Spaces:
Configuration error
Configuration error
File size: 8,101 Bytes
77bcbf1 | 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 | """
CASCADE SDK - Universal AI Observation Layer
Usage:
import cascade
cascade.init()
# Now every call emits a receipt automatically
import openai
response = openai.chat.completions.create(...) # Receipt emitted
"""
import threading
import queue
from typing import Optional, Dict, Any, List
from datetime import datetime, timezone
# Import our observation infrastructure
from .observation import ObservationManager
from .identity import ModelRegistry
from .genesis import ProvenanceChain
class CascadeSDK:
"""Main SDK singleton - manages patching and emission."""
_instance = None
_initialized = False
def __new__(cls):
if cls._instance is None:
cls._instance = super().__new__(cls)
return cls._instance
def __init__(self):
if CascadeSDK._initialized:
return
self.observation_manager = ObservationManager()
self.model_registry = ModelRegistry()
self.emission_queue = queue.Queue()
self.background_thread = None
self.running = False
self.patched_providers = set()
self.config = {
"emit_async": True,
"lattice_path": "lattice/observations",
"verbose": False,
}
CascadeSDK._initialized = True
def init(self, **kwargs):
"""
Initialize CASCADE and auto-patch available providers.
Args:
emit_async: Whether to emit receipts in background (default: True)
verbose: Print when receipts are emitted (default: False)
providers: List of providers to patch, or 'all' (default: 'all')
"""
self.config.update(kwargs)
# Start background emission thread
if self.config["emit_async"] and not self.running:
self.running = True
self.background_thread = threading.Thread(
target=self._emission_worker,
daemon=True
)
self.background_thread.start()
# Auto-patch available providers
providers = kwargs.get("providers", "all")
self._patch_providers(providers)
if self.config["verbose"]:
print(f"[CASCADE] Initialized. Patched: {self.patched_providers}")
return self
def _patch_providers(self, providers):
"""Patch LLM provider libraries."""
from .patches import (
patch_openai,
patch_anthropic,
patch_huggingface,
patch_ollama,
patch_litellm,
)
patch_map = {
"openai": patch_openai,
"anthropic": patch_anthropic,
"huggingface": patch_huggingface,
"ollama": patch_ollama,
"litellm": patch_litellm,
}
if providers == "all":
providers = list(patch_map.keys())
for provider in providers:
if provider in patch_map:
try:
patch_map[provider](self)
self.patched_providers.add(provider)
except ImportError:
# Provider not installed, skip
pass
except Exception as e:
if self.config["verbose"]:
print(f"[CASCADE] Failed to patch {provider}: {e}")
def _emission_worker(self):
"""Background thread that processes emission queue."""
while self.running:
try:
receipt_data = self.emission_queue.get(timeout=1.0)
self._emit_receipt(receipt_data)
except queue.Empty:
continue
except Exception as e:
if self.config["verbose"]:
print(f"[CASCADE] Emission error: {e}")
def _emit_receipt(self, receipt_data: Dict[str, Any]):
"""Actually write the receipt to lattice."""
import hashlib
import uuid
try:
# Create provenance chain for this observation
model_id = receipt_data["model_id"]
input_text = receipt_data["input"][:1000] # Truncate
output_text = receipt_data["output"][:2000] # Truncate
# Compute hashes
input_hash = hashlib.sha256(input_text.encode()).hexdigest()[:16]
model_hash = hashlib.sha256(model_id.encode()).hexdigest()[:16]
session_id = str(uuid.uuid4())[:8]
chain = ProvenanceChain(
session_id=session_id,
model_id=model_id,
model_hash=model_hash,
input_hash=input_hash,
)
# Add inference record
from cascade.core.provenance import ProvenanceRecord
import time
record = ProvenanceRecord(
layer_name="inference",
layer_idx=0,
state_hash=hashlib.sha256(output_text.encode()).hexdigest()[:16],
parent_hashes=[input_hash],
params_hash=model_hash,
shape=[len(output_text)],
dtype="text",
stats={
**receipt_data.get("metrics", {}),
"provider": receipt_data.get("context", {}).get("provider", "unknown"),
"timestamp": receipt_data.get("timestamp", datetime.now(timezone.utc).isoformat()),
},
execution_order=0,
)
chain.add_record(record)
chain.finalize()
observation = self.observation_manager.observe_model(
model_id=model_id,
chain=chain,
user_hash=receipt_data.get("user_hash"),
)
if self.config["verbose"]:
print(f"[CASCADE] Receipt: {observation.merkle_root[:16]}... -> {model_id}")
return observation
except Exception as e:
if self.config["verbose"]:
import traceback
print(f"[CASCADE] Failed to emit: {e}")
traceback.print_exc()
return None
def observe(
self,
model_id: str,
input_data: Any,
output_data: Any,
metrics: Optional[Dict] = None,
context: Optional[Dict] = None
):
"""
Manually emit an observation receipt.
Called automatically by patches, but can be called directly.
"""
receipt_data = {
"model_id": model_id,
"input": str(input_data),
"output": str(output_data),
"metrics": metrics or {},
"context": context or {},
"timestamp": datetime.now(timezone.utc).isoformat(),
}
if self.config["emit_async"]:
self.emission_queue.put(receipt_data)
else:
self._emit_receipt(receipt_data)
def shutdown(self):
"""Stop background emission and flush queue."""
self.running = False
if self.background_thread:
self.background_thread.join(timeout=5.0)
# Flush remaining items
while not self.emission_queue.empty():
try:
receipt_data = self.emission_queue.get_nowait()
self._emit_receipt(receipt_data)
except queue.Empty:
break
# Global SDK instance
_sdk = CascadeSDK()
def init(**kwargs):
"""Initialize CASCADE observation layer."""
return _sdk.init(**kwargs)
def observe(model_id: str, input_data: Any, output_data: Any, **kwargs):
"""Manually emit an observation."""
return _sdk.observe(model_id, input_data, output_data, **kwargs)
def shutdown():
"""Shutdown CASCADE (flush pending receipts)."""
return _sdk.shutdown()
# Convenience: allow `import cascade; cascade.init()`
__all__ = ["init", "observe", "shutdown", "CascadeSDK"]
|