstance(tool_args, str) and tool_args.strip() == ""):
tool_args = "{}"
# Compose a message for the tool call
tool_call_msg = {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call.id,
"type": tool_call.type,
"function": {
"name": tool_call.function.name,
"arguments": tool_args,
},
}
],
}
# Store for later atomic addition with response
self._pending_tool_calls[tool_call.id] = tool_call_msg
# Save the tool call details for later matching with output
# This is important for non-streaming mode to track tool calls properly
if not hasattr(self._converter, "recent_tool_calls"):
self._converter.recent_tool_calls = {}
# Store the tool call by ID for later reference
import time
self._converter.recent_tool_calls[tool_call.id] = {
"name": tool_call.function.name,
"arguments": tool_call.function.arguments,
"start_time": time.time(),
"execution_info": {"start_time": time.time()},
}
# Log the assistant tool call message
tool_calls_list = []
for tool_call in assistant_msg.tool_calls:
tool_calls_list.append(
{
"id": tool_call.id,
"type": tool_call.type,
"function": {
"name": tool_call.function.name,
"arguments": tool_call.function.arguments,
},
}
)
self.logger.log_assistant_message(None, tool_calls_list)
# If the assistant message is just text, add it as well
elif hasattr(assistant_msg, "content") and assistant_msg.content:
asst_msg = {"role": "assistant", "content": assistant_msg.content}
self.add_to_message_history(asst_msg)
# Log the assistant message
self.logger.log_assistant_message(assistant_msg.content)
# En no-streaming, también necesitamos añadir cualquier tool output al message_history
# Esto se hace procesando los items de output del ModelResponse
items = self._converter.message_to_output_items(response.choices[0].message)
# Además, necesitamos añadir los tool outputs que se hayan generado
# durante la ejecución de las herramientas
if hasattr(_Converter, "tool_outputs"):
for call_id, output_content in self._converter.tool_outputs.items():
# Verificar si ya existe un mensaje tool con este call_id en message_history
tool_msg_exists = any(
msg.get("role") == "tool" and msg.get("tool_call_id") == call_id
for msg in message_history
)
if not tool_msg_exists:
# Añadir el mensaje tool al message_history
tool_msg = {
"role": "tool",
"tool_call_id": call_id,
"content": output_content,
}
self.add_to_message_history(tool_msg)
# Log the complete response for the session
self.logger.rec_training_data(
{
"model": str(self.model),
"messages": converted_messages,
"stream": False,
"tools": [t.params_json_schema for t in tools] if tools else [],
"tool_choice": model_settings.tool_choice,
},
response,
self.total_cost,
self.agent_name,
)
usage = (
Usage(
requests=1,
input_tokens=input_tokens,
output_tokens=output_tokens,
total_tokens=input_tokens + output_tokens,
)
if response.usage or input_tokens > 0
else Usage()
)
if tracing.include_data():
span_generation.span_data.output = [response.choices[0].message.model_dump()]
span_generation.span_data.usage = {
"input_tokens": usage.input_tokens,
"output_tokens": usage.output_tokens,
}
items = self._converter.message_to_output_items(response.choices[0].message)
# For non-streaming responses, make sure we also log token usage with compatible field names
# This ensures both streaming and non-streaming use consistent naming
if not hasattr(response, "usage"):
response.usage = {}
if hasattr(response.usage, "prompt_tokens") and not hasattr(
response.usage, "input_tokens"
):
response.usage.input_tokens = response.usage.prompt_tokens
if hasattr(response.usage, "completion_tokens") and not hasattr(
response.usage, "output_tokens"
):
response.usage.output_tokens = response.usage.completion_tokens
# Ensure cost is properly initialized
if not hasattr(response, "cost"):
response.cost = None
return ModelResponse(
output=items,
usage=usage,
referenceable_id=None,
)
# Stop active timer and start idle timer when response is complete
stop_active_timer()
start_idle_timer()
async def stream_response(
self,
system_instructions: str | None,
input: str | list[TResponseInputItem],
model_settings: ModelSettings,
tools: list[Tool],
output_schema: AgentOutputSchema | None,
handoffs: list[Handoff],
tracing: ModelTracing,
) -> AsyncIterator[TResponseStreamEvent]:
"""
Yields a partial message as it is generated, as well as the usage information.
"""
# Initialize streaming contexts as None
streaming_context = None
thinking_context = None
stream_interrupted = False
try:
# IMPORTANT: Pre-process input to ensure it's in the correct format
# for streaming. This helps prevent errors during stream handling.
if not isinstance(input, str):
# Convert input items to messages and verify structure
try:
input_items = list(input) # Make sure it's a list
# Pre-verify the input messages to avoid errors during streaming
from cai.util import fix_message_list
# Apply fix_message_list to the input items that are dictionaries
dict_items = [item for item in input_items if isinstance(item, dict)]
if dict_items:
fixed_dict_items = fix_message_list(dict_items)
# Replace the original dict items with fixed ones while preserving non-dict items
new_input = []
dict_index = 0
for item in input_items:
if isinstance(item, dict):
if dict_index < len(fixed_dict_items):
new_input.append(fixed_dict_items[dict_index])
dict_index += 1
else:
new_input.append(item)
# Update input with the fixed version
input = new_input
except Exception as e:
# Silently continue with original input if pre-processing failed
# This is not critical and shouldn't show warnings
pass
# Increment the interaction counter for CLI display
self.interaction_counter += 1
self._intermediate_logs()
# Stop idle timer and start active timer to track LLM processing time
stop_idle_timer()
start_active_timer()
# --- Check if streaming should be shown in rich panel ---
should_show_rich_stream = (
os.getenv("CAI_STREAM", "false").lower() == "true"
and not self.disable_rich_streaming
)
# Create streaming context if needed
if should_show_rich_stream:
try:
streaming_context = create_agent_streaming_context(
agent_name=self.agent_name,
counter=self.interaction_counter,
model=str(self.model),
)
except Exception as e:
# Silently fall back to non-streaming display
streaming_context = None
with generation_span(
model=str(self.model),
model_config=dataclasses.asdict(model_settings)
| {"base_url": str(self._get_client().base_url)},
disabled=tracing.is_disabled(),
) as span_generation:
# Prepare messages for consistent token counting
converted_messages = self._converter.items_to_messages(input, model_instance=self)
if system_instructions:
converted_messages.insert(
0,
{
"content": system_instructions,
"role": "system",
},
)
# Add support for prompt caching for claude (not automatically applied)
# Gemini supports it too
# https://www.anthropic.com/news/token-saving-updates
# Maximize cache efficiency by using up to 4 cache_control blocks
if (str(self.model).startswith("claude") or "gemini" in str(self.model)) and len(
converted_messages
) > 0:
# Strategy: Cache the most valuable messages for maximum savings
# 1. System message (always first priority)
# 2. Long user messages (high token count)
# 3. Assistant messages with tool calls (complex context)
# 4. Recent context (last message)
cache_candidates = []
# Always cache system message if present
for i, msg in enumerate(converted_messages):
if msg.get("role") == "system":
cache_candidates.append((i, len(str(msg.get("content", ""))), "system"))
break
# Find long user messages and assistant messages with tool calls
for i, msg in enumerate(converted_messages):
content_len = len(str(msg.get("content", "")))
role = msg.get("role")
if role == "user" and content_len > 500: # Long user messages
cache_candidates.append((i, content_len, "user"))
elif role == "assistant" and msg.get("tool_calls"): # Tool calls
cache_candidates.append(
(i, content_len + 200, "assistant_tools")
) # Bonus for tool calls
# Always consider the last message for recent context
if len(converted_messages) > 1:
last_idx = len(converted_messages) - 1
last_msg = converted_messages[last_idx]
last_content_len = len(str(last_msg.get("content", "")))
cache_candidates.append((last_idx, last_content_len, "recent"))
# Sort by value (content length) and select top 4 unique indices
cache_candidates.sort(key=lambda x: x[1], reverse=True)
selected_indices = []
for idx, _, msg_type in cache_candidates:
if idx not in selected_indices:
selected_indices.append(idx)
if len(selected_indices) >= 4: # Max 4 cache blocks
break
# Apply cache_control to selected messages
for idx in selected_indices:
msg_copy = converted_messages[idx].copy()
msg_copy["cache_control"] = {"type": "ephemeral"}
converted_messages[idx] = msg_copy
# # --- Add to message_history: user, system prompts ---
# if system_instructions:
# sys_msg = {
# "role": "system",
# "content": system_instructions
# }
# self.add_to_message_history(sys_msg)
if isinstance(input, str):
user_msg = {"role": "user", "content": input}
self.add_to_message_history(user_msg)
# Log the user message
self.logger.log_user_message(input)
elif isinstance(input, list):
for item in input:
if isinstance(item, dict):
if item.get("role") == "user":
user_msg = {"role": "user", "content": item.get("content", "")}
self.add_to_message_history(user_msg)
# Log the user message
if item.get("content"):
self.logger.log_user_message(item.get("content"))
# Get token count estimate before API call for consistent counting
estimated_input_tokens, _ = count_tokens_with_tiktoken(converted_messages)
# Check if auto-compaction is needed
input, system_instructions, compacted = await self._auto_compact_if_needed(estimated_input_tokens, input, system_instructions)
# If compaction occurred, recalculate tokens with new input
if compacted:
converted_messages = self._converter.items_to_messages(input, model_instance=self)
if system_instructions:
converted_messages.insert(0, {"role": "system", "content": system_instructions})
estimated_input_tokens, _ = count_tokens_with_tiktoken(converted_messages)
# Pre-check price limit using estimated input tokens and a conservative estimate for output
# This prevents starting a stream that would immediately exceed the price limit
if hasattr(COST_TRACKER, "check_price_limit"):
# Use a conservative estimate for output tokens (roughly equal to input)
estimated_cost = calculate_model_cost(
str(self.model), estimated_input_tokens, estimated_input_tokens
) # Conservative estimate
try:
COST_TRACKER.check_price_limit(estimated_cost)
except Exception:
# Ensure streaming context is cleaned up in case of errors
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
except Exception:
pass
# Stop active timer and start idle timer before re-raising the exception
stop_active_timer()
start_idle_timer()
raise
response, stream = await self._fetch_response(
system_instructions,
input,
model_settings,
tools,
output_schema,
handoffs,
span_generation,
tracing,
stream=True,
)
usage: CompletionUsage | None = None
state = _StreamingState()
# Manual token counting (when API doesn't provide it)
output_text = ""
estimated_output_tokens = 0
# Initialize a streaming text accumulator for rich display
streaming_text_buffer = ""
# For tool call streaming, accumulate tool_calls to add to message_history at the end
streamed_tool_calls = []
# Initialize Claude thinking display if applicable
if should_show_rich_stream: # Only show thinking in rich streaming mode
thinking_context = start_claude_thinking_if_applicable(
str(self.model), self.agent_name, self.interaction_counter
)
# Ollama specific: accumulate full content to check for function calls at the end
# Some Ollama models output the function call as JSON in the text content
ollama_full_content = ""
is_ollama = False
model_str = str(self.model).lower()
is_ollama = (
self.is_ollama
or "ollama" in model_str
or ":" in model_str
or "qwen" in model_str
)
# Add visual separation before agent output
if streaming_context and should_show_rich_stream:
# If we're using rich context, we'll add separation through that
pass
else:
# Removed clear visual separator to avoid blank lines during streaming
pass
try:
async for chunk in stream:
# Check if we've been interrupted
if stream_interrupted:
break
if not state.started:
state.started = True
yield ResponseCreatedEvent(
response=response,
type="response.created",
)
# The usage is only available in the last chunk
if hasattr(chunk, "usage"):
usage = chunk.usage
# For Ollama/LiteLLM streams that don't have usage attribute
else:
usage = None
# Handle different stream chunk formats
if hasattr(chunk, "choices") and chunk.choices:
choices = chunk.choices
elif hasattr(chunk, "delta") and chunk.delta:
# Some providers might return delta directly
choices = [{"delta": chunk.delta}]
elif isinstance(chunk, dict) and "choices" in chunk:
choices = chunk["choices"]
# Special handling for Qwen/Ollama chunks
elif isinstance(chunk, dict) and (
