Spaces:
Sleeping
Sleeping
File size: 10,486 Bytes
6085b61 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 | import os
import time
from typing import Any, Callable, Optional
from langchain_community.chat_models import ChatLiteLLM
from langchain_core.messages import SystemMessage
from observability.langfuse_client import get_langfuse
_SKIP_ERRORS = (
"ResourceExhausted",
"RateLimit",
"QuotaExceeded",
"APIConnectionError",
"AuthenticationError",
"BadRequestError",
)
_TRANSIENT_ERROR_NAMES = (
"RateLimitError",
"ResourceExhausted",
"APIConnectionError",
"Timeout",
"ConnectionError",
"ServiceUnavailable",
"InternalServerError",
)
_TRANSIENT_EXCEPTIONS = (Exception,)
class AgentRuntime:
cost_tracker: Any = None
circuit_breaker: Any = None
circuit_events: list = []
fallback_models: list[str] = []
tools: list = []
executor_node: Any = None
export_ui_state_fn: Optional[Callable] = None
_runtime = AgentRuntime()
def configure_runtime(
cost_tracker,
circuit_breaker,
circuit_events: list,
fallback_models: list[str],
tools: list,
executor_node,
export_ui_state_fn: Optional[Callable] = None,
) -> None:
_runtime.cost_tracker = cost_tracker
_runtime.circuit_breaker = circuit_breaker
_runtime.circuit_events = circuit_events
_runtime.fallback_models = fallback_models
_runtime.tools = tools
_runtime.executor_node = executor_node
_runtime.export_ui_state_fn = export_ui_state_fn
def get_runtime() -> AgentRuntime:
return _runtime
def _model_available(model: str) -> bool:
if (
model.startswith("gemini/")
and not os.environ.get("GOOGLE_API_KEY")
and not os.environ.get("GEMINI_API_KEY")
):
return False
if model.startswith("groq/") and not os.environ.get("GROQ_API_KEY"):
return False
return True
def _make_llm(model: str, tools_list: list):
return ChatLiteLLM(model=model, temperature=0).bind_tools(tools_list)
def _is_transient(e: Exception) -> bool:
name = type(e).__name__
msg = str(e).lower()
return (
any(t in name for t in _TRANSIENT_ERROR_NAMES)
or "rate limit" in msg
or "timeout" in msg
or "connection" in msg
or "503" in msg
or "502" in msg
or "529" in msg
)
def _call_with_retry(model: str, msgs: list, max_retries: int, base_delay: float):
for attempt in range(max_retries + 1):
try:
llm = _make_llm(model, _runtime.tools)
return llm.invoke(msgs)
except Exception as e:
if attempt >= max_retries or not _is_transient(e):
raise
delay = min(base_delay * (2.0**attempt), 30.0)
print(
f"[RETRY] {model} attempt {attempt + 1}/{max_retries} failed ({type(e).__name__}). Retrying in {delay:.1f}s..."
)
time.sleep(delay)
raise RuntimeError("Unreachable")
def _extract_usage(response) -> Optional[dict]:
usage = getattr(response, "usage_metadata", None) or getattr(
response, "response_metadata", {}
).get("usage", None)
if usage is None:
return None
input_tokens = (
getattr(usage, "prompt_token_count", None)
or getattr(usage, "input_tokens", None)
or (usage.get("prompt_tokens") if isinstance(usage, dict) else None)
or 0
)
output_tokens = (
getattr(usage, "candidates_token_count", None)
or getattr(usage, "output_tokens", None)
or (usage.get("completion_tokens") if isinstance(usage, dict) else None)
or 0
)
return {"input": input_tokens, "output": output_tokens, "unit": "TOKENS"}
def invoke_agent(
system_prompt: str,
state: dict,
node_name: str,
*,
extra_messages: Optional[list] = None,
context_window: int = 10,
) -> dict:
cost_tracker = _runtime.cost_tracker
circuit_breaker = _runtime.circuit_breaker
circuit_events = _runtime.circuit_events
export_ui_state_fn = _runtime.export_ui_state_fn
langfuse = get_langfuse()
trace_id = state.get("langfuse_trace_id")
if langfuse.is_enabled() and not trace_id:
trace = langfuse.create_trace(
name=f"auto-swe-agent",
metadata={
"task": state.get("current_task", "unknown")[:200],
"workspace": state.get("workspace_dir", "unknown"),
"mode": "multi-agent",
},
)
if trace is not None and hasattr(trace, "id"):
trace_id = trace.id
trimmed = []
for msg in state["messages"][-context_window:]:
if (
hasattr(msg, "content")
and isinstance(msg.content, str)
and len(msg.content) > 4000
):
from langchain_core.messages import ToolMessage
if isinstance(msg, ToolMessage):
msg = ToolMessage(
content=msg.content[:4000] + "\n[TRUNCATED]",
tool_call_id=msg.tool_call_id,
)
trimmed.append(msg)
msgs = [SystemMessage(content=system_prompt)] + (extra_messages or []) + trimmed
last_input = trimmed[-1].content if trimmed else ""
agent_span = None
if langfuse.is_enabled() and trace_id:
agent_span = langfuse.span(
trace_id=trace_id,
name=f"agent-{node_name}",
input={
"messages_count": len(msgs),
"context_window": context_window,
"last_input_preview": str(last_input)[:300],
},
)
for model in _runtime.fallback_models:
if not _model_available(model):
print(f"[SKIP] {model} — no API key set.")
