Spaces:
Sleeping
Sleeping
File size: 9,649 Bytes
ce675d4 d6fbf54 ce675d4 d6fbf54 ce675d4 d6fbf54 ce675d4 d6fbf54 ce675d4 | 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 | from __future__ import annotations
import copy
import json
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any
from osint_env.domain.models import (
EnvironmentConfig,
LLMConfig,
NodeType,
SeedingConfig,
SeedEdgeSpec,
SeedNodeSpec,
SeedQuestionSpec,
SpawnRewardConfig,
SwarmConfig,
)
@dataclass(slots=True)
class RuntimeDefaults:
default_episodes: int = 20
leaderboard_path: str = "artifacts/leaderboard.json"
dashboard_path: str = "artifacts/osint_dashboard.html"
sweep_dashboard_dir: str = "artifacts/sweep_dashboards"
@dataclass(slots=True)
class SharedConfig:
environment: EnvironmentConfig = field(default_factory=EnvironmentConfig)
runtime: RuntimeDefaults = field(default_factory=RuntimeDefaults)
def clone_environment_config(config: EnvironmentConfig) -> EnvironmentConfig:
return copy.deepcopy(config)
def _as_dict(value: Any) -> dict[str, Any]:
return value if isinstance(value, dict) else {}
def _parse_int(value: Any, default: int) -> int:
try:
return int(value)
except (TypeError, ValueError):
return default
def _parse_float(value: Any, default: float) -> float:
try:
return float(value)
except (TypeError, ValueError):
return default
def _parse_bool(value: Any, default: bool) -> bool:
if isinstance(value, bool):
return value
if isinstance(value, str):
lowered = value.strip().lower()
if lowered in {"1", "true", "yes", "y", "on"}:
return True
if lowered in {"0", "false", "no", "n", "off"}:
return False
return default
def _infer_node_type(node_id: str) -> NodeType:
prefix = str(node_id).split("_", 1)[0].lower()
mapping = {
"user": NodeType.USER,
"alias": NodeType.ALIAS,
"org": NodeType.ORG,
"loc": NodeType.LOCATION,
"location": NodeType.LOCATION,
"post": NodeType.POST,
"thr": NodeType.THREAD,
"thread": NodeType.THREAD,
"event": NodeType.EVENT,
}
return mapping.get(prefix, NodeType.USER)
def _parse_node_type(value: Any, node_id: str) -> NodeType:
if isinstance(value, NodeType):
return value
if isinstance(value, str):
raw = value.strip().lower()
try:
return NodeType(raw)
except ValueError:
return _infer_node_type(node_id)
return _infer_node_type(node_id)
def _parse_seed_edge(item: dict[str, Any]) -> SeedEdgeSpec | None:
src = str(item.get("src", "")).strip()
rel = str(item.get("rel", "")).strip()
dst = str(item.get("dst", "")).strip()
if not src or not rel or not dst:
return None
confidence = _parse_float(item.get("confidence", 1.0), 1.0)
return SeedEdgeSpec(src=src, rel=rel, dst=dst, confidence=confidence)
def _parse_seeding(data: dict[str, Any]) -> SeedingConfig:
seeded_nodes: list[SeedNodeSpec] = []
for item in data.get("seeded_nodes", []):
row = _as_dict(item)
node_id = str(row.get("node_id", "")).strip()
if not node_id:
continue
node_type = _parse_node_type(row.get("node_type"), node_id)
attrs = _as_dict(row.get("attrs"))
seeded_nodes.append(SeedNodeSpec(node_id=node_id, node_type=node_type, attrs=attrs))
seeded_edges: list[SeedEdgeSpec] = []
for item in data.get("seeded_edges", []):
edge = _parse_seed_edge(_as_dict(item))
if edge is not None:
seeded_edges.append(edge)
seeded_questions: list[SeedQuestionSpec] = []
for item in data.get("seeded_questions", []):
row = _as_dict(item)
question = str(row.get("question", "")).strip()
if not question:
continue
answer_val = row.get("answer")
answer = str(answer_val).strip() if answer_val is not None and str(answer_val).strip() else None
task_type = str(row.get("task_type", "seeded")).strip() or "seeded"
support_edges: list[SeedEdgeSpec] = []
for edge_item in row.get("supporting_edges", []):
edge = _parse_seed_edge(_as_dict(edge_item))
if edge is not None:
support_edges.append(edge)
metadata = _as_dict(row.get("metadata"))
seeded_questions.append(
SeedQuestionSpec(
question=question,
answer=answer,
task_type=task_type,
supporting_edges=support_edges,
metadata=metadata,
)
)
return SeedingConfig(
seeded_nodes=seeded_nodes,
seeded_edges=seeded_edges,
seeded_questions=seeded_questions,
llm_generate_remaining_graph=_parse_bool(data.get("llm_generate_remaining_graph"), True),
llm_generate_remaining_tasks=_parse_bool(data.get("llm_generate_remaining_tasks"), True),
llm_generated_edge_budget=max(0, _parse_int(data.get("llm_generated_edge_budget"), 6)),
llm_generated_task_budget=max(0, _parse_int(data.get("llm_generated_task_budget"), 8)),
llm_generation_parallel=_parse_bool(data.get("llm_generation_parallel"), True),
llm_generation_workers=max(1, _parse_int(data.get("llm_generation_workers"), 3)),
llm_generation_retries=max(1, _parse_int(data.get("llm_generation_retries"), 2)),
allow_template_fallback_on_llm_failure=_parse_bool(
data.get("allow_template_fallback_on_llm_failure"),
False,
),
)
def load_seeding_config(path: str | Path) -> SeedingConfig:
payload = json.loads(Path(path).read_text(encoding="utf-8"))
if not isinstance(payload, dict):
raise ValueError("Seed file must contain a JSON object.")