"content" in chunk or "function_call" in chunk
):
# Qwen direct delta format - convert to standard
choices = [{"delta": chunk}]
else:
# Skip chunks that don't contain choice data
continue
if not choices or len(choices) == 0:
continue
# Get the delta content
delta = None
if hasattr(choices[0], "delta"):
delta = choices[0].delta
elif isinstance(choices[0], dict) and "delta" in choices[0]:
delta = choices[0]["delta"]
if not delta:
continue
# Handle Claude reasoning content first (before regular content)
reasoning_content = None
# Check for Claude reasoning in different possible formats
if (
hasattr(delta, "reasoning_content")
and delta.reasoning_content is not None
):
reasoning_content = delta.reasoning_content
elif (
isinstance(delta, dict)
and "reasoning_content" in delta
and delta["reasoning_content"] is not None
):
reasoning_content = delta["reasoning_content"]
# Also check for thinking_blocks structure (Claude 4 format)
thinking_blocks = None
if hasattr(delta, "thinking_blocks") and delta.thinking_blocks is not None:
thinking_blocks = delta.thinking_blocks
elif (
isinstance(delta, dict)
and "thinking_blocks" in delta
and delta["thinking_blocks"] is not None
):
thinking_blocks = delta["thinking_blocks"]
# Extract reasoning content from thinking blocks if available
if thinking_blocks and not reasoning_content:
for block in thinking_blocks:
if isinstance(block, dict) and block.get("type") == "thinking":
reasoning_content = block.get("thinking", "")
break
elif (
isinstance(block, dict)
and block.get("type") == "text"
and "thinking" in str(block)
):
# Sometimes thinking content comes as text blocks
reasoning_content = block.get("text", "")
break
# Check for direct thinking field (some Claude models)
if not reasoning_content:
if hasattr(delta, "thinking") and delta.thinking is not None:
reasoning_content = delta.thinking
elif (
isinstance(delta, dict)
and "thinking" in delta
and delta["thinking"] is not None
):
reasoning_content = delta["thinking"]
# Update thinking display if we have reasoning content
if reasoning_content:
if thinking_context:
# Streaming mode: Update the rich thinking display
from cai.util import update_claude_thinking_content
update_claude_thinking_content(thinking_context, reasoning_content)
else:
# Non-streaming mode: Use simple text output
from cai.util import (
detect_claude_thinking_in_stream,
print_claude_reasoning_simple,
)
# Check if model supports reasoning (Claude or DeepSeek)
model_str_lower = str(self.model).lower()
if (
detect_claude_thinking_in_stream(str(self.model))
or "deepseek" in model_str_lower
):
print_claude_reasoning_simple(
reasoning_content, self.agent_name, str(self.model)
)
# Handle text
content = None
if hasattr(delta, "content") and delta.content is not None:
content = delta.content
elif (
isinstance(delta, dict)
and "content" in delta
and delta["content"] is not None
):
content = delta["content"]
if content:
# IMPORTANT: If we have content and thinking_context is active,
# it means thinking is complete and normal content is starting
# Close the thinking display automatically
if thinking_context:
from cai.util import finish_claude_thinking_display
finish_claude_thinking_display(thinking_context)
thinking_context = None # Clear the context
# For Ollama, we need to accumulate the full content to check for function calls
if is_ollama:
ollama_full_content += content
# Add to the streaming text buffer
streaming_text_buffer += content
# Update streaming display if enabled - ALWAYS respect CAI_STREAM setting
# Both thinking and regular content should stream if streaming is enabled
if streaming_context:
# Calculate cost for current interaction
current_cost = calculate_model_cost(
str(self.model), estimated_input_tokens, estimated_output_tokens
)
# Check price limit only for paid models
if (
current_cost > 0
and hasattr(COST_TRACKER, "check_price_limit")
and estimated_output_tokens % 50 == 0
):
try:
COST_TRACKER.check_price_limit(current_cost)
except Exception:
# Ensure streaming context is cleaned up
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
except Exception:
pass
# Stop timers and re-raise the exception
stop_active_timer()
start_idle_timer()
raise
# Update session total cost for real-time display
# This is a temporary estimate during streaming that will be properly updated at the end
estimated_session_total = getattr(
COST_TRACKER, "session_total_cost", 0.0
)
# For free models, don't add to the total cost
display_total_cost = estimated_session_total
if current_cost > 0:
display_total_cost += current_cost
# Create token stats with both current interaction cost and updated total cost
token_stats = {
"input_tokens": estimated_input_tokens,
"output_tokens": estimated_output_tokens,
"cost": current_cost,
"total_cost": display_total_cost,
}
update_agent_streaming_content(
streaming_context, content, token_stats
)
# More accurate token counting for text content
output_text += content
token_count, _ = count_tokens_with_tiktoken(output_text)
estimated_output_tokens = token_count
# Periodically check price limit during streaming
# This allows early termination if price limit is reached mid-stream
if (
estimated_output_tokens > 0 and estimated_output_tokens % 50 == 0
): # Check every ~50 tokens
# Calculate current estimated cost
current_estimated_cost = calculate_model_cost(
str(self.model), estimated_input_tokens, estimated_output_tokens
)
# Check price limit only for paid models
if current_estimated_cost > 0 and hasattr(
COST_TRACKER, "check_price_limit"
):
try:
COST_TRACKER.check_price_limit(current_estimated_cost)
except Exception:
# Ensure streaming context is cleaned up
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
except Exception:
pass
# Stop timers and re-raise the exception
stop_active_timer()
start_idle_timer()
raise
# Update the COST_TRACKER with the running cost for accurate display
if hasattr(COST_TRACKER, "interaction_cost"):
COST_TRACKER.interaction_cost = current_estimated_cost
# Also update streaming context if available for live display
if streaming_context:
# For free models, don't add to the session total
if current_estimated_cost == 0:
session_total = getattr(
COST_TRACKER, "session_total_cost", 0.0
)
else:
session_total = (
getattr(COST_TRACKER, "session_total_cost", 0.0)
+ current_estimated_cost
)
updated_token_stats = {
"input_tokens": estimated_input_tokens,
"output_tokens": estimated_output_tokens,
"cost": current_estimated_cost,
"total_cost": session_total,
}
update_agent_streaming_content(
streaming_context, "", updated_token_stats
)
if not state.text_content_index_and_output:
# Initialize a content tracker for streaming text
state.text_content_index_and_output = (
0 if not state.refusal_content_index_and_output else 1,
ResponseOutputText(
text="",
type="output_text",
annotations=[],
),
)
# Start a new assistant message stream
assistant_item = ResponseOutputMessage(
id=FAKE_RESPONSES_ID,
content=[],
role="assistant",
type="message",
status="in_progress",
)
# Notify consumers of the start of a new output message + first content part
yield ResponseOutputItemAddedEvent(
item=assistant_item,
output_index=0,
type="response.output_item.added",
)
yield ResponseContentPartAddedEvent(
content_index=state.text_content_index_and_output[0],
item_id=FAKE_RESPONSES_ID,
output_index=0,
part=ResponseOutputText(
text="",
type="output_text",
annotations=[],
),
type="response.content_part.added",
)
# Emit the delta for this segment of content
yield ResponseTextDeltaEvent(
content_index=state.text_content_index_and_output[0],
delta=content,
item_id=FAKE_RESPONSES_ID,
output_index=0,
type="response.output_text.delta",
)
# Accumulate the text into the response part
state.text_content_index_and_output[1].text += content
# Handle refusals (model declines to answer)
refusal_content = None
if hasattr(delta, "refusal") and delta.refusal:
refusal_content = delta.refusal
elif isinstance(delta, dict) and "refusal" in delta and delta["refusal"]:
refusal_content = delta["refusal"]
if refusal_content:
if not state.refusal_content_index_and_output:
# Initialize a content tracker for streaming refusal text
state.refusal_content_index_and_output = (
0 if not state.text_content_index_and_output else 1,
ResponseOutputRefusal(refusal="", type="refusal"),
)
# Start a new assistant message if one doesn't exist yet (in-progress)
assistant_item = ResponseOutputMessage(
id=FAKE_RESPONSES_ID,
content=[],
role="assistant",
type="message",
status="in_progress",
)
# Notify downstream that assistant message + first content part are starting
yield ResponseOutputItemAddedEvent(
item=assistant_item,
output_index=0,
type="response.output_item.added",
)
yield ResponseContentPartAddedEvent(
content_index=state.refusal_content_index_and_output[0],
item_id=FAKE_RESPONSES_ID,
output_index=0,
part=ResponseOutputText(
text="",
type="output_text",
annotations=[],
),
type="response.content_part.added",
)
# Emit the delta for this segment of refusal
yield ResponseRefusalDeltaEvent(
content_index=state.refusal_content_index_and_output[0],
delta=refusal_content,
item_id=FAKE_RESPONSES_ID,
output_index=0,
type="response.refusal.delta",
)
# Accumulate the refusal string in the output part
state.refusal_content_index_and_output[1].refusal += refusal_content
# Handle tool calls
# Because we don't know the name of the function until the end of the stream, we'll
# save everything and yield events at the end
tool_calls = self._detect_and_format_function_calls(delta)
if tool_calls:
for tc_delta in tool_calls:
tc_index = (
tc_delta.index
if hasattr(tc_delta, "index")
else tc_delta.get("index", 0)
)
if tc_index not in state.function_calls:
state.function_calls[tc_index] = ResponseFunctionToolCall(
id=FAKE_RESPONSES_ID,
arguments="",
name="",
type="function_call",
call_id="",
)
tc_function = None
if hasattr(tc_delta, "function"):
tc_function = tc_delta.function
elif isinstance(tc_delta, dict) and "function" in tc_delta:
tc_function = tc_delta["function"]
if tc_function:
# Handle both object and dict formats
args = ""
if hasattr(tc_function, "arguments"):
args = tc_function.arguments or ""
elif (
isinstance(tc_function, dict) and "arguments" in tc_function
):
args = tc_function.get("arguments", "") or ""
name = ""
if hasattr(tc_function, "name"):
name = tc_function.name or ""
elif isinstance(tc_function, dict) and "name" in tc_function:
name = tc_function.get("name", "") or ""
state.function_calls[tc_index].arguments += args
state.function_calls[tc_index].name += name
# Handle call_id in both formats
call_id = ""
if hasattr(tc_delta, "id"):
call_id = tc_delta.id or ""
elif isinstance(tc_delta, dict) and "id" in tc_delta:
call_id = tc_delta.get("id", "") or ""
else:
# For Qwen models, generate a predictable ID if none is provided
if state.function_calls[tc_index].name:
# Generate a stable ID from the function name and arguments
call_id = f"call_{hashlib.md5(state.function_calls[tc_index].name.encode()).hexdigest()[:8]}"
state.function_calls[tc_index].call_id += call_id
# --- Accumulate tool call for message_history ---
# Only add if not already present (avoid duplicates in streaming)
# Handle empty arguments before storing
tool_args = state.function_calls[tc_index].arguments
if tool_args is None or (isinstance(tool_args, str) and tool_args.strip() == ""):
tool_args = "{}"
tool_call_msg = {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": state.function_calls[tc_index].call_id,
"type": "function",
"function": {
"name": state.function_calls[tc_index].name,
"arguments": tool_args,
},
}
],
}
# Only add if not already in streamed_tool_calls
if tool_call_msg not in streamed_tool_calls:
streamed_tool_calls.append(tool_call_msg)
# Don't add to message history here - wait for tool output
# to add both tool call and response atomically
# NEW: Display tool call immediately when detected in streaming mode
# But only if it has complete arguments and name
if (
state.function_calls[tc_index].name
and state.function_calls[tc_index].arguments
and state.function_calls[tc_index].call_id
):
# First, finish any existing streaming context if it exists
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
streaming_context = None
except Exception:
pass
# Create a message-like object for displaying the function call
tool_msg = type(
"ToolCallStreamDisplay",
(),
{
"content": None,
"tool_calls": [
type(
"ToolCallDetail",
(),
{
"function": type(
"FunctionDetail",
(),
{
"name": state.function_calls[
tc_index
].name,
"arguments": state.function_calls[
tc_index
].arguments,
},
),
"id": state.function_calls[
tc_index
].call_id,
"type": "function",
},
)
],
},
)
# Display the tool call during streaming
cli_print_agent_messages(
agent_name=getattr(self, "agent_name", "Agent"),
message=tool_msg,
counter=getattr(self, "interaction_counter", 0),
model=str(self.model),
debug=False,
interaction_input_tokens=estimated_input_tokens,
