File size: 19,645 Bytes
a54fd97 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 | """
MCP Protocol Handler - JSON-RPC 2.0 over SSE
Implements the Model Context Protocol for remote clients like Claude Desktop.
"""
import json
import asyncio
from typing import Any, Optional, AsyncGenerator
from dataclasses import dataclass, asdict
from .auth.models import User, MemoryEntry
from .database.vector_store import MultiTenantVectorStore
from .core.memory_builder import MemoryBuilder
from .core.retriever import Retriever
from .core.answer_generator import AnswerGenerator
# Type alias for client manager (supports both OpenRouter and Ollama)
ClientManager = object # Duck-typed: can be OpenRouterClientManager or OllamaClientManager
@dataclass
class JsonRpcRequest:
jsonrpc: str
method: str
id: Optional[int | str]
params: Optional[dict] = None
@dataclass
class JsonRpcResponse:
jsonrpc: str = "2.0"
id: Optional[int | str] = None
result: Optional[Any] = None
error: Optional[dict] = None
def to_dict(self):
d = {"jsonrpc": self.jsonrpc, "id": self.id}
if self.error:
d["error"] = self.error
else:
d["result"] = self.result
return d
# MCP Protocol Constants
MCP_VERSION = "2025-03-26" # Streamable HTTP transport
SERVER_NAME = "simplemem"
SERVER_VERSION = "1.0.0"
class MCPHandler:
"""
Handles MCP protocol messages for a specific user session.
"""
def __init__(
self,
user: User,
api_key: str,
vector_store: MultiTenantVectorStore,
client_manager: ClientManager,
settings: Any,
):
self.user = user
self.api_key = api_key
self.vector_store = vector_store
self.client_manager = client_manager
self.settings = settings
self.initialized = False
# Lazy-loaded components
self._memory_builder: Optional[MemoryBuilder] = None
self._retriever: Optional[Retriever] = None
self._answer_generator: Optional[AnswerGenerator] = None
def _get_client(self):
return self.client_manager.get_client(self.api_key)
def _get_memory_builder(self) -> MemoryBuilder:
if not self._memory_builder:
self._memory_builder = MemoryBuilder(
llm_client=self._get_client(),
vector_store=self.vector_store,
table_name=self.user.table_name,
window_size=self.settings.window_size,
overlap_size=self.settings.overlap_size,
temperature=self.settings.llm_temperature,
)
return self._memory_builder
def _get_retriever(self) -> Retriever:
if not self._retriever:
self._retriever = Retriever(
llm_client=self._get_client(),
vector_store=self.vector_store,
table_name=self.user.table_name,
semantic_top_k=self.settings.semantic_top_k,
keyword_top_k=self.settings.keyword_top_k,
enable_planning=self.settings.enable_planning,
enable_reflection=self.settings.enable_reflection,
max_reflection_rounds=self.settings.max_reflection_rounds,
temperature=self.settings.llm_temperature,
)
return self._retriever
def _get_answer_generator(self) -> AnswerGenerator:
if not self._answer_generator:
self._answer_generator = AnswerGenerator(
llm_client=self._get_client(),
temperature=self.settings.llm_temperature,
)
return self._answer_generator
async def handle_message(self, message: str) -> str:
"""Handle a JSON-RPC message and return response"""
try:
data = json.loads(message)
request = JsonRpcRequest(
jsonrpc=data.get("jsonrpc", "2.0"),
method=data.get("method", ""),
id=data.get("id"),
params=data.get("params", {}),
)
response = await self._dispatch(request)
return json.dumps(response.to_dict(), ensure_ascii=False)
except json.JSONDecodeError as e:
return json.dumps(JsonRpcResponse(
error={"code": -32700, "message": f"Parse error: {e}"}
).to_dict())
except Exception as e:
return json.dumps(JsonRpcResponse(
error={"code": -32603, "message": f"Internal error: {e}"}
).to_dict())
async def _dispatch(self, request: JsonRpcRequest) -> JsonRpcResponse:
"""Dispatch request to appropriate handler"""
method = request.method
params = request.params or {}
handlers = {
"initialize": self._handle_initialize,
"initialized": self._handle_initialized,
"ping": self._handle_ping,
"tools/list": self._handle_tools_list,
"tools/call": self._handle_tools_call,
"resources/list": self._handle_resources_list,
"resources/read": self._handle_resources_read,
}
handler = handlers.get(method)
if not handler:
return JsonRpcResponse(
id=request.id,
error={"code": -32601, "message": f"Method not found: {method}"}
)
try:
result = await handler(params)
return JsonRpcResponse(id=request.id, result=result)
except Exception as e:
return JsonRpcResponse(
id=request.id,
error={"code": -32603, "message": str(e)}
)
async def _handle_initialize(self, params: dict) -> dict:
"""Handle initialize request"""
self.initialized = True
return {
"protocolVersion": MCP_VERSION,
"capabilities": {
"tools": {},
"resources": {},
},
"serverInfo": {
"name": SERVER_NAME,
"version": SERVER_VERSION,
"description": "SimpleMem - Advanced Lifelong Memory System for LLM Agents. "
"Features: Semantic lossless compression, coreference resolution, "
"temporal anchoring, hybrid retrieval (semantic + lexical + symbolic), "
"and intelligent query planning with reflection.",
},
"instructions": """SimpleMem is your long-term memory system. Use it to:
1. STORE conversations: Use memory_add or memory_add_batch to save dialogues.
The system automatically extracts facts, resolves pronouns, and anchors timestamps.
Memories are stored immediately - no manual flush needed.
2. RECALL information: Use memory_query to ask questions about past conversations.
The system retrieves relevant memories and synthesizes answers.
3. BROWSE memories: Use memory_retrieve to see raw stored facts with metadata.
4. MANAGE: Use memory_stats to check status, memory_clear to reset (careful!).
Tips:
- Use memory_query for natural questions, memory_retrieve for browsing
- Each memory_add call processes and stores data immediately""",
}
async def _handle_initialized(self, params: dict) -> dict:
"""Handle initialized notification"""
return {}
async def _handle_ping(self, params: dict) -> dict:
"""Handle ping request"""
return {}
async def _handle_tools_list(self, params: dict) -> dict:
"""Handle tools/list request"""
return {
"tools": [
{
"name": "memory_add",
"description": """Add a dialogue to SimpleMem long-term memory system.
SimpleMem is an advanced lifelong memory system that:
- Stores conversations as atomic, self-contained facts (no pronouns, absolute timestamps)
- Uses semantic compression to extract key information (persons, locations, entities, topics)
- Supports hybrid retrieval (semantic + keyword + metadata filtering)
The dialogue is processed immediately by LLM and stored. No manual flush needed.
Example: memory_add(speaker="Alice", content="I'll meet Bob at Starbucks tomorrow at 2pm")
→ Stored as: "Alice will meet Bob at Starbucks on 2025-01-14 at 14:00"
with metadata: persons=["Alice","Bob"], location="Starbucks", topic="Meeting arrangement\"""",
"inputSchema": {
"type": "object",
"properties": {
"speaker": {
"type": "string",
"description": "Name of the speaker (will be used for coreference resolution)",
},
"content": {
"type": "string",
"description": "Content of the dialogue (pronouns will be resolved, relative times will be converted to absolute)",
},
"timestamp": {
"type": "string",
"description": "ISO 8601 timestamp of when this was said (used for temporal anchoring). Defaults to now.",
},
},
"required": ["speaker", "content"],
},
},
{
"name": "memory_add_batch",
"description": """Add multiple dialogues to SimpleMem at once.
Efficient for importing conversation history. Each dialogue is processed with:
- Coreference resolution (he/she → actual names)
- Temporal anchoring (tomorrow → actual date)
- Entity extraction (persons, locations, organizations)
All dialogues are processed immediately and stored. No manual flush needed.""",
"inputSchema": {
"type": "object",
"properties": {
"dialogues": {
"type": "array",
"description": "List of dialogues to add",
"items": {
"type": "object",
"properties": {
"speaker": {"type": "string", "description": "Speaker name"},
"content": {"type": "string", "description": "Dialogue content"},
"timestamp": {"type": "string", "description": "ISO 8601 timestamp"},
},
"required": ["speaker", "content"],
},
},
},
"required": ["dialogues"],
},
},
{
"name": "memory_query",
"description": """Query SimpleMem and get an AI-generated answer based on stored memories.
This is the primary way to retrieve information from long-term memory. The system:
1. Analyzes query complexity (simple fact vs multi-hop reasoning)
2. Generates targeted search queries
3. Performs hybrid retrieval (semantic similarity + keyword matching + metadata filtering)
4. Optionally reflects to find missing information for complex queries
5. Synthesizes a concise answer from retrieved contexts
Best for: "When did Alice and Bob plan to meet?", "What does Alice think about the project?", "Summarize recent events with Bob"
Returns: answer, reasoning, confidence level, and number of memory entries used.""",
"inputSchema": {
"type": "object",
"properties": {
"question": {
"type": "string",
"description": "Natural language question about stored memories",
},
"enable_reflection": {
"type": "boolean",
"description": "Enable iterative refinement for complex multi-hop queries. Default: true. Disable for simple factual lookups to save tokens.",
},
},
"required": ["question"],
},
},
{
"name": "memory_retrieve",
"description": """Retrieve relevant memory entries without generating an answer.
