File size: 11,969 Bytes
9abace2 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 | """
Paradigm 4: Cyclic Bitmaps & Shared Memory (ctypes + multiprocessing)
Part of SOVEREIGN PYTHON LLM ENGINE
Zero-copy state sharing between orchestrator and daemon processes.
RAM-speed, no serialization, no disk I/O, no TCP overhead.
Two backends:
A. ctypes.create_string_buffer()
- Single process / multiple threads
- Direct memory pointer sharing
- Fastest possible: byte mutation at array index
B. multiprocessing.shared_memory (Python 3.8+ stdlib)
- Cross-process on same machine
- Named shared memory block, accessible by name
- Survives across process restarts (by name)
Memory layout (fixed 4KB block):
Offset Size Field
0 8 MAGIC = b'SOVSTATE'
8 8 version (uint64 BE) — incremented on every write
16 8 agent_id hash (uint64 BE) — FNV-1a of agent_id string
24 8 step (uint64 BE)
32 8 timestamp_ns (uint64 BE)
40 8 bitmask (uint64 BE) — 64 boolean flags
48 8 sequence (uint64 BE) — monotonic write counter
56 8 checksum (uint64 BE) — FNV-1a of bytes 0-55
64 960 payload (raw bytes, caller-defined)
1024 3072 reserved
Total fixed block = 4096 bytes (one memory page)
"""
from __future__ import annotations
import ctypes
import struct
import time
from pathlib import Path
from typing import Any
# ─────────────────────────────────────────────
# Constants
# ─────────────────────────────────────────────
BLOCK_SIZE = 4096
HEADER_SIZE = 64
PAYLOAD_SIZE = 960
MAGIC = b'SOVSTATE'
HDR_FMT = '>8sQQQQQQQ' # magic(8) ver(8) id_hash(8) step(8) ts(8) mask(8) seq(8) chk(8)
HDR_BYTES = struct.calcsize(HDR_FMT) # should be 64
assert HDR_BYTES == HEADER_SIZE, f"Header size mismatch: {HDR_BYTES}"
# ─────────────────────────────────────────────
# FNV-1a hash (fast, no imports)
# ─────────────────────────────────────────────
def _fnv1a_64(data: bytes) -> int:
h = 0xcbf29ce484222325
for b in data:
h ^= b
h = (h * 0x100000001b3) & 0xFFFFFFFFFFFFFFFF
return h
# ─────────────────────────────────────────────
# Shared State Block (in-process, ctypes buffer)
# ─────────────────────────────────────────────
class SharedStateBlock:
"""
Single 4KB memory block for agent state.
Uses ctypes.create_string_buffer — works within one process
across threads with no locking needed for reads
(writes are single-byte atomic on x86/ARM).
Usage (single process, multiple threads):
block = SharedStateBlock(agent_id="react_1")
block.set_flag(2) # set bit 2
block.set_step(5)
block.write_payload(b"custom data")
# From another thread:
print(block.get_flags()) # fast bitmask read
print(block.get_step())
"""
def __init__(self, agent_id: str, block_size: int = BLOCK_SIZE):
self._agent_id = agent_id
self._id_hash = _fnv1a_64(agent_id.encode('utf-8'))
self._version = 0
self._sequence = 0
self._buf = ctypes.create_string_buffer(block_size)
self._block_size = block_size
self._flush_header(mask=0, step=0)
def _flush_header(self, mask: int, step: int) -> None:
self._version += 1
self._sequence += 1
ts = time.time_ns()
# Build header without checksum first
hdr_no_chk = struct.pack(
'>8sQQQQQQ',
MAGIC,
self._version,
self._id_hash,
step,
ts,
mask,
self._sequence
)
checksum = _fnv1a_64(hdr_no_chk)
hdr = hdr_no_chk + struct.pack('>Q', checksum)
# Write to ctypes buffer
self._buf[:HEADER_SIZE] = hdr
def _read_header(self) -> tuple | None:
raw = bytes(self._buf[:HEADER_SIZE])
try:
(magic, ver, id_hash, step, ts, mask, seq, chk) = struct.unpack(HDR_FMT, raw)
if magic != MAGIC:
return None
# Verify checksum
expected = _fnv1a_64(raw[:56])
if chk != expected:
return None
return (ver, id_hash, step, ts, mask, seq)
except Exception:
return None
# ── Flag operations ────────────────────────
def set_flag(self, bit: int) -> None:
hdr = self._read_header()
mask = hdr[4] if hdr else 0
step = hdr[2] if hdr else 0
mask |= (1 << (bit & 63))
self._flush_header(mask=mask, step=step)
def clear_flag(self, bit: int) -> None:
hdr = self._read_header()
mask = hdr[4] if hdr else 0
step = hdr[2] if hdr else 0
mask &= ~(1 << (bit & 63))
self._flush_header(mask=mask, step=step)
def get_flags(self) -> int:
hdr = self._read_header()
return hdr[4] if hdr else 0
def has_flag(self, bit: int) -> bool:
return bool(self.get_flags() & (1 << (bit & 63)))
def set_flags_from_set(self, bits: set[int]) -> None:
mask = sum(1 << (b & 63) for b in bits)
hdr = self._read_header()
step = hdr[2] if hdr else 0
self._flush_header(mask=mask, step=step)