interaction_output_tokens=estimated_output_tokens,
interaction_reasoning_tokens=0, # Not available during streaming yet
total_input_tokens=getattr(
self, "total_input_tokens", 0
)
+ estimated_input_tokens,
total_output_tokens=getattr(
self, "total_output_tokens", 0
)
+ estimated_output_tokens,
total_reasoning_tokens=getattr(
self, "total_reasoning_tokens", 0
),
interaction_cost=None,
total_cost=None,
tool_output=None, # Will be shown once tool is executed
suppress_empty=True, # Prevent empty panels
)
# Set flag to suppress final output to avoid duplication
self.suppress_final_output = True
except KeyboardInterrupt:
# Handle interruption during streaming
stream_interrupted = True
print("\n[Streaming interrupted by user]", file=sys.stderr)
# Let the exception propagate after cleanup
raise
except Exception as e:
# Handle other exceptions during streaming
logger.error(f"Error during streaming: {e}")
if "token" in str(e).lower() or "limit" in str(e).lower():
print("\n📏 Token limit exceeded - Response truncated")
raise
# Special handling for Ollama - check if accumulated text contains a valid function call
if is_ollama and ollama_full_content and len(state.function_calls) == 0:
# Look for JSON object that might be a function call
try:
# Try to extract a JSON object from the content
json_start = ollama_full_content.find("{")
json_end = ollama_full_content.rfind("}") + 1
if json_start >= 0 and json_end > json_start:
json_str = ollama_full_content[json_start:json_end]
# Try to parse the JSON
parsed = json.loads(json_str)
# Check if it looks like a function call
if "name" in parsed and "arguments" in parsed:
logger.debug(
f"Found valid function call in Ollama output: {json_str}"
)
# Create a tool call ID
tool_call_id = f"call_{hashlib.md5((parsed['name'] + str(time.time())).encode()).hexdigest()[:8]}"
# Ensure arguments is a valid JSON string
arguments_str = ""
if isinstance(parsed["arguments"], dict):
# Remove 'ctf' field if it exists
if "ctf" in parsed["arguments"]:
del parsed["arguments"]["ctf"]
arguments_str = json.dumps(parsed["arguments"])
elif isinstance(parsed["arguments"], str):
# If it's already a string, check if it's valid JSON
try:
# Try parsing to validate and remove 'ctf' if present
args_dict = json.loads(parsed["arguments"])
if isinstance(args_dict, dict) and "ctf" in args_dict:
del args_dict["ctf"]
arguments_str = json.dumps(args_dict)
except:
# If not valid JSON, encode it as a JSON string
arguments_str = json.dumps(parsed["arguments"])
else:
# For any other type, convert to string and then JSON
arguments_str = json.dumps(str(parsed["arguments"]))
# Add it to our function_calls state
state.function_calls[0] = ResponseFunctionToolCall(
id=FAKE_RESPONSES_ID,
arguments=arguments_str,
name=parsed["name"],
type="function_call",
call_id=tool_call_id[:40],
)
# Display the tool call in CLI
try:
# First, finish any existing streaming context if it exists
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
streaming_context = None
except Exception:
pass
# Create a message-like object to display the function call
tool_msg = type(
"ToolCallWrapper",
(),
{
"content": None,
"tool_calls": [
type(
"ToolCallDetail",
(),
{
"function": type(
"FunctionDetail",
(),
{
"name": parsed["name"],
"arguments": arguments_str,
},
),
"id": tool_call_id[:40],
"type": "function",
},
)
],
},
)
# Print the tool call using the CLI utility
cli_print_agent_messages(
agent_name=getattr(self, "agent_name", "Agent"),
message=tool_msg,
counter=getattr(self, "interaction_counter", 0),
model=str(self.model),
debug=False,
interaction_input_tokens=estimated_input_tokens,
interaction_output_tokens=estimated_output_tokens,
interaction_reasoning_tokens=0, # Not available for Ollama
total_input_tokens=getattr(self, "total_input_tokens", 0)
+ estimated_input_tokens,
total_output_tokens=getattr(self, "total_output_tokens", 0)
+ estimated_output_tokens,
total_reasoning_tokens=getattr(
self, "total_reasoning_tokens", 0
),
interaction_cost=None,
total_cost=None,
tool_output=None, # Will be shown once the tool is executed
suppress_empty=True, # Suppress empty panels during streaming
)
# Set flag to suppress final output to avoid duplication
self.suppress_final_output = True
except Exception as e:
# Silently log the error - don't disrupt the flow
logger.debug(f"Display error (non-critical): {e}")
# Add to message history
tool_call_msg = {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call_id,
"type": "function",
"function": {
"name": parsed["name"],
"arguments": arguments_str,
},
}
],
}
streamed_tool_calls.append(tool_call_msg)
# Don't add to message history here - wait for tool output
# to add both tool call and response atomically
logger.debug(
f"Added function call: {parsed['name']} with args: {arguments_str}"
)
except Exception:
pass
function_call_starting_index = 0
if state.text_content_index_and_output:
function_call_starting_index += 1
# Send end event for this content part
yield ResponseContentPartDoneEvent(
content_index=state.text_content_index_and_output[0],
item_id=FAKE_RESPONSES_ID,
output_index=0,
part=state.text_content_index_and_output[1],
type="response.content_part.done",
)
if state.refusal_content_index_and_output:
function_call_starting_index += 1
# Send end event for this content part
yield ResponseContentPartDoneEvent(
content_index=state.refusal_content_index_and_output[0],
item_id=FAKE_RESPONSES_ID,
output_index=0,
part=state.refusal_content_index_and_output[1],
type="response.content_part.done",
)
# Actually send events for the function calls
for function_call in state.function_calls.values():
# First, a ResponseOutputItemAdded for the function call
yield ResponseOutputItemAddedEvent(
item=ResponseFunctionToolCall(
id=FAKE_RESPONSES_ID,
call_id=function_call.call_id[:40],
arguments=function_call.arguments,
name=function_call.name,
type="function_call",
),
output_index=function_call_starting_index,
type="response.output_item.added",
)
# Then, yield the args
yield ResponseFunctionCallArgumentsDeltaEvent(
delta=function_call.arguments,
item_id=FAKE_RESPONSES_ID,
output_index=function_call_starting_index,
type="response.function_call_arguments.delta",
)
# Finally, the ResponseOutputItemDone
yield ResponseOutputItemDoneEvent(
item=ResponseFunctionToolCall(
id=FAKE_RESPONSES_ID,
call_id=function_call.call_id[:40],
arguments=function_call.arguments,
name=function_call.name,
type="function_call",
),
output_index=function_call_starting_index,
type="response.output_item.done",
)
# Finally, send the Response completed event
outputs: list[ResponseOutputItem] = []
if state.text_content_index_and_output or state.refusal_content_index_and_output:
assistant_msg = ResponseOutputMessage(
id=FAKE_RESPONSES_ID,
content=[],
role="assistant",
type="message",
status="completed",
)
if state.text_content_index_and_output:
assistant_msg.content.append(state.text_content_index_and_output[1])
if state.refusal_content_index_and_output:
assistant_msg.content.append(state.refusal_content_index_and_output[1])
outputs.append(assistant_msg)
# send a ResponseOutputItemDone for the assistant message
yield ResponseOutputItemDoneEvent(
item=assistant_msg,
output_index=0,
type="response.output_item.done",
)
for function_call in state.function_calls.values():
outputs.append(function_call)
final_response = response.model_copy()
final_response.output = outputs
# Get final token counts using consistent method
input_tokens = estimated_input_tokens
output_tokens = estimated_output_tokens
# Use API token counts if available and reasonable
if usage and hasattr(usage, "prompt_tokens") and usage.prompt_tokens > 0:
input_tokens = usage.prompt_tokens
if usage and hasattr(usage, "completion_tokens") and usage.completion_tokens > 0:
output_tokens = usage.completion_tokens
# Create a proper usage object with our token counts
final_response.usage = CustomResponseUsage(
input_tokens=input_tokens,
output_tokens=output_tokens,
total_tokens=input_tokens + output_tokens,
output_tokens_details=OutputTokensDetails(
reasoning_tokens=usage.completion_tokens_details.reasoning_tokens
if usage
and hasattr(usage, "completion_tokens_details")
and usage.completion_tokens_details
and hasattr(usage.completion_tokens_details, "reasoning_tokens")
and usage.completion_tokens_details.reasoning_tokens
else 0
),
input_tokens_details={
"prompt_tokens": input_tokens,
"cached_tokens": usage.prompt_tokens_details.cached_tokens
if usage
and hasattr(usage, "prompt_tokens_details")
and usage.prompt_tokens_details
and hasattr(usage.prompt_tokens_details, "cached_tokens")
and usage.prompt_tokens_details.cached_tokens
else 0,
},
)
yield ResponseCompletedEvent(
response=final_response,
type="response.completed",
)
# Update token totals for CLI display
if final_response.usage:
# Always update the total counters with the best available counts
self.total_input_tokens += final_response.usage.input_tokens
self.total_output_tokens += final_response.usage.output_tokens
if final_response.usage.output_tokens_details and hasattr(
final_response.usage.output_tokens_details, "reasoning_tokens"
):
self.total_reasoning_tokens += (
final_response.usage.output_tokens_details.reasoning_tokens
)
# Prepare final statistics for display
interaction_input = final_response.usage.input_tokens if final_response.usage else 0
interaction_output = (
final_response.usage.output_tokens if final_response.usage else 0
)
total_input = getattr(self, "total_input_tokens", 0)
total_output = getattr(self, "total_output_tokens", 0)
# Calculate costs for this model
model_name = str(self.model)
interaction_cost = calculate_model_cost(
model_name, interaction_input, interaction_output
)
# Get the previous total cost and add this interaction's cost
# Don't recalculate cost for all tokens - that causes double-counting
previous_total = getattr(COST_TRACKER, "session_total_cost", 0.0)
total_cost = previous_total + interaction_cost
# If interaction cost is zero, this is a free model
if interaction_cost == 0:
# For free models, keep existing total and ensure cost tracking system knows it's free
total_cost = getattr(COST_TRACKER, "session_total_cost", 0.0)
if hasattr(COST_TRACKER, "reset_cost_for_local_model"):
COST_TRACKER.reset_cost_for_local_model(model_name)
# Explicit conversion to float with fallback to ensure they're never None or 0
interaction_cost = float(interaction_cost if interaction_cost is not None else 0.0)
total_cost = float(total_cost if total_cost is not None else 0.0)
# Process costs through COST_TRACKER only once per interaction
if interaction_cost > 0.0:
# Check price limit before processing the new cost
if hasattr(COST_TRACKER, "check_price_limit"):
try:
COST_TRACKER.check_price_limit(interaction_cost)
except Exception:
# Ensure streaming context is cleaned up
if streaming_context:
try:
finish_agent_streaming(streaming_context, None)
except Exception:
pass
# Stop timers and re-raise the exception
stop_active_timer()
start_idle_timer()
raise
# Process the interaction cost (updates internal tracking)
COST_TRACKER.process_interaction_cost(
model_name,
interaction_input,
interaction_output,
final_response.usage.output_tokens_details.reasoning_tokens
if final_response.usage
and final_response.usage.output_tokens_details
and hasattr(final_response.usage.output_tokens_details, "reasoning_tokens")
else 0,
interaction_cost
)
# Process the total cost (updates session total correctly)
total_cost = COST_TRACKER.process_total_cost(
model_name,
total_input,
total_output,
getattr(self, "total_reasoning_tokens", 0),
None # Let it calculate from tokens
)
# Track usage globally
GLOBAL_USAGE_TRACKER.track_usage(
model_name=model_name,
input_tokens=interaction_input,
output_tokens=interaction_output,
cost=interaction_cost,
agent_name=self.agent_name
)
else:
# For free models, still track token usage
GLOBAL_USAGE_TRACKER.track_usage(
model_name=model_name,
input_tokens=interaction_input,
output_tokens=interaction_output,
cost=0.0,
agent_name=self.agent_name
)
# Store the total cost for future recording
self.total_cost = total_cost
# Create final stats with explicit type conversion for all values
final_stats = {
"interaction_input_tokens": int(interaction_input),
"interaction_output_tokens": int(interaction_output),
"interaction_reasoning_tokens": int(
final_response.usage.output_tokens_details.reasoning_tokens
if final_response.usage
and final_response.usage.output_tokens_details
and hasattr(final_response.usage.output_tokens_details, "reasoning_tokens")
else 0
),
"total_input_tokens": int(total_input),
"total_output_tokens": int(total_output),
"total_reasoning_tokens": int(getattr(self, "total_reasoning_tokens", 0)),
"interaction_cost": float(interaction_cost),
"total_cost": float(total_cost),
}
# At the end of streaming, finish the streaming context if we were using it
if streaming_context:
# Create a direct copy of the costs to ensure they remain as floats
direct_stats = final_stats.copy()
