Relace MCP Server

from typing import Any from .constants import MAX_TOOL_RESULT_CHARS def truncate_for_context( text: str, max_chars: int = MAX_TOOL_RESULT_CHARS, tool_hint: str = "" ) -> str: """Truncate overly long tool result to avoid context overflow. Args: text: Text to truncate. max_chars: Maximum characters. tool_hint: Tool hint message shown when truncated. """ if len(text) <= max_chars: return text shown = max_chars for _ in range(5): hint_msg = f"\n... [truncated] ({len(text)} chars total, showing {shown})" if tool_hint: hint_msg += f"\n{tool_hint}" allowed = max_chars - len(hint_msg) if allowed < 0: allowed = 0 if allowed == shown: break shown = allowed hint_msg = f"\n... [truncated] ({len(text)} chars total, showing {shown})" if tool_hint: hint_msg += f"\n{tool_hint}" # If the hint itself exceeds max_chars, return as much of the hint as fits. if len(hint_msg) >= max_chars: return hint_msg[:max_chars] truncated = text[:shown] return truncated + hint_msg def estimate_context_size(messages: list[dict[str, Any]]) -> int: """Estimate total character count of messages.""" total = 0 for msg in messages: content = msg.get("content", "") if isinstance(content, str): total += len(content) elif isinstance(content, list): # OpenAI-style multimodal content: [{"type":"text","text":"..."}, ...] for part in content: if isinstance(part, str): total += len(part) elif isinstance(part, dict): for key in ("text", "input_text", "content"): value = part.get(key) if isinstance(value, str): total += len(value) break # tool_calls also take space tool_calls = msg.get("tool_calls", []) for tc in tool_calls: func = tc.get("function", {}) total += len(func.get("arguments", "")) return total