Returns raw memory entries with full metadata. Use this when you need:
- Direct access to stored facts
- To process/analyze memories yourself
- To show the user what's stored about a topic
Each entry contains: content (self-contained fact), timestamp, location, persons, entities, topic.""",
"inputSchema": {
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "Search query (semantic search + keyword matching)",
},
"top_k": {
"type": "integer",
"description": "Maximum number of entries to return. Default: 10",
},
},
"required": ["query"],
},
},
{
"name": "memory_clear",
"description": """Clear ALL memories for this user. This action CANNOT be undone.
Use with caution. This removes all stored memory entries from the vector database.""",
"inputSchema": {
"type": "object",
"properties": {},
},
},
{
"name": "memory_stats",
"description": """Get statistics about the memory store.
Returns:
- Total number of stored memory entries
- User ID and table info
Use to check if memories are being stored correctly.""",
"inputSchema": {
"type": "object",
"properties": {},
},
},
]
}
async def _handle_tools_call(self, params: dict) -> dict:
"""Handle tools/call request"""
name = params.get("name", "")
arguments = params.get("arguments", {})
tool_handlers = {
"memory_add": self._tool_memory_add,
"memory_add_batch": self._tool_memory_add_batch,
"memory_query": self._tool_memory_query,
"memory_retrieve": self._tool_memory_retrieve,
"memory_clear": self._tool_memory_clear,
"memory_stats": self._tool_memory_stats,
}
handler = tool_handlers.get(name)
if not handler:
raise ValueError(f"Unknown tool: {name}")
result = await handler(arguments)
return {
"content": [
{
"type": "text",
"text": json.dumps(result, ensure_ascii=False, indent=2),
}
]
}
async def _tool_memory_add(self, args: dict) -> dict:
builder = self._get_memory_builder()
return await builder.add_dialogue(
speaker=args["speaker"],
content=args["content"],
timestamp=args.get("timestamp"),
)
async def _tool_memory_add_batch(self, args: dict) -> dict:
builder = self._get_memory_builder()
return await builder.add_dialogues(
dialogues=args["dialogues"],
)
async def _tool_memory_query(self, args: dict) -> dict:
retriever = self._get_retriever()
generator = self._get_answer_generator()
contexts = await retriever.retrieve(
query=args["question"],
enable_reflection=args.get("enable_reflection", True),
)
answer_result = await generator.generate_answer(
query=args["question"],
contexts=contexts,
)
return {
"question": args["question"],
"answer": answer_result["answer"],
"reasoning": answer_result["reasoning"],
"confidence": answer_result["confidence"],
"contexts_used": len(contexts),
}
async def _tool_memory_retrieve(self, args: dict) -> dict:
retriever = self._get_retriever()
top_k = args.get("top_k", 10)
entries = await retriever.retrieve(
query=args["query"],
enable_reflection=False,
)
return {
"query": args["query"],
"results": [
{
"content": e.lossless_restatement,
"timestamp": e.timestamp,
"location": e.location,
"persons": e.persons,
"entities": e.entities,
"topic": e.topic,
}
for e in entries[:top_k]
],
"total": len(entries),
}
async def _tool_memory_clear(self, args: dict) -> dict:
success = await self.vector_store.clear_table(self.user.table_name)
return {
"success": success,
"message": "All memories cleared" if success else "Failed",
}
async def _tool_memory_stats(self, args: dict) -> dict:
stats = self.vector_store.get_stats(self.user.table_name)
builder = self._get_memory_builder()
builder_stats = builder.get_stats()
return {
"user_id": self.user.user_id,
"entry_count": stats.get("entry_count", 0),
"total_dialogues_processed": builder_stats.get("total_dialogues_processed", 0),
}
async def _handle_resources_list(self, params: dict) -> dict:
"""Handle resources/list request"""
return {
"resources": [
{
"uri": f"memory://{self.user.user_id}/stats",
"name": "Memory Statistics",
"description": "Statistics about your memory store",
"mimeType": "application/json",
},
{
"uri": f"memory://{self.user.user_id}/all",
"name": "All Memories",
"description": "All stored memory entries",
"mimeType": "application/json",
},
]
}
async def _handle_resources_read(self, params: dict) -> dict:
"""Handle resources/read request"""
uri = params.get("uri", "")
if uri.endswith("/stats"):
stats = self.vector_store.get_stats(self.user.table_name)
content = json.dumps(stats, ensure_ascii=False)
elif uri.endswith("/all"):
entries = await self.vector_store.get_all_entries(self.user.table_name)
content = json.dumps({
"entries": [e.to_dict() for e in entries],
"total": len(entries),
}, ensure_ascii=False)
else:
raise ValueError(f"Unknown resource: {uri}")
return {
"contents": [
{
"uri": uri,
"mimeType": "application/json",
"text": content,
}
]
}
|