direct_stats["interaction_cost"] = float(interaction_cost)
direct_stats["total_cost"] = float(total_cost)
# Use the direct copy with guaranteed float costs
finish_agent_streaming(streaming_context, direct_stats)
streaming_context = None
# Removed extra newline after streaming completes to avoid blank lines
pass
# Finish Claude thinking display if it was active
if thinking_context:
from cai.util import finish_claude_thinking_display
finish_claude_thinking_display(thinking_context)
# Note: Content is now displayed during streaming, no need to show it again here
if tracing.include_data():
span_generation.span_data.output = [final_response.model_dump()]
span_generation.span_data.usage = {
"input_tokens": input_tokens,
"output_tokens": output_tokens,
}
# --- DEFERRED: Tool calls are no longer added immediately ---
# Store pending tool calls but don't add to history yet
if not hasattr(self, "_pending_tool_calls"):
self._pending_tool_calls = {}
for tool_call_msg in streamed_tool_calls:
# Extract tool call ID from the message
if tool_call_msg.get("tool_calls"):
for tc in tool_call_msg["tool_calls"]:
self._pending_tool_calls[tc["id"]] = tool_call_msg
# Log the assistant tool call message if any tool calls were collected
if streamed_tool_calls:
tool_calls_list = []
for tool_call_msg in streamed_tool_calls:
for tool_call in tool_call_msg.get("tool_calls", []):
tool_calls_list.append(tool_call)
self.logger.log_assistant_message(None, tool_calls_list)
# Always log text content if it exists, regardless of suppress_final_output
# The suppress_final_output flag is only for preventing duplicate tool call display
if (
state.text_content_index_and_output
and state.text_content_index_and_output[1].text
):
asst_msg = {
"role": "assistant",
"content": state.text_content_index_and_output[1].text,
}
self.add_to_message_history(asst_msg)
# Log the assistant message
self.logger.log_assistant_message(state.text_content_index_and_output[1].text)
# Reset the suppress flag for future requests
self.suppress_final_output = False
# Log the complete response
self.logger.rec_training_data(
{
"model": str(self.model),
"messages": converted_messages,
"stream": True,
"tools": [t.params_json_schema for t in tools] if tools else [],
"tool_choice": model_settings.tool_choice,
},
final_response,
self.total_cost,
self.agent_name,
)
# Stop active timer and start idle timer when streaming is complete
stop_active_timer()
start_idle_timer()
except KeyboardInterrupt:
# Handle keyboard interruption specifically
stream_interrupted = True
# Ensure message history consistency by adding synthetic tool results
# for any tool calls that were added but don't have corresponding results
try:
# Find all tool calls in recent assistant messages
orphaned_tool_calls = []
for msg in reversed(self.message_history[-10:]): # Check recent messages
if msg.get("role") == "assistant" and msg.get("tool_calls"):
for tool_call in msg["tool_calls"]:
call_id = tool_call.get("id")
if call_id:
# Check if this tool call has a corresponding tool result
has_result = any(
m.get("role") == "tool" and m.get("tool_call_id") == call_id
for m in self.message_history
)
if not has_result:
orphaned_tool_calls.append((call_id, tool_call))
# Add synthetic tool results for orphaned tool calls
for call_id, tool_call in orphaned_tool_calls:
tool_response_msg = {
"role": "tool",
"tool_call_id": call_id,
"content": "Tool execution interrupted"
}
self.add_to_message_history(tool_response_msg)
except Exception as cleanup_error:
# Don't let cleanup errors mask the original KeyboardInterrupt
logger.debug(f"Error during interrupt cleanup: {cleanup_error}")
# Make sure to clean up and re-raise
raise
except Exception as e:
# Handle other exceptions
logger.error(f"Error in stream_response: {e}")
raise
finally:
# Always clean up resources
# This block executes whether the try block succeeds, fails, or is interrupted
# Clean up streaming context
if streaming_context:
try:
# Check if we need to force stop the streaming panel
if streaming_context.get("is_started", False) and streaming_context.get("live"):
streaming_context["live"].stop()
# Remove from active streaming contexts
if hasattr(create_agent_streaming_context, "_active_streaming"):
for key, value in list(
create_agent_streaming_context._active_streaming.items()
):
if value is streaming_context:
del create_agent_streaming_context._active_streaming[key]
break
except Exception as cleanup_error:
logger.debug(f"Error cleaning up streaming context: {cleanup_error}")
# Clean up thinking context
if thinking_context:
try:
# Force finish the thinking display
from cai.util import finish_claude_thinking_display
finish_claude_thinking_display(thinking_context)
except Exception as cleanup_error:
logger.debug(f"Error cleaning up thinking context: {cleanup_error}")
# Clean up any live streaming panels
if hasattr(cli_print_tool_output, "_streaming_sessions"):
# Find any sessions related to this stream
for call_id in list(cli_print_tool_output._streaming_sessions.keys()):
if call_id in _LIVE_STREAMING_PANELS:
try:
live = _LIVE_STREAMING_PANELS[call_id]
live.stop()
del _LIVE_STREAMING_PANELS[call_id]
except Exception:
pass
# Stop active timer and start idle timer
try:
stop_active_timer()
start_idle_timer()
except Exception:
pass
# Stream cleanup completed
@overload
async def _fetch_response(
self,
system_instructions: str | None,
input: str | list[TResponseInputItem],
model_settings: ModelSettings,
tools: list[Tool],
output_schema: AgentOutputSchema | None,
handoffs: list[Handoff],
span: Span[GenerationSpanData],
tracing: ModelTracing,
stream: Literal[True],
) -> tuple[Response, AsyncStream[ChatCompletionChunk]]: ...
@overload
async def _fetch_response(
self,
system_instructions: str | None,
input: str | list[TResponseInputItem],
model_settings: ModelSettings,
tools: list[Tool],
output_schema: AgentOutputSchema | None,
handoffs: list[Handoff],
span: Span[GenerationSpanData],
tracing: ModelTracing,
stream: Literal[False],
) -> ChatCompletion: ...
async def _fetch_response(
self,
system_instructions: str | None,
input: str | list[TResponseInputItem],
model_settings: ModelSettings,
tools: list[Tool],
output_schema: AgentOutputSchema | None,
handoffs: list[Handoff],
span: Span[GenerationSpanData],
tracing: ModelTracing,
stream: bool = False,
) -> ChatCompletion | tuple[Response, AsyncStream[ChatCompletionChunk]]:
# start by re-fetching self.is_ollama
self.is_ollama = os.getenv("OLLAMA") is not None and os.getenv("OLLAMA").lower() == "true"
# IMPORTANT: Include existing message history for context
converted_messages = []
# First, add all existing messages from history
if self.message_history:
for msg in self.message_history:
msg_copy = msg.copy() # Use copy to avoid modifying original
# Remove any existing cache_control to avoid exceeding the 4-block limit
if "cache_control" in msg_copy:
del msg_copy["cache_control"]
converted_messages.append(msg_copy)
# Then convert and add the new input
new_messages = self._converter.items_to_messages(input, model_instance=self)
converted_messages.extend(new_messages)
if system_instructions:
# Check if we already have a system message
has_system = any(msg.get("role") == "system" for msg in converted_messages)
if not has_system:
converted_messages.insert(
0,
{
"content": system_instructions,
"role": "system",
},
)
# Add support for prompt caching for claude (not automatically applied)
# Gemini supports it too
# https://www.anthropic.com/news/token-saving-updates
# Maximize cache efficiency by using up to 4 cache_control blocks
if (str(self.model).startswith("claude") or "gemini" in str(self.model)) and len(
converted_messages
) > 0:
# Strategy: Cache the most valuable messages for maximum savings
# 1. System message (always first priority)
# 2. Long user messages (high token count)
# 3. Assistant messages with tool calls (complex context)
# 4. Recent context (last message)
cache_candidates = []
# Always cache system message if present
for i, msg in enumerate(converted_messages):
if msg.get("role") == "system":
cache_candidates.append((i, len(str(msg.get("content", ""))), "system"))
break
# Find long user messages and assistant messages with tool calls
for i, msg in enumerate(converted_messages):
content_len = len(str(msg.get("content", "")))
role = msg.get("role")
if role == "user" and content_len > 500: # Long user messages
cache_candidates.append((i, content_len, "user"))
elif role == "assistant" and msg.get("tool_calls"): # Tool calls
cache_candidates.append(
(i, content_len + 200, "assistant_tools")
) # Bonus for tool calls
# Always consider the last message for recent context
if len(converted_messages) > 1:
last_idx = len(converted_messages) - 1
last_msg = converted_messages[last_idx]
last_content_len = len(str(last_msg.get("content", "")))
cache_candidates.append((last_idx, last_content_len, "recent"))
# Sort by value (content length) and select top 4 unique indices
cache_candidates.sort(key=lambda x: x[1], reverse=True)
selected_indices = []
for idx, _, msg_type in cache_candidates:
if idx not in selected_indices:
selected_indices.append(idx)
if len(selected_indices) >= 4: # Max 4 cache blocks
break
# Apply cache_control to selected messages
for idx in selected_indices:
msg_copy = converted_messages[idx].copy()
msg_copy["cache_control"] = {"type": "ephemeral"}
converted_messages[idx] = msg_copy
if tracing.include_data():
span.span_data.input = converted_messages
# IMPORTANT: Always sanitize the message list to prevent tool call errors
# This is critical to fix common errors with tool/assistant sequences
try:
from cai.util import fix_message_list
prev_length = len(converted_messages)
converted_messages = fix_message_list(converted_messages)
new_length = len(converted_messages)
# Log if the message list was changed significantly
if new_length != prev_length:
logger.debug(f"Message list was fixed: {prev_length} -> {new_length} messages")
except Exception:
pass
parallel_tool_calls = (
True if model_settings.parallel_tool_calls and tools and len(tools) > 0 else NOT_GIVEN
)
tool_choice = self._converter.convert_tool_choice(model_settings.tool_choice)
response_format = self._converter.convert_response_format(output_schema)
converted_tools = [ToolConverter.to_openai(tool) for tool in tools] if tools else []
for handoff in handoffs:
converted_tools.append(ToolConverter.convert_handoff_tool(handoff))
if _debug.DONT_LOG_MODEL_DATA:
logger.debug("Calling LLM")
else:
logger.debug(
f"{json.dumps(converted_messages, indent=2)}\n"
f"Tools:\n{json.dumps(converted_tools, indent=2)}\n"
f"Stream: {stream}\n"
f"Tool choice: {tool_choice}\n"
f"Response format: {response_format}\n"
f"Using OLLAMA: {self.is_ollama}\n"
)
# Use NOT_GIVEN for store if not explicitly set to avoid compatibility issues
store = self._non_null_or_not_given(model_settings.store)
# Check if we should use the agent's model instead of self.model
# This prioritizes the model from Agent when available
agent_model = None
if hasattr(model_settings, "agent_model") and model_settings.agent_model:
agent_model = model_settings.agent_model
logger.debug(f"Using agent model: {agent_model} instead of {self.model}")
# Prepare kwargs for the API call
kwargs = {
"model": agent_model if agent_model else self.model,
"messages": converted_messages,
"tools": converted_tools or NOT_GIVEN,
"temperature": self._non_null_or_not_given(model_settings.temperature),
"top_p": self._non_null_or_not_given(model_settings.top_p),
"frequency_penalty": self._non_null_or_not_given(model_settings.frequency_penalty),
"presence_penalty": self._non_null_or_not_given(model_settings.presence_penalty),
"max_tokens": self._non_null_or_not_given(model_settings.max_tokens),
"tool_choice": tool_choice,
"response_format": response_format,
"parallel_tool_calls": parallel_tool_calls,
"stream": stream,
"stream_options": {"include_usage": True} if stream else NOT_GIVEN,
"store": store,
"extra_headers": _HEADERS,
}
# Determine provider based on model string
model_str = str(kwargs["model"]).lower()
if "alias" in model_str:
kwargs["api_base"] = "http://api.aliasrobotics.com:666/"
kwargs["custom_llm_provider"] = "openai"
kwargs["api_key"] = os.getenv("ALIAS_API_KEY", "REDACTED_ALIAS_KEY")
elif "/" in model_str:
# Handle provider/model format
provider = model_str.split("/")[0]
# Apply provider-specific configurations
if provider == "deepseek":
litellm.drop_params = True
kwargs.pop("parallel_tool_calls", None)
kwargs.pop("store", None) # DeepSeek doesn't support store parameter
# Remove tool_choice if no tools are specified
if not converted_tools:
kwargs.pop("tool_choice", None)
# Add reasoning support for DeepSeek
# DeepSeek supports reasoning_effort parameter
if hasattr(model_settings, "reasoning_effort") and model_settings.reasoning_effort:
kwargs["reasoning_effort"] = model_settings.reasoning_effort
else:
# Default to "low" reasoning effort if model supports it
kwargs["reasoning_effort"] = "low"
elif provider == "claude" or "claude" in model_str:
litellm.drop_params = True
kwargs.pop("store", None)
kwargs.pop(
"parallel_tool_calls", None
) # Claude doesn't support parallel tool calls
# Remove tool_choice if no tools are specified
if not converted_tools:
kwargs.pop("tool_choice", None)
# Add extended reasoning support for Claude models
# Supports Claude 3.7, Claude 4, and any model with "thinking" in the name
has_reasoning_capability = (
"thinking" in model_str
or
# Claude 4 models support reasoning
"-4-" in model_str
or "sonnet-4" in model_str
or "haiku-4" in model_str
or "opus-4" in model_str
or "3.7" in model_str
)
if has_reasoning_capability:
# Clean the model name by removing "thinking" before sending to API
clean_model = kwargs["model"]
if isinstance(clean_model, str) and "thinking" in clean_model.lower():
# Remove "thinking" and clean up any extra spaces/separators
clean_model = re.sub(
r"[_-]?thinking[_-]?", "", clean_model, flags=re.IGNORECASE
)
clean_model = re.sub(
r"[-_]{2,}", "-", clean_model
) # Clean up multiple separators
clean_model = clean_model.strip(
"-_"
) # Clean up leading/trailing separators
kwargs["model"] = clean_model
# Check if message history is compatible with reasoning
messages = kwargs.get("messages", [])
is_compatible = _check_reasoning_compatibility(messages)
if is_compatible:
kwargs["reasoning_effort"] = (
"low" # Use reasoning_effort instead of thinking
)
elif provider == "gemini":
kwargs.pop("parallel_tool_calls", None)
# Add any specific gemini settings if needed
else:
# Handle models without provider prefix
if "claude" in model_str or "anthropic" in model_str:
litellm.drop_params = True
# Remove parameters that Anthropic doesn't support
kwargs.pop("store", None)
kwargs.pop("parallel_tool_calls", None)
# Remove tool_choice if no tools are specified
if not converted_tools:
kwargs.pop("tool_choice", None)
# Add extended reasoning support for Claude models
# Supports Claude 3.7, Claude 4, and any model with "thinking" in the name
has_reasoning_capability = "thinking" in model_str
if has_reasoning_capability:
# Clean the model name by removing "thinking" before sending to API
clean_model = kwargs["model"]
if isinstance(clean_model, str) and "thinking" in clean_model.lower():
# Remove "thinking" and clean up any extra spaces/separators
clean_model = re.sub(
r"[_-]?thinking[_-]?", "", clean_model, flags=re.IGNORECASE
)
clean_model = re.sub(
r"[-_]{2,}", "-", clean_model
) # Clean up multiple separators
clean_model = clean_model.strip(
"-_"
) # Clean up leading/trailing separators
kwargs["model"] = clean_model
# Check if message history is compatible with reasoning
messages = kwargs.get("messages", [])
is_compatible = _check_reasoning_compatibility(messages)
if is_compatible:
kwargs["reasoning_effort"] = (
"low" # Use reasoning_effort instead of thinking
)
elif "gemini" in model_str:
kwargs.pop("parallel_tool_calls", None)
elif "qwen" in model_str or ":" in model_str:
# Handle Ollama-served models with custom formats (e.g., alias0)
# These typically need the Ollama provider
litellm.drop_params = True
kwargs.pop("parallel_tool_calls", None)
kwargs.pop("store", None) # Ollama doesn't support store parameter
# These models may not support certain parameters
if not converted_tools:
kwargs.pop("tool_choice", None)
# Don't add custom_llm_provider here to avoid duplication with Ollama provider
if self.is_ollama:
# Clean kwargs for ollama to avoid parameter conflicts
for param in ["custom_llm_provider"]:
kwargs.pop(param, None)
elif any(x in model_str for x in ["o1", "o3", "o4"]):
# Handle OpenAI reasoning models (o1, o3, o4)
kwargs.pop("parallel_tool_calls", None)
# Add reasoning effort if provided
if hasattr(model_settings, "reasoning_effort"):
kwargs["reasoning_effort"] = model_settings.reasoning_effort
# Filter out NotGiven values to avoid JSON serialization issues
filtered_kwargs = {}
for key, value in kwargs.items():
if value is not NOT_GIVEN:
filtered_kwargs[key] = value
kwargs = filtered_kwargs
# Add retry logic for rate limits
max_retries = 3
retry_count = 0
while retry_count < max_retries:
try:
if self.is_ollama:
return await self._fetch_response_litellm_ollama(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
else:
return await self._fetch_response_litellm_openai(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
except litellm.exceptions.RateLimitError as e:
retry_count += 1
if retry_count >= max_retries:
print(f"\n❌ Rate limit exceeded after {max_retries} retries")
raise
print(f"\n⏳ Rate limit reached - Too many requests (attempt {retry_count}/{max_retries})")
# Try to extract retry delay from error response or use default
retry_delay = 60 # Default delay in seconds
try:
# Extract the JSON part from the error message
json_str = str(e.message).split("VertexAIException - ")[-1]
error_details = json.loads(json_str)
retry_info = next(
(
detail
for detail in error_details.get("error", {}).get("details", [])
if detail.get("@type") == "type.googleapis.com/google.rpc.RetryInfo"
),
None,
)
if retry_info and "retryDelay" in retry_info:
retry_delay = int(retry_info["retryDelay"].rstrip("s"))
except Exception:
# Try other common formats
import re
error_str = str(e)
# Look for "Retry-After" header or similar patterns
retry_match = re.search(r'retry[_-]?after[:\s]+(\d+)', error_str, re.IGNORECASE)
if retry_match:
retry_delay = int(retry_match.group(1))
# Look for "wait X seconds" patterns
elif wait_match := re.search(r'wait\s+(\d+)\s+seconds?', error_str, re.IGNORECASE):
retry_delay = int(wait_match.group(1))
# Look for explicit retry delay mentions
elif delay_match := re.search(r'retry\s+in\s+(\d+)\s+seconds?', error_str, re.IGNORECASE):
retry_delay = int(delay_match.group(1))
# Use exponential backoff with jitter if no explicit delay found
if retry_count > 1 and retry_delay == 60:
import random
retry_delay = min(300, retry_delay * retry_count) + random.randint(0, 10)
print(f"💤 Waiting {retry_delay}s before retry... (Rate limit protection)")
await asyncio.sleep(retry_delay) # Use async sleep instead of time.sleep
continue # Retry the request
except litellm.exceptions.BadRequestError as e:
error_msg = str(e)
# Handle Claude reasoning/thinking compatibility errors
if (
"Expected `thinking` or `redacted_thinking`, but found `text`" in error_msg
or "When `thinking` is enabled, a final `assistant` message must start with a thinking block"
in error_msg
):
# Retry without reasoning_effort
retry_kwargs = kwargs.copy()
retry_kwargs.pop("reasoning_effort", None)
try:
if stream:
response = Response(
id=FAKE_RESPONSES_ID,
created_at=time.time(),
model=self.model,
object="response",
output=[],
tool_choice="auto"
if tool_choice is None or tool_choice == NOT_GIVEN
else cast(Literal["auto", "required", "none"], tool_choice),
top_p=model_settings.top_p,
temperature=model_settings.temperature,
tools=[],
parallel_tool_calls=parallel_tool_calls or False,
)
stream_obj = await litellm.acompletion(**retry_kwargs)
return response, stream_obj
else:
ret = await litellm.acompletion(**retry_kwargs)
return ret
except Exception:
# If retry also fails, raise the original error
raise e
# print(color("BadRequestError encountered: " + str(e), fg="yellow"))
if "LLM Provider NOT provided" in str(e):
model_str = str(self.model).lower()
provider = None
is_qwen = "qwen" in model_str or ":" in model_str
# Special handling for Qwen models
if is_qwen:
try:
# Use the specialized Qwen approach first
return await self._fetch_response_litellm_ollama(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
except Exception as qwen_e:
print(qwen_e)
# If that fails, try our direct OpenAI approach
qwen_params = kwargs.copy()
qwen_params["api_base"] = get_ollama_api_base()
qwen_params["custom_llm_provider"] = "openai" # Use openai provider
# Make sure tools are passed
if "tools" in kwargs and kwargs["tools"]:
qwen_params["tools"] = kwargs["tools"]
if "tool_choice" in kwargs and kwargs["tool_choice"] is not NOT_GIVEN:
qwen_params["tool_choice"] = kwargs["tool_choice"]
try:
if stream:
# Streaming case
response = Response(
id=FAKE_RESPONSES_ID,
created_at=time.time(),
model=self.model,
object="response",
output=[],
tool_choice="auto"
if tool_choice is None or tool_choice == NOT_GIVEN
else cast(Literal["auto", "required", "none"], tool_choice),
top_p=model_settings.top_p,
temperature=model_settings.temperature,
tools=[],
parallel_tool_calls=parallel_tool_calls or False,
)
stream_obj = await litellm.acompletion(**qwen_params)
return response, stream_obj
else:
# Non-streaming case
ret = await litellm.acompletion(**qwen_params)
return ret
except Exception as direct_e:
# All approaches failed, log and raise the original error
print(
f"All Qwen approaches failed. Original error: {str(e)}, Direct error: {str(direct_e)}"
)
raise e
# Try to detect provider from model string
if "/" in model_str:
provider = model_str.split("/")[0]
if provider:
# Add provider-specific settings based on detected provider
provider_kwargs = kwargs.copy()
if provider == "deepseek":
provider_kwargs["custom_llm_provider"] = "deepseek"
provider_kwargs.pop(
"store", None
) # DeepSeek doesn't support store parameter
provider_kwargs.pop(
"parallel_tool_calls", None
) # DeepSeek doesn't support parallel tool calls
# Add reasoning support for DeepSeek
if (
hasattr(model_settings, "reasoning_effort")
and model_settings.reasoning_effort
):
provider_kwargs["reasoning_effort"] = model_settings.reasoning_effort
else:
# Default to "low" reasoning effort
provider_kwargs["reasoning_effort"] = "low"
elif provider == "claude" or "claude" in model_str:
provider_kwargs["custom_llm_provider"] = "anthropic"
provider_kwargs.pop("store", None) # Claude doesn't support store parameter
provider_kwargs.pop(
"parallel_tool_calls", None
) # Claude doesn't support parallel tool calls
# Add extended reasoning support for Claude models
if "thinking" in model_str:
# Clean the model name by removing "thinking" before sending to API
clean_model = provider_kwargs["model"]
if isinstance(clean_model, str) and "thinking" in clean_model.lower():
# Remove "thinking" and clean up any extra spaces/separators
clean_model = re.sub(
r"[_-]?thinking[_-]?", "", clean_model, flags=re.IGNORECASE
)
clean_model = re.sub(
r"[-_]{2,}", "-", clean_model
) # Clean up multiple separators
clean_model = clean_model.strip(
"-_"
) # Clean up leading/trailing separators
provider_kwargs["model"] = clean_model
# Check if message history is compatible with reasoning
messages = provider_kwargs.get("messages", [])
is_compatible = _check_reasoning_compatibility(messages)
if is_compatible:
provider_kwargs["reasoning_effort"] = (
"low" # Use reasoning_effort instead of thinking
)
elif provider == "gemini":
provider_kwargs["custom_llm_provider"] = "gemini"
provider_kwargs.pop("store", None) # Gemini doesn't support store parameter
provider_kwargs.pop(
"parallel_tool_calls", None
) # Gemini doesn't support parallel tool calls
else:
# For unknown providers, try ollama as fallback
return await self._fetch_response_litellm_ollama(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
# Check for message sequence errors
if (
"An assistant message with 'tool_calls'" in str(e)
or "`tool_use` blocks must be followed by a user message with `tool_result`"
in str(e) # noqa: E501 # pylint: disable=C0301
or "`tool_use` ids were found without `tool_result` blocks immediately after"
in str(e) # noqa: E501 # pylint: disable=C0301
or "An assistant message with 'tool_calls' must be followed by tool messages"
in str(e)
or "messages with role 'tool' must be a response to a preceeding message with 'tool_calls'"
in str(e)
):
print("⚠️ Message sequence error - Tool calls and results are out of order")
# Use the pretty message history printer instead of the simple loop
try:
from cai.util import print_message_history
print("\n📋 Current message sequence:")
print_message_history(kwargs["messages"], title="Message History")
except ImportError:
# Fall back to simple printing if the function isn't available
print("\n📋 Current message sequence:")
for i, msg in enumerate(kwargs["messages"]):
role = msg.get("role", "unknown")
content_type = (
"text"
if isinstance(msg.get("content"), str)
else "list"
if isinstance(msg.get("content"), list)
else "None"
if msg.get("content") is None
else type(msg.get("content")).__name__
)
tool_calls = "with tool_calls" if msg.get("tool_calls") else ""
tool_call_id = (
f", tool_call_id: {msg.get('tool_call_id')}"
if msg.get("tool_call_id")
else ""
)
print(
f" [{i}] {role}{tool_call_id} (content: {content_type}) {tool_calls}"
)
# NOTE: EDGE CASE: Report Agent CTRL C error
#
# This fix CTRL-C error when message list is incomplete
# When a tool is not finished but the LLM generates a tool call
try:
from cai.util import fix_message_list
print("🔧 Auto-fixing message sequence...")