def active_bits(self) -> set[int]:
mask = self.get_flags()
return {i for i in range(64) if mask & (1 << i)}
# ── Step ────────────────────────────────────
def set_step(self, step: int) -> None:
hdr = self._read_header()
mask = hdr[4] if hdr else 0
self._flush_header(mask=mask, step=step)
def get_step(self) -> int:
hdr = self._read_header()
return hdr[2] if hdr else 0
# ── Payload ─────────────────────────────────
def write_payload(self, data: bytes) -> None:
n = min(len(data), PAYLOAD_SIZE)
self._buf[HEADER_SIZE:HEADER_SIZE + n] = data[:n]
# Zero rest
if n < PAYLOAD_SIZE:
self._buf[HEADER_SIZE + n:HEADER_SIZE + PAYLOAD_SIZE] = b'\x00' * (PAYLOAD_SIZE - n)
def read_payload(self, length: int | None = None) -> bytes:
n = length if length is not None else PAYLOAD_SIZE
n = min(n, PAYLOAD_SIZE)
return bytes(self._buf[HEADER_SIZE:HEADER_SIZE + n])
# ── Snapshot ────────────────────────────────
def snapshot(self) -> dict[str, Any]:
hdr = self._read_header()
if not hdr:
return {"valid": False}
ver, id_hash, step, ts, mask, seq = hdr
return {
"valid": True,
"agent_id": self._agent_id,
"version": ver,
"step": step,
"timestamp_ns": ts,
"flags_mask": mask,
"active_bits": list(self.active_bits()),
"sequence": seq,
}
def raw_bytes(self) -> bytes:
return bytes(self._buf[:self._block_size])
# ─────────────────────────────────────────────
# Cross-process shared memory (multiprocessing.shared_memory)
# ─────────────────────────────────────────────
class SharedMemoryState:
"""
Cross-process state block using Python 3.8+ multiprocessing.shared_memory.
The memory block is identified by name — any process on the same
machine can attach to it by name.
Survives process restarts: if the block already exists,
the new process attaches and reads existing state.
Usage:
# Process A (orchestrator)
sm = SharedMemoryState("sovereign_react_1", create=True)
sm.set_step(5)
sm.set_flag(2)
# Process B (daemon) — attaches to existing block
sm2 = SharedMemoryState("sovereign_react_1", create=False)
print(sm2.get_step()) # 5
print(sm2.has_flag(2)) # True
"""
def __init__(self, name: str, create: bool = True, size: int = BLOCK_SIZE):
from multiprocessing import shared_memory
self._name = name
self._size = size
try:
if create:
try:
self._shm = shared_memory.SharedMemory(name=name, create=True, size=size)
# Initialize with zeros
self._shm.buf[:size] = b'\x00' * size
except FileExistsError:
# Already exists — attach
self._shm = shared_memory.SharedMemory(name=name, create=False, size=size)
else:
self._shm = shared_memory.SharedMemory(name=name, create=False, size=size)
except Exception as e:
# Fallback to ctypes buffer if shared memory unavailable
self._shm = None
self._fallback = ctypes.create_string_buffer(size)
self._fallback_mode = True
return
self._fallback_mode = False
def _read(self, offset: int, length: int) -> bytes:
if self._fallback_mode:
return bytes(self._fallback[offset:offset + length])
return bytes(self._shm.buf[offset:offset + length])
def _write(self, offset: int, data: bytes) -> None:
if self._fallback_mode:
self._fallback[offset:offset + len(data)] = data
else:
self._shm.buf[offset:offset + len(data)] = data
def set_step(self, step: int) -> None:
self._write(24, struct.pack('>Q', step))
def get_step(self) -> int:
raw = self._read(24, 8)
return struct.unpack('>Q', raw)[0]
def set_flag(self, bit: int) -> None:
raw = self._read(40, 8)
mask = struct.unpack('>Q', raw)[0]
mask |= (1 << (bit & 63))
self._write(40, struct.pack('>Q', mask))
def clear_flag(self, bit: int) -> None:
raw = self._read(40, 8)
mask = struct.unpack('>Q', raw)[0]
mask &= ~(1 << (bit & 63))
self._write(40, struct.pack('>Q', mask))
def has_flag(self, bit: int) -> bool:
raw = self._read(40, 8)
mask = struct.unpack('>Q', raw)[0]
return bool(mask & (1 << (bit & 63)))
def get_flags(self) -> int:
raw = self._read(40, 8)
return struct.unpack('>Q', raw)[0]
def write_payload(self, data: bytes) -> None:
n = min(len(data), PAYLOAD_SIZE)
self._write(HEADER_SIZE, data[:n])
def read_payload(self, length: int | None = None) -> bytes:
n = length if length is not None else PAYLOAD_SIZE
return self._read(HEADER_SIZE, min(n, PAYLOAD_SIZE))
def close(self) -> None:
if not self._fallback_mode and self._shm:
self._shm.close()
def unlink(self) -> None:
if not self._fallback_mode and self._shm:
self._shm.unlink()
@property
def name(self) -> str:
return self._name
|