continue
if not circuit_breaker.can_call(model):
event = f"[CIRCUIT OPEN] Skipping {model} (cooldown active)"
print(event)
circuit_events.append(event)
continue
print(f"\n--- [NODE] {node_name.upper()} | model={model} ---")
try:
response = _call_with_retry(
model,
msgs,
max_retries=state.get("_retry_max", 3),
base_delay=state.get("_retry_delay", 2.0),
)
circuit_breaker.record_success(model)
usage_dict = _extract_usage(response)
if usage_dict:
input_tokens = usage_dict["input"]
output_tokens = usage_dict["output"]
estimated = False
else:
input_tokens = len(msgs) * 500
output_tokens = len(str(response.content)) // 4
estimated = True
print(
f"[COST] Token counts unavailable — using estimates (in={input_tokens}, out={output_tokens})"
)
# Langfuse generation trace
if langfuse.is_enabled() and trace_id:
gen_params = {
"trace_id": trace_id,
"name": f"llm-{node_name}",
"model": model,
"input": str(last_input)[:500],
"output": (
str(response.content)[:1000]
if hasattr(response, "content")
else str(response)[:1000]
),
}
if usage_dict:
gen_params["usage"] = usage_dict
langfuse.generation(**gen_params)
cost_tracker.add_call(
model, input_tokens, output_tokens, node_name, estimated
)
total_cost = cost_tracker.get_total_cost()
print(
f"[COST] ${total_cost:.6f} total | this call: in={input_tokens} out={output_tokens} tokens"
)
if agent_span is not None:
agent_span.update(output={"status": "success", "model_used": model})
if cost_tracker.check_budget_exceeded():
print(
f"[COST] Budget exceeded (${total_cost:.4f} > ${cost_tracker.budget_usd})."
)
budget_msg = SystemMessage(
content=f"Budget exceeded (${total_cost:.4f} > ${cost_tracker.budget_usd}). Halting."
)
result = {
"messages": [response, budget_msg],
"iteration_count": state["iteration_count"] + 1,
"total_cost_usd": total_cost,
"budget_exceeded": True,
"tests_passed": False,
"current_node": node_name,
"current_agent": node_name,
}
if trace_id:
result["langfuse_trace_id"] = trace_id
if export_ui_state_fn:
export_ui_state_fn({**state, **result}, node_name)
return result
result = {
"messages": [response],
"iteration_count": state["iteration_count"] + 1,
"total_cost_usd": total_cost,
"budget_exceeded": False,
"current_node": node_name,
"current_agent": node_name,
}
if trace_id:
result["langfuse_trace_id"] = trace_id
if export_ui_state_fn:
export_ui_state_fn({**state, **result}, node_name)
return result
except Exception as e:
err_name = type(e).__name__
is_permanent = (
any(t in err_name for t in _SKIP_ERRORS)
or "Missing" in str(e)
or "key" in str(e).lower()
)
if not is_permanent:
circuit_breaker.record_failure(model)
status = circuit_breaker.get_status().get(model, {})
if status.get("state") == "open":
event = f"[CIRCUIT OPENED] {model} after {status.get('failures')} failures"
circuit_events.append(event)
print(f"[FALLBACK] {model} failed: {err_name}. Trying next model...")
if agent_span is not None:
agent_span.update(
output={"status": "fallback", "error": err_name, "model": model}
)
continue
if agent_span is not None:
agent_span.update(output={"status": "error", "error": "All models exhausted"})
raise RuntimeError("All models in fallback chain exhausted.")
|