fixed_messages = fix_message_list(kwargs["messages"])
# Show the fixed messages if they're different
if fixed_messages != kwargs["messages"]:
try:
from cai.util import print_message_history
print_message_history(fixed_messages, title="Fixed Message Sequence")
except ImportError:
print("✅ Message sequence fixed successfully")
kwargs["messages"] = fixed_messages
except Exception:
pass
return await self._fetch_response_litellm_openai(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
# this captures an error related to the fact
# that the messages list contains an empty
# content position
if "expected a string, got null" in str(e):
print("⚠️ Empty content detected - Filling with placeholder")
# Fix for null content in messages
kwargs["messages"] = [
msg if msg.get("content") is not None else {**msg, "content": ""}
for msg in kwargs["messages"]
]
return await self._fetch_response_litellm_openai(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
# Handle Anthropic error for empty text content blocks
if "text content blocks must be non-empty" in str(
e
) or "cache_control cannot be set for empty text blocks" in str(e): # noqa
# Print the error message only once
print("⚠️ Empty text blocks detected - Adding placeholder content") if not self.empty_content_error_shown else None
self.empty_content_error_shown = True
# Fix for empty content in messages for Anthropic models
kwargs["messages"] = [
msg
if msg.get("content") not in [None, ""]
else {**msg, "content": "Empty content block"}
for msg in kwargs["messages"]
]
return await self._fetch_response_litellm_openai(
kwargs, model_settings, tool_choice, stream, parallel_tool_calls
)
# Check for Python formatting errors - NOT context errors
if "Cannot specify ',' with 's'" in str(e):
print("\n❌ Python formatting error - Not a context error")
print("⚠️ There's a bug in the code trying to format strings as numbers")
print(f"Error: {str(e)}")
raise
# Check for context length errors in BadRequestError
if (
"context_length_exceeded" in str(e)
or "prompt is too long" in str(e).lower()
or "maximum context length" in str(e).lower()
or "max_tokens" in str(e) and "exceeded" in str(e).lower()
or "too many tokens" in str(e).lower()
or "token limit" in str(e).lower()
):
print("\n📦 Context window exceeded - Message history too long")
# Try to extract token info from different error formats
import re
error_str = str(e)
# Pattern 1: "X tokens > Y maximum" (Anthropic)
match1 = re.search(r'(\d+)\s*tokens?\s*>\s*(\d+)\s*maximum', error_str)
# Pattern 2: "requested X tokens...maximum context length is Y" (OpenAI)
match2 = re.search(r'requested\s+(\d+)\s+tokens.*maximum.*?(\d+)', error_str)
# Pattern 3: "This model's maximum context length is X tokens, however you requested Y"
match3 = re.search(r'maximum context length is\s+(\d+).*requested\s+(\d+)', error_str)
if match1:
used_tokens = int(match1.group(1))
max_tokens = int(match1.group(2))
print(f"🎯 Actual: {used_tokens:,} / {max_tokens:,} tokens")
elif match2:
used_tokens = int(match2.group(1))
max_tokens = int(match2.group(2))
print(f"🎯 Requested: {used_tokens:,} tokens (max: {max_tokens:,})")
elif match3:
max_tokens = int(match3.group(1))
used_tokens = int(match3.group(2))
print(f"🎯 Requested: {used_tokens:,} tokens (max: {max_tokens:,})")
elif 'estimated_input_tokens' in locals():
print(f"📊 Estimated tokens: ~{estimated_input_tokens:,}")
# Get model's max tokens
model_max = self._get_model_max_tokens(str(self.model))
print(f"🎯 Model limit: {model_max:,} tokens")
print("\n💡 Quick fixes:")
print(" • /flush - Clear conversation history")
print(" • /compact - Manually compact context")
print(" • /model <larger-model> - Switch to model with more context")
raise
else:
raise e
async def _fetch_response_litellm_openai(
self,
kwargs: dict,
model_settings: ModelSettings,
tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven,
stream: bool,
parallel_tool_calls: bool,
) -> ChatCompletion | tuple[Response, AsyncStream[ChatCompletionChunk]]:
"""
Handle standard LiteLLM API calls for OpenAI and compatible models.
If a ContextWindowExceededError occurs due to a tool_call id being
too long, truncate all tool_call ids in the messages to 40 characters
and retry once silently.
"""
try:
if stream:
# Standard LiteLLM handling for streaming
ret = await litellm.acompletion(**kwargs)
stream_obj = await litellm.acompletion(**kwargs)
response = Response(
id=FAKE_RESPONSES_ID,
created_at=time.time(),
model=self.model,
object="response",
output=[],
tool_choice="auto"
if tool_choice is None or tool_choice == NOT_GIVEN
else cast(Literal["auto", "required", "none"], tool_choice),
top_p=model_settings.top_p,
temperature=model_settings.temperature,
tools=[],
parallel_tool_calls=parallel_tool_calls or False,
)
return response, stream_obj
else:
# Standard OpenAI handling for non-streaming
ret = await litellm.acompletion(**kwargs)
return ret
except Exception as e:
error_msg = str(e)
# Handle both OpenAI and Anthropic error messages for tool_call_id
if (
"string too long" in error_msg
or "Invalid 'messages" in error_msg
and "tool_call_id" in error_msg
and "maximum length" in error_msg
):
# Truncate all tool_call ids in all messages to 40 characters
messages = kwargs.get("messages", [])
for msg in messages:
# Truncate tool_call_id in the message itself if present
if (
"tool_call_id" in msg
and isinstance(msg["tool_call_id"], str)
and len(msg["tool_call_id"]) > 40
):
msg["tool_call_id"] = msg["tool_call_id"][:40]
# Truncate tool_call ids in tool_calls if present
if "tool_calls" in msg and isinstance(msg["tool_calls"], list):
for tool_call in msg["tool_calls"]:
if (
isinstance(tool_call, dict)
and "id" in tool_call
and isinstance(tool_call["id"], str)
and len(tool_call["id"]) > 40
):
tool_call["id"] = tool_call["id"][:40]
kwargs["messages"] = messages
# Retry once, silently
if stream:
ret = await litellm.acompletion(**kwargs)
stream_obj = await litellm.acompletion(**kwargs)
response = Response(
id=FAKE_RESPONSES_ID,
created_at=time.time(),
model=self.model,
object="response",
output=[],
tool_choice="auto"
if tool_choice is None or tool_choice == NOT_GIVEN
else cast(Literal["auto", "required", "none"], tool_choice),
top_p=model_settings.top_p,
temperature=model_settings.temperature,
tools=[],
parallel_tool_calls=parallel_tool_calls or False,
)
return response, stream_obj
else:
ret = await litellm.acompletion(**kwargs)
return ret
else:
raise
async def _fetch_response_litellm_ollama(
self,
kwargs: dict,
model_settings: ModelSettings,
tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven,
stream: bool,
parallel_tool_calls: bool,
) -> ChatCompletion | tuple[Response, AsyncStream[ChatCompletionChunk]]:
"""
Fetches a response from an Ollama or Qwen model using LiteLLM, ensuring
that the 'format' parameter is not set to a JSON string, which can cause
issues with the Ollama API.
Args:
kwargs (dict): Parameters for the completion request.
model_settings (ModelSettings): Model configuration.
tool_choice (ChatCompletionToolChoiceOptionParam | NotGiven): Tool choice.
stream (bool): Whether to stream the response.
parallel_tool_calls (bool): Whether to allow parallel tool calls.
Returns:
ChatCompletion or tuple[Response, AsyncStream[ChatCompletionChunk]]:
The completion response or a tuple for streaming.
"""
# Extract only supported parameters for Ollama
ollama_supported_params = {
"model": kwargs.get("model", ""),
"messages": kwargs.get("messages", []),
"stream": kwargs.get("stream", False),
}
# Add optional parameters if they exist and are not NOT_GIVEN
for param in ["temperature", "top_p", "max_tokens"]:
if param in kwargs and kwargs[param] is not NOT_GIVEN:
ollama_supported_params[param] = kwargs[param]
# Add extra headers if available
if "extra_headers" in kwargs:
ollama_supported_params["extra_headers"] = kwargs["extra_headers"]
# Add tools for compatibility with Qwen
if "tools" in kwargs and kwargs.get("tools") and kwargs.get("tools") is not NOT_GIVEN:
ollama_supported_params["tools"] = kwargs.get("tools")
# Remove None values and filter out unsupported parameters
ollama_kwargs = {
k: v
for k, v in ollama_supported_params.items()
if v is not None and k not in ["response_format", "store"]
}
# Check if this is a Qwen model
model_str = str(self.model).lower()
is_qwen = "qwen" in model_str
api_base = get_ollama_api_base()
if stream:
response = Response(
id=FAKE_RESPONSES_ID,
created_at=time.time(),
model=self.model,
object="response",
output=[],
tool_choice="auto"
if tool_choice is None or tool_choice == NOT_GIVEN
else cast(Literal["auto", "required", "none"], tool_choice),
top_p=model_settings.top_p,
temperature=model_settings.temperature,
tools=[],
parallel_tool_calls=parallel_tool_calls or False,
)
# Get streaming response
stream_obj = await litellm.acompletion(
**ollama_kwargs, api_base=api_base, custom_llm_provider="openai"
)
return response, stream_obj
else:
# Get completion response
return await litellm.acompletion(
**ollama_kwargs,
api_base=api_base,
custom_llm_provider="openai",
)
def _get_model_max_tokens(self, model_name: str) -> int:
"""Get the maximum input tokens for a model from pricing.json or default."""
try:
import pathlib
pricing_path = pathlib.Path("pricing.json")
if pricing_path.exists():
with open(pricing_path, encoding="utf-8") as f:
pricing_data = json.load(f)
model_info = pricing_data.get(model_name, {})
return model_info.get("max_input_tokens", 200000)
except Exception:
pass
# Default to 200k if not found
return 200000
async def _auto_compact_if_needed(self, estimated_tokens: int, input: str | list[TResponseInputItem], system_instructions: str | None) -> tuple[str | list[TResponseInputItem], str | None, bool]:
"""Check if auto-compaction is needed and perform it if necessary.