source = _as_dict(payload.get("seeding", payload))
return _parse_seeding(source)
def _parse_environment(payload: dict[str, Any]) -> EnvironmentConfig:
env_data = _as_dict(payload.get("environment", payload))
swarm_data = _as_dict(payload.get("swarm", env_data.get("swarm", {})))
spawn_data = _as_dict(payload.get("spawn_reward", env_data.get("spawn_reward", {})))
seeding_data = _as_dict(payload.get("seeding", env_data.get("seeding", {})))
llm_data = _as_dict(payload.get("llm", env_data.get("llm", {})))
env = EnvironmentConfig(
n_users=max(4, _parse_int(env_data.get("n_users"), 40)),
alias_density=max(0.0, min(1.0, _parse_float(env_data.get("alias_density"), 0.35))),
noise_level=max(0.0, min(1.0, _parse_float(env_data.get("noise_level"), 0.15))),
red_herring_rate=max(0.0, min(1.0, _parse_float(env_data.get("red_herring_rate"), 0.1))),
max_steps=max(2, _parse_int(env_data.get("max_steps"), 18)),
seed=_parse_int(env_data.get("seed"), 7),
)
env.swarm = SwarmConfig(
enabled=_parse_bool(swarm_data.get("enabled"), False),
max_agents=max(1, _parse_int(swarm_data.get("max_agents"), 3)),
max_breadth=max(1, _parse_int(swarm_data.get("max_breadth"), 2)),
max_width=max(1, _parse_int(swarm_data.get("max_width"), 2)),
max_depth=max(1, _parse_int(swarm_data.get("max_depth"), 2)),
planner_rounds=max(1, _parse_int(swarm_data.get("planner_rounds"), 2)),
tools_per_agent=max(1, _parse_int(swarm_data.get("tools_per_agent"), 1)),
)
env.spawn_reward = SpawnRewardConfig(
lambda_parallel=max(0.0, _parse_float(spawn_data.get("lambda_parallel"), 0.15)),
lambda_finish=max(0.0, _parse_float(spawn_data.get("lambda_finish"), 0.2)),
anneal=max(0.0, min(1.0, _parse_float(spawn_data.get("anneal"), 1.0))),
max_parallel_hint=max(1, _parse_int(spawn_data.get("max_parallel_hint"), 3)),
)
env.seeding = _parse_seeding(seeding_data)
env.llm = LLMConfig(
provider=str(llm_data.get("provider", "mock")).strip() or "mock",
model=str(llm_data.get("model", "qwen3:2b")).strip() or "qwen3:2b",
temperature=_parse_float(llm_data.get("temperature"), 0.1),
max_tokens=max(1, _parse_int(llm_data.get("max_tokens"), 256)),
timeout_seconds=max(1, _parse_int(llm_data.get("timeout_seconds"), 240)),
ollama_base_url=str(llm_data.get("ollama_base_url", "http://127.0.0.1:11434")).strip()
or "http://127.0.0.1:11434",
openai_base_url=str(llm_data.get("openai_base_url", "https://api.openai.com/v1")).strip()
or "https://api.openai.com/v1",
openai_api_key_env=str(llm_data.get("openai_api_key_env", "OPENAI_API_KEY")).strip() or "OPENAI_API_KEY",
openai_api_key=str(llm_data.get("openai_api_key", "")).strip(),
)
return env
def _parse_runtime(payload: dict[str, Any]) -> RuntimeDefaults:
runtime = _as_dict(payload.get("runtime", {}))
return RuntimeDefaults(
default_episodes=max(1, _parse_int(runtime.get("default_episodes"), 20)),
leaderboard_path=str(runtime.get("leaderboard_path", "artifacts/leaderboard.json")),
dashboard_path=str(runtime.get("dashboard_path", "artifacts/osint_dashboard.html")),
sweep_dashboard_dir=str(runtime.get("sweep_dashboard_dir", "artifacts/sweep_dashboards")),
)
def load_shared_config(path: str | Path | None) -> SharedConfig:
if not path:
return SharedConfig()
file_path = Path(path)
if not file_path.exists():
return SharedConfig()
payload = json.loads(file_path.read_text(encoding="utf-8"))
if not isinstance(payload, dict):
raise ValueError("Shared config file must contain a JSON object.")
return SharedConfig(environment=_parse_environment(payload), runtime=_parse_runtime(payload))
|