File size: 9,320 Bytes
aa15bce
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
"""Simplified Execution Agent Runtime."""

import inspect
import json
from typing import Dict, Any, List, Optional, Tuple
from dataclasses import dataclass

from .agent import ExecutionAgent
from .tools import get_tool_schemas, get_tool_registry
from ...config import get_settings
from ...openrouter_client import request_chat_completion
from ...logging_config import logger


@dataclass
class ExecutionResult:
    """Result from an execution agent."""
    agent_name: str
    success: bool
    response: str
    error: Optional[str] = None
    tools_executed: List[str] = None


class ExecutionAgentRuntime:
    """Manages the execution of a single agent request."""

    MAX_TOOL_ITERATIONS = 8

    # Initialize execution agent runtime with settings, tools, and agent instance
    def __init__(self, agent_name: str):
        settings = get_settings()
        self.agent = ExecutionAgent(agent_name)
        self.api_key = settings.api_key
        self.model = settings.execution_agent_model
        self.tool_registry = get_tool_registry(agent_name=agent_name)
        self.tool_schemas = get_tool_schemas()

        if not self.api_key:
            raise ValueError("API key not configured. Set API_KEY environment variable.")

    # Main execution loop for running agent with LLM calls and tool execution
    async def execute(self, instructions: str) -> ExecutionResult:
        """Execute the agent with given instructions."""
        try:
            # Build system prompt with history
            system_prompt = self.agent.build_system_prompt_with_history()

            # Start conversation with the instruction
            messages = [{"role": "user", "content": instructions}]
            tools_executed: List[str] = []
            final_response: Optional[str] = None

            for iteration in range(self.MAX_TOOL_ITERATIONS):
                logger.info(
                    f"[{self.agent.name}] Requesting plan (iteration {iteration + 1})"
                )
                response = await self._make_llm_call(system_prompt, messages, with_tools=True)
                assistant_message = response.get("choices", [{}])[0].get("message", {})

                if not assistant_message:
                    raise RuntimeError("LLM response did not include an assistant message")

                raw_tool_calls = assistant_message.get("tool_calls", []) or []
                parsed_tool_calls = self._extract_tool_calls(raw_tool_calls)

                assistant_entry: Dict[str, Any] = {
                    "role": "assistant",
                    "content": assistant_message.get("content", "") or "",
                }
                if raw_tool_calls:
                    assistant_entry["tool_calls"] = raw_tool_calls
                messages.append(assistant_entry)

                if not parsed_tool_calls:
                    final_response = assistant_entry["content"] or "No action required."
                    break

                for tool_call in parsed_tool_calls:
                    tool_name = tool_call.get("name", "")
                    tool_args = tool_call.get("arguments", {})
                    call_id = tool_call.get("id")

                    if not tool_name:
                        logger.warning("Tool call missing name: %s", tool_call)
                        failure = {"error": "Tool call missing name; unable to execute."}
                        tool_message = {
                            "role": "tool",
                            "tool_call_id": call_id or "unknown_tool",
                            "content": self._format_tool_result(
                                tool_name or "<unknown>", False, failure, tool_args
                            ),
                        }
                        messages.append(tool_message)
                        continue

                    tools_executed.append(tool_name)
                    logger.info(f"[{self.agent.name}] Executing tool: {tool_name}")

                    success, result = await self._execute_tool(tool_name, tool_args)

                    if success:
                        logger.info(f"[{self.agent.name}] Tool {tool_name} completed successfully")
                        record_payload = self._safe_json_dump(result)
                    else:
                        error_detail = result.get("error") if isinstance(result, dict) else str(result)
                        logger.warning(f"[{self.agent.name}] Tool {tool_name} failed: {error_detail}")
                        record_payload = error_detail

                    self.agent.record_tool_execution(
                        tool_name,
                        self._safe_json_dump(tool_args),
                        record_payload
                    )

                    tool_message = {
                        "role": "tool",
                        "tool_call_id": call_id or tool_name,
                        "content": self._format_tool_result(tool_name, success, result, tool_args),
                    }
                    messages.append(tool_message)

            else:
                raise RuntimeError("Reached tool iteration limit without final response")

            if final_response is None:
                raise RuntimeError("LLM did not return a final response")

            self.agent.record_response(final_response)

            return ExecutionResult(
                agent_name=self.agent.name,
                success=True,
                response=final_response,
                tools_executed=tools_executed
            )

        except Exception as e:
            logger.error(f"[{self.agent.name}] Execution failed: {e}")
            error_msg = str(e)
            failure_text = f"Failed to complete task: {error_msg}"
            self.agent.record_response(f"Error: {error_msg}")

            return ExecutionResult(
                agent_name=self.agent.name,
                success=False,
                response=failure_text,
                error=error_msg
            )

    # Execute API call with system prompt, messages, and optional tool schemas
    async def _make_llm_call(self, system_prompt: str, messages: List[Dict], with_tools: bool) -> Dict:
        """Make an LLM call."""
        tools_to_send = self.tool_schemas if with_tools else None
        logger.info(f"[{self.agent.name}] Calling LLM with model: {self.model}, tools: {len(tools_to_send) if tools_to_send else 0}")
        return await request_chat_completion(
            model=self.model,
            messages=messages,
            system=system_prompt,
            api_key=self.api_key,
            tools=tools_to_send
        )

    # Parse and validate tool calls from LLM response into structured format
    def _extract_tool_calls(self, raw_tools: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
        """Extract tool calls from an assistant message."""
        tool_calls: List[Dict[str, Any]] = []

        for tool in raw_tools:
            function = tool.get("function", {})
            name = function.get("name", "")
            args = function.get("arguments", "")

            if isinstance(args, str):
                try:
                    args = json.loads(args) if args else {}
                except json.JSONDecodeError:
                    args = {}

            if name:
                tool_calls.append({
                    "id": tool.get("id"),
                    "name": name,
                    "arguments": args,
                })

        return tool_calls

    # Safely convert objects to JSON with fallback to string representation
    def _safe_json_dump(self, payload: Any) -> str:
        """Serialize payload to JSON, falling back to string representation."""
        try:
            return json.dumps(payload, default=str)
        except TypeError:
            return str(payload)

    # Format tool execution results into JSON structure for LLM consumption
    def _format_tool_result(
        self,
        tool_name: str,
        success: bool,
        result: Any,
        arguments: Dict[str, Any],
    ) -> str:
        """Build a structured string for tool responses."""
        if success:
            payload: Dict[str, Any] = {
                "tool": tool_name,
                "status": "success",
                "arguments": arguments,
                "result": result,
            }
        else:
            error_detail = result.get("error") if isinstance(result, dict) else str(result)
            payload = {
                "tool": tool_name,
                "status": "error",
                "arguments": arguments,
                "error": error_detail,
            }
        return self._safe_json_dump(payload)

    # Execute tool function from registry with error handling and async support
    async def _execute_tool(self, tool_name: str, arguments: Dict) -> Tuple[bool, Any]:
        """Execute a tool. Returns (success, result)."""
        tool_func = self.tool_registry.get(tool_name)
        if not tool_func:
            return False, {"error": f"Unknown tool: {tool_name}"}

        try:
            result = tool_func(**arguments)
            if inspect.isawaitable(result):
                result = await result
            return True, result
        except Exception as e:
            return False, {"error": str(e)}