Returns:
tuple: (potentially modified input, potentially modified system_instructions, whether compaction occurred)
"""
# Check if auto-compaction is disabled
if os.getenv("CAI_AUTO_COMPACT", "true").lower() == "false":
return input, system_instructions, False
max_tokens = self._get_model_max_tokens(str(self.model))
threshold_percent = float(os.getenv("CAI_AUTO_COMPACT_THRESHOLD", "0.8"))
threshold = max_tokens * threshold_percent
if estimated_tokens <= threshold:
return input, system_instructions, False
# Auto-compaction needed
from rich.console import Console
console = Console()
# Update context usage in environment for toolbar
context_usage = estimated_tokens / max_tokens
os.environ['CAI_CONTEXT_USAGE'] = str(context_usage)
console.print(f"\n[yellow]⚠️ Context usage at {(estimated_tokens/max_tokens)*100:.1f}% ({estimated_tokens:,}/{max_tokens:,} tokens)[/yellow]")
console.print("[yellow]Triggering automatic context compaction...[/yellow]\n")
# Import compact command components
try:
from cai.repl.commands.memory import MEMORY_COMMAND_INSTANCE
# Generate AI summary of the conversation
summary = await MEMORY_COMMAND_INSTANCE._ai_summarize_history(self.agent_name)
if summary:
# Store the summary
from cai.repl.commands.memory import COMPACTED_SUMMARIES
COMPACTED_SUMMARIES[self.agent_name] = summary
# Clear the message history and keep only essential messages
self.message_history.clear()
# Reset context usage after clearing
os.environ['CAI_CONTEXT_USAGE'] = '0.0'
# Reset context usage since we cleared history
os.environ['CAI_CONTEXT_USAGE'] = '0.0'
# Create new input with summary
new_system_instructions = system_instructions or ""
if new_system_instructions:
new_system_instructions += "\n\n"
new_system_instructions += f"Previous conversation summary:\n{summary}"
# Keep only the current input (user's latest message)
if isinstance(input, str):
new_input = input
else:
# For list input, keep only user messages
new_input = []
for item in input:
if hasattr(item, 'role') and item.role == 'user':
new_input.append(item)
elif isinstance(item, dict) and item.get('role') == 'user':
new_input.append(item)
# If no user messages found, keep the original input
if not new_input:
new_input = input
# Re-estimate tokens with compacted context
test_messages = self._converter.items_to_messages(new_input, model_instance=self)
if new_system_instructions:
test_messages.insert(0, {"role": "system", "content": new_system_instructions})
new_tokens, _ = count_tokens_with_tiktoken(test_messages)
console.print(f"[green]✓ Context compacted: {estimated_tokens:,}{new_tokens:,} tokens ({(1-new_tokens/estimated_tokens)*100:.1f}% reduction)[/green]\n")
# Update context usage after compaction
new_context_usage = new_tokens / max_tokens if max_tokens > 0 else 0.0
os.environ['CAI_CONTEXT_USAGE'] = str(new_context_usage)
return new_input, new_system_instructions, True
except Exception as e:
console.print(f"[red]Auto-compaction failed: {e}[/red]")
console.print("[yellow]Continuing with full context...[/yellow]\n")
return input, system_instructions, False
def _intermediate_logs(self):
"""Intermediate logging if conditions are met."""
if (
self.logger
and self.interaction_counter > 0
and self.interaction_counter % self.INTERMEDIATE_LOG_INTERVAL == 0
):
process_intermediate_logs(self.logger.filename, self.logger.session_id)
def _get_client(self) -> AsyncOpenAI:
if self._client is None:
self._client = AsyncOpenAI()
return self._client
# Helper function to detect and format function calls from various models
def _detect_and_format_function_calls(self, delta):
"""
Helper to detect function calls in different formats and normalize them.
Handles Qwen specifics where function calls may be formatted differently.
Returns: List of normalized tool calls or None
"""
# Standard OpenAI-style tool_calls format
if hasattr(delta, "tool_calls") and delta.tool_calls:
return delta.tool_calls
elif isinstance(delta, dict) and "tool_calls" in delta and delta["tool_calls"]:
return delta["tool_calls"]
# Qwen/Ollama function_call format
if isinstance(delta, dict) and "function_call" in delta:
function_call = delta["function_call"]
return [
{
"index": 0,
"id": f"call_{time.time_ns()}", # Generate a unique ID
"type": "function",
"function": {
"name": function_call.get("name", ""),
"arguments": function_call.get("arguments", ""),
},
}
]
if isinstance(delta, dict) and "content" in delta:
content = delta["content"]
# Try to detect if the content is a JSON string with function call format
try:
if isinstance(content, str) and "{" in content and "}" in content:
# Try to extract JSON from the content (it might be embedded in text)
json_start = content.find("{")
json_end = content.rfind("}") + 1
if json_start >= 0 and json_end > json_start:
json_str = content[json_start:json_end]
parsed = json.loads(json_str)
if "name" in parsed and "arguments" in parsed:
# This looks like a function call in JSON format
return [
{
"index": 0,
"id": f"call_{time.time_ns()}", # Generate a unique ID
"type": "function",
"function": {
"name": parsed["name"],
"arguments": json.dumps(parsed["arguments"])
if isinstance(parsed["arguments"], dict)
else parsed["arguments"],
},
}
]
except Exception:
# If JSON parsing fails, just continue with normal processing
pass
# Anthropic-style tool_use format
if hasattr(delta, "tool_use") and delta.tool_use:
tool_use = delta.tool_use
return [
{
"index": 0,
"id": tool_use.get("id", f"tool_{time.time_ns()}"),
"type": "function",
"function": {
"name": tool_use.get("name", ""),
"arguments": tool_use.get("input", "{}"),
},
}
]
elif isinstance(delta, dict) and "tool_use" in delta and delta["tool_use"]:
tool_use = delta["tool_use"]
return [
{
"index": 0,
"id": tool_use.get("id", f"tool_{time.time_ns()}"),
"type": "function",
"function": {
"name": tool_use.get("name", ""),
"arguments": tool_use.get("input", "{}"),
},
}
]
return None
class _Converter:
def __init__(self):
"""Initialize converter with instance-based state."""
self.recent_tool_calls = {}
self.tool_outputs = {}
def convert_tool_choice(
self, tool_choice: Literal["auto", "required", "none"] | str | None
) -> ChatCompletionToolChoiceOptionParam | NotGiven:
if tool_choice is None:
return "auto"
elif tool_choice == "auto":
return "auto"
elif tool_choice == "required":
return "required"
elif tool_choice == "none":
return "none"
else:
return {
"type": "function",
"function": {
"name": tool_choice,
},
}
def convert_response_format(
self, final_output_schema: AgentOutputSchema | None
) -> ResponseFormat | NotGiven:
if not final_output_schema or final_output_schema.is_plain_text():
return None
return {
"type": "json_schema",
"json_schema": {
"name": "final_output",
"strict": final_output_schema.strict_json_schema,
"schema": final_output_schema.json_schema(),
},
}
def message_to_output_items(self, message: ChatCompletionMessage) -> list[TResponseOutputItem]:
items: list[TResponseOutputItem] = []
message_item = ResponseOutputMessage(
id=FAKE_RESPONSES_ID,
content=[],
role="assistant",
type="message",
status="completed",
)
if message.content:
message_item.content.append(
ResponseOutputText(text=message.content, type="output_text", annotations=[])
)
if hasattr(message, "refusal") and message.refusal:
message_item.content.append(
ResponseOutputRefusal(refusal=message.refusal, type="refusal")
)
if hasattr(message, "audio") and message.audio:
raise AgentsException("🎵 Audio output not supported - Text responses only")
if message_item.content:
items.append(message_item)
if hasattr(message, "tool_calls") and message.tool_calls:
for tool_call in message.tool_calls:
items.append(
ResponseFunctionToolCall(
id=FAKE_RESPONSES_ID,
call_id=tool_call.id[:40],
arguments=tool_call.function.arguments,
name=tool_call.function.name,
type="function_call",
)
)
return items
def maybe_easy_input_message(self, item: Any) -> EasyInputMessageParam | None:
if not isinstance(item, dict):
return None
keys = item.keys()
# EasyInputMessageParam only has these two keys
if keys != {"content", "role"}:
return None
role = item.get("role", None)
if role not in ("user", "assistant", "system", "developer"):
return None
if "content" not in item:
return None
return cast(EasyInputMessageParam, item)
def maybe_input_message(self, item: Any) -> Message | None:
if (
isinstance(item, dict)
and item.get("type") == "message"
and item.get("role")
in (
"user",
"system",
"developer",
)
):
return cast(Message, item)
return None
def maybe_file_search_call(self, item: Any) -> ResponseFileSearchToolCallParam | None:
if isinstance(item, dict) and item.get("type") == "file_search_call":
return cast(ResponseFileSearchToolCallParam, item)
return None
def maybe_function_tool_call(self, item: Any) -> ResponseFunctionToolCallParam | None:
if isinstance(item, dict) and item.get("type") == "function_call":
return cast(ResponseFunctionToolCallParam, item)
return None
def maybe_function_tool_call_output(
self,
item: Any,
) -> FunctionCallOutput | None:
if isinstance(item, dict) and item.get("type") == "function_call_output":
return cast(FunctionCallOutput, item)
return None
def maybe_item_reference(self, item: Any) -> ItemReference | None:
if isinstance(item, dict) and item.get("type") == "item_reference":
return cast(ItemReference, item)
return None
def maybe_response_output_message(self, item: Any) -> ResponseOutputMessageParam | None:
# ResponseOutputMessage is only used for messages with role assistant
if (
isinstance(item, dict)
and item.get("type") == "message"
and item.get("role") == "assistant"
):
return cast(ResponseOutputMessageParam, item)
return None
def extract_text_content(
self, content: str | Iterable[ResponseInputContentParam]
) -> str | list[ChatCompletionContentPartTextParam]:
all_content = self.extract_all_content(content)
if isinstance(all_content, str):
return all_content
out: list[ChatCompletionContentPartTextParam] = []
for c in all_content:
if c.get("type") == "text":
out.append(cast(ChatCompletionContentPartTextParam, c))
return out
def extract_all_content(
self, content: str | Iterable[ResponseInputContentParam]
) -> str | list[ChatCompletionContentPartParam]:
if isinstance(content, str):
return content
out: list[ChatCompletionContentPartParam] = []
for c in content:
if isinstance(c, dict) and c.get("type") == "input_text":
casted_text_param = cast(ResponseInputTextParam, c)
out.append(
ChatCompletionContentPartTextParam(
type="text",
text=casted_text_param["text"],
)
)
elif isinstance(c, dict) and c.get("type") == "input_image":
casted_image_param = cast(ResponseInputImageParam, c)
if "image_url" not in casted_image_param or not casted_image_param["image_url"]:
raise UserError(
"🖼️ Image URLs required - Upload images to a URL first"
)
out.append(
ChatCompletionContentPartImageParam(
type="image_url",
image_url={
"url": casted_image_param["image_url"],
"detail": casted_image_param["detail"],
},
)
)
elif isinstance(c, dict) and c.get("type") == "input_file":
raise UserError("📄 File uploads not supported - Use image URLs or text content")
else:
raise UserError(f"❓ Unrecognized content type - Expected 'input_text' or 'input_image'")
return out
def items_to_messages(
self,
items: str | Iterable[TResponseInputItem],
model_instance=None,
) -> list[ChatCompletionMessageParam]:
"""
Convert a sequence of 'Item' objects into a list of ChatCompletionMessageParam.
Rules:
- EasyInputMessage or InputMessage (role=user) => ChatCompletionUserMessageParam
- EasyInputMessage or InputMessage (role=system) => ChatCompletionSystemMessageParam
- EasyInputMessage or InputMessage (role=developer) => ChatCompletionDeveloperMessageParam
- InputMessage (role=assistant) => Start or flush a ChatCompletionAssistantMessageParam
- response_output_message => Also produces/flushes a ChatCompletionAssistantMessageParam
- tool calls get attached to the *current* assistant message, or create one if none.
- tool outputs => ChatCompletionToolMessageParam
"""
if isinstance(items, str):
return [
ChatCompletionUserMessageParam(
role="user",
content=items,
)
]
result: list[ChatCompletionMessageParam] = []
current_assistant_msg: ChatCompletionAssistantMessageParam | None = None
def flush_assistant_message() -> None:
nonlocal current_assistant_msg
if current_assistant_msg is not None:
# The API doesn't support empty arrays for tool_calls
if not current_assistant_msg.get("tool_calls"):
# Ensure content is not None if tool_calls are absent and content is also None
# Some models like Anthropic require some content, even if it's just a placeholder.
if current_assistant_msg.get("content") is None:
current_assistant_msg["content"] = (
"(No text content in this assistant message)" # Or just an empty string if preferred
)
current_assistant_msg.pop(
"tool_calls", None
) # Use pop with default to avoid KeyError
result.append(current_assistant_msg)
current_assistant_msg = None
def ensure_assistant_message() -> ChatCompletionAssistantMessageParam:
nonlocal current_assistant_msg
if current_assistant_msg is None:
current_assistant_msg = ChatCompletionAssistantMessageParam(role="assistant")
current_assistant_msg["tool_calls"] = []
return current_assistant_msg
for item in items:
# NEW: Handle 'tool' messages from history
if (
isinstance(item, dict)
and item.get("role") == "tool"
and "tool_call_id" in item
and "content" in item
):
flush_assistant_message() # Ensure any pending assistant message is flushed
tool_message: ChatCompletionToolMessageParam = {
"role": "tool",
"tool_call_id": item["tool_call_id"],
"content": str(item["content"] or ""), # Ensure content is a string
}
result.append(tool_message)
continue
# 0) Assistant messages with tool_calls only (from memory)
if (
isinstance(item, dict)
and item.get("role") == "assistant"
and item.get("tool_calls")
):
flush_assistant_message()
tool_calls_param: list[ChatCompletionMessageToolCallParam] = []
for tc in item["tool_calls"]:
function_details = tc.get("function", {})
arguments = function_details.get("arguments")
# Ensure arguments is a valid JSON string, defaulting to "{}" if empty or None
if arguments is None or (
isinstance(arguments, str) and arguments.strip() == ""
):
arguments = "{}"
elif isinstance(arguments, dict):
# Ensure it's a string if it's a dict (should already be string per schema)
arguments = json.dumps(arguments)
tool_calls_param.append(
ChatCompletionMessageToolCallParam(
id=tc.get("id", "")[:40],
type=tc.get("type", "function"),
function={
"name": function_details.get("name", "unknown_function"),
"arguments": arguments, # Use sanitized arguments
},
)
)
msg_asst: ChatCompletionAssistantMessageParam = {
"role": "assistant",
"content": item.get("content"), # Content can be None here
"tool_calls": tool_calls_param,
}
result.append(msg_asst)
# Skip further processing for this item
continue
# 1) Check easy input message
if easy_msg := self.maybe_easy_input_message(item):
role = easy_msg["role"]
content = easy_msg["content"]
if role == "user":
flush_assistant_message()
msg_user: ChatCompletionUserMessageParam = {
"role": "user",
"content": self.extract_all_content(content),
}
result.append(msg_user)
elif role == "system":
flush_assistant_message()
msg_system: ChatCompletionSystemMessageParam = {
"role": "system",
"content": self.extract_text_content(content),
}
result.append(msg_system)
elif role == "developer":
flush_assistant_message()
msg_developer: ChatCompletionDeveloperMessageParam = {
"role": "developer",
"content": self.extract_text_content(content),
}
result.append(msg_developer)
elif role == "assistant":
flush_assistant_message()
msg_assistant: ChatCompletionAssistantMessageParam = {
"role": "assistant",
"content": self.extract_text_content(content),
}
result.append(msg_assistant)
else:
raise UserError(f"👥 Invalid role '{role}' - Use: user, assistant, system, or developer")
# 2) Check input message
elif in_msg := self.maybe_input_message(item):
role = in_msg["role"]
content = in_msg["content"]
flush_assistant_message()
if role == "user":
msg_user = {
"role": "user",
"content": self.extract_all_content(content),
}
result.append(msg_user)
elif role == "system":
msg_system = {
"role": "system",
"content": self.extract_text_content(content),
}
result.append(msg_system)
elif role == "developer":
msg_developer = {
"role": "developer",
"content": self.extract_text_content(content),
}
result.append(msg_developer)
else:
raise UserError(f"👥 Invalid message role '{role}' - Must be: user, system, or developer")
# 3) response output message => assistant
elif resp_msg := self.maybe_response_output_message(item):
flush_assistant_message()
new_asst = ChatCompletionAssistantMessageParam(role="assistant")
contents = resp_msg["content"]
text_segments = []
for c in contents:
if c["type"] == "output_text":
text_segments.append(c["text"])
elif c["type"] == "refusal":
new_asst["refusal"] = c["refusal"]
elif c["type"] == "output_audio":
# Can't handle this, b/c chat completions expects an ID which we dont have
raise UserError(
"🎵 Audio content must use audio IDs - Direct audio data not supported"
)
else:
raise UserError("❓ Unknown assistant message content - Check message format")
if text_segments:
combined = "\n".join(text_segments)
new_asst["content"] = combined
new_asst["tool_calls"] = []
current_assistant_msg = new_asst
# 4) function/file-search calls => attach to assistant
elif file_search := self.maybe_file_search_call(item):
asst = ensure_assistant_message()
tool_calls = list(asst.get("tool_calls", []))
new_tool_call = ChatCompletionMessageToolCallParam(
id=file_search["id"][:40],
type="function",
function={
"name": "file_search_call",
"arguments": json.dumps(
{
"queries": file_search.get("queries", []),
"status": file_search.get("status"),
}
),
},
)
tool_calls.append(new_tool_call)
asst["tool_calls"] = tool_calls
elif func_call := self.maybe_function_tool_call(item):
asst = ensure_assistant_message()
tool_calls = list(asst.get("tool_calls", []))
# Save the tool call details for later matching with output
if not hasattr(self, "recent_tool_calls"):
self.recent_tool_calls = {}
# Store the tool call by ID for later reference
# Also store the current time for execution timing
import time
self.recent_tool_calls[func_call["call_id"]] = {
"name": func_call["name"],
"arguments": func_call["arguments"],
"start_time": time.time(),
"execution_info": {"start_time": time.time()},
}
arguments = func_call.get("arguments") # func_call is a dict here
# Ensure arguments is a valid JSON string, defaulting to "{}" if empty or None
if arguments is None or (isinstance(arguments, str) and arguments.strip() == ""):
arguments = "{}"
elif isinstance(arguments, dict):
arguments = json.dumps(arguments)
new_tool_call = ChatCompletionMessageToolCallParam(
id=func_call["call_id"][:40],
type="function",
function={
"name": func_call["name"],
"arguments": arguments, # Use sanitized arguments
},
)
tool_calls.append(new_tool_call)
asst["tool_calls"] = tool_calls
# 5) function call output => tool message
elif func_output := self.maybe_function_tool_call_output(item):
# Store the output for this call_id
call_id = func_output["call_id"]
output_content = func_output["output"]
# IMPORTANT: Truncate call_id to 40 characters for consistency
truncated_call_id = call_id[:40] if call_id else call_id
# Update execution timing if we have the start time
if hasattr(self, "recent_tool_calls") and call_id in self.recent_tool_calls:
tool_call_details = self.recent_tool_calls[call_id] # Renamed for clarity
if "start_time" in tool_call_details:
end_time = time.time()
tool_execution_time = end_time - tool_call_details["start_time"]
# Update the execution info
if "execution_info" in tool_call_details:
tool_call_details["execution_info"]["end_time"] = end_time
tool_call_details["execution_info"]["tool_time"] = tool_execution_time
# If this is the first tool being executed, record the total time from conversation start
if not hasattr(self, "conversation_start_time"):
self.conversation_start_time = tool_call_details["start_time"]
total_time = end_time - getattr(
self, "conversation_start_time", tool_call_details["start_time"]
)
tool_call_details["execution_info"]["total_time"] = total_time
# Store the output so it can be accessed later
if not hasattr(self, "tool_outputs"):
self.tool_outputs = {}
self.tool_outputs[call_id] = output_content
# Display the tool output immediately with the matched tool call
from cai.util import cli_print_tool_output
# Look up the original tool call to get the name and arguments
tool_name = "Unknown Tool"
tool_args = {}
execution_info = {}
if hasattr(self, "recent_tool_calls") and call_id in self.recent_tool_calls:
tool_call_details = self.recent_tool_calls[call_id] # Renamed for clarity
tool_name = tool_call_details.get("name", "Unknown Tool")
tool_args = tool_call_details.get("arguments", {})
execution_info = tool_call_details.get("execution_info", {})
# Get token counts from the OpenAIChatCompletionsModel if available
model_instance = None
for frame in inspect.stack():
if "self" in frame.frame.f_locals:
self_obj = frame.frame.f_locals["self"]
if isinstance(self_obj, OpenAIChatCompletionsModel):
model_instance = self_obj
break
# Always create a token_info dictionary, even if some values are zero
token_info = {
"interaction_input_tokens": getattr(
model_instance, "interaction_input_tokens", 0
),
"interaction_output_tokens": getattr(
model_instance, "interaction_output_tokens", 0
),
"interaction_reasoning_tokens": getattr(
model_instance, "interaction_reasoning_tokens", 0
),
"total_input_tokens": getattr(model_instance, "total_input_tokens", 0),
"total_output_tokens": getattr(model_instance, "total_output_tokens", 0),
"total_reasoning_tokens": getattr(model_instance, "total_reasoning_tokens", 0),
"model": str(getattr(model_instance, "model", "")),
"agent_name": getattr(model_instance, "agent_name", "Agent"),
}
# Use already-calculated costs from COST_TRACKER instead of recalculating
if model_instance and hasattr(model_instance, "model"):
from cai.util import COST_TRACKER
# Use the last recorded costs instead of recalculating
token_info["interaction_cost"] = getattr(COST_TRACKER, "last_interaction_cost", 0.0)
token_info["total_cost"] = getattr(COST_TRACKER, "last_total_cost", 0.0)
# Check if we're in streaming mode
is_streaming_enabled = os.environ.get("CAI_STREAM", "false").lower() == "true"
# Check if this output was already displayed during streaming
# For async sessions, we always display since they don't have real streaming
should_display = True
# If streaming is enabled, check if this was already shown
if (
is_streaming_enabled
and hasattr(self, "recent_tool_calls")
and call_id in self.recent_tool_calls
):
tool_call_info = self.recent_tool_calls[call_id]
# Check if this tool was executed very recently (within last 5 seconds)
# This indicates it was likely shown during streaming
if "start_time" in tool_call_info:
time_since_execution = time.time() - tool_call_info["start_time"]
# For generic_linux_command executed recently in streaming mode, skip display
# But always display for async session commands (they have session_id in args)
# and always display for non-generic_linux_command tools
if time_since_execution < 5.0 and "_command" in tool_name.lower():
# Parse arguments to check if this is an async session command
try:
import json
args_dict = (
json.loads(tool_args)
if isinstance(tool_args, str)
else tool_args
)
# If it has session_id, it's an async command - always show
if not (
isinstance(args_dict, dict) and args_dict.get("session_id")
):
should_display = False
except:
should_display = False
# Only display if it hasn't been shown during streaming
if should_display:
cli_print_tool_output(
tool_name=tool_name,
args=tool_args,
output=output_content,
call_id=call_id,
execution_info=execution_info,
token_info=token_info,
)
# Continue with normal processing
flush_assistant_message()
# ATOMIC ADDITION: Add pending tool call and response together
# This ensures we never have tool calls without responses in history
if model_instance and hasattr(model_instance, "_pending_tool_calls"):
# Check if we have a pending tool call for this ID
if call_id in model_instance._pending_tool_calls:
# Add the assistant message with tool call first
pending_msg = model_instance._pending_tool_calls[call_id]
model_instance.add_to_message_history(pending_msg)
# Now add the tool response
tool_response_msg = {
"role": "tool",
"tool_call_id": truncated_call_id,
"content": func_output["output"],
}
model_instance.add_to_message_history(tool_response_msg)
# Remove from pending
del model_instance._pending_tool_calls[call_id]
# Log both messages
if hasattr(model_instance, "logger"):
# Log the tool call with its response
# Note: Tool responses are logged as part of the training data recording,
# not as separate events
pass
# Now add the tool message with truncated call_id
msg: ChatCompletionToolMessageParam = {
"role": "tool",
"tool_call_id": truncated_call_id,
"content": func_output["output"],
}
result.append(msg)
# 6) item reference => handle or raise
elif item_ref := self.maybe_item_reference(item):
raise UserError(
"🔗 Item references not supported - Include content directly"
)
# 7) If we haven't recognized it => fail or ignore
else:
raise UserError("❌ Invalid message format - Check documentation for supported types")
flush_assistant_message()
return result
class ToolConverter:
@classmethod
def to_openai(cls, tool: Tool) -> ChatCompletionToolParam:
if isinstance(tool, FunctionTool):
return {
"type": "function",
"function": {
"name": tool.name,
"description": tool.description or "",
"parameters": tool.params_json_schema,
},
}
raise UserError(
f"Hosted tools are not supported with the ChatCompletions API. FGot tool type: "
f"{type(tool)}, tool: {tool}"
)
@classmethod
def convert_handoff_tool(cls, handoff: Handoff[Any]) -> ChatCompletionToolParam:
return {
"type": "function",
"function": {
"name": handoff.tool_name,
"description": handoff.tool_description,
"parameters": handoff.input_json_schema,
},
}