File size: 13,986 Bytes
beb8990 | 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 | """
Generic HTTP Environment Server for MCP Integration.
This module provides a custom HTTP server that wraps any MCP environment
and exposes OpenEnv-compliant HTTP endpoints (/reset, /step, /state).
The server is fully generic and works with any MCP integration.
All MCP-specific configuration is loaded from config.py.
"""
import asyncio
import logging
import sqlite3
from typing import Any, Dict, Optional, List
from fastapi import Body, FastAPI, Request, Query
from openenv.core.env_server.http_server import HTTPEnvServer
from .config import (
MCP_NAME,
HTTP_HEADERS,
HTTP_HEADER_DEFAULTS,
get_session_manager_class,
get_seed_data_function
)
logger = logging.getLogger(__name__)
class MCPHTTPEnvServer(HTTPEnvServer):
"""
Generic HTTP Environment Server for any MCP integration.
This server wraps any MCP environment and provides HTTP endpoints
for OpenEnv integration. It's fully generic and works with any MCP.
HTTP headers are configured via config.py:HTTP_HEADERS, making it easy
to adapt to different MCP authentication and multi-tenancy patterns.
"""
def __init__(self, env, action_cls, observation_cls):
"""Initialize custom HTTP server with MCP session manager."""
# Store classes before calling super().__init__()
self.action_cls = action_cls
self.observation_cls = observation_cls
# Call parent init
super().__init__(env=env, action_cls=action_cls, observation_cls=observation_cls)
# Create a persistent environment instance for HTTP endpoints
# The parent class stores env_factory in self._env_factory
# We create one instance for the HTTP endpoints (not WebSocket)
if callable(self._env_factory):
self.env = self._env_factory()
else:
self.env = self._env_factory
# Dynamically load the session manager from config
SessionManagerClass = get_session_manager_class()
self.session_manager = SessionManagerClass()
def _get_header_value(self, headers: dict, header_key: str) -> Optional[str]:
"""
Get header value using configured header name.
Args:
headers: Request headers dictionary
header_key: Key in HTTP_HEADERS config (e.g., "database_id", "access_token")
Returns:
Header value or default from config
"""
header_name = HTTP_HEADERS.get(header_key)
if not header_name:
logger.warning(f"Header key '{header_key}' not configured in HTTP_HEADERS")
return HTTP_HEADER_DEFAULTS.get(header_key)
value = headers.get(header_name)
if value is None:
value = HTTP_HEADER_DEFAULTS.get(header_key)
return value
def register_routes(self, app: Any) -> None:
if not isinstance(app, FastAPI):
raise TypeError("app must be a FastAPI instance")
# Register custom reset endpoint
@app.post("/reset")
async def reset_with_database_refresh(
request: Request,
body: Optional[Dict[str, Any]] = Body(default=None)
) -> Dict[str, Any]:
"""
Reset the environment and optionally reset the database.
The database_id can be provided via:
1. Request body: {"database_id": "my_db", "sql_content": "INSERT INTO..."}
2. HTTP header: x-database-id
3. Default value if neither provided
Args (in request body, all optional):
database_id: Database identifier for multi-tenancy
sql_content: Custom SQL content to use for seeding instead of default
Returns:
Observation with reset status and database reset result
"""
headers = dict(request.headers)
# Get database_id from body first, then header, then default
body = body or {}
database_id = body.get("database_id") or self._get_header_value(headers, "database_id")
access_token = self._get_header_value(headers, "access_token")
# Get optional sql_content from body
sql_content = body.get("sql_content")
logger.info(f"Reset request for database_id={database_id}, custom_sql={'yes' if sql_content else 'no'}")
# Reset database to original state (with optional custom SQL)
db_reset_result = self._reset_database(database_id, sql_content=sql_content)
# Set request context in environment
self.env.set_request_context(
database_id=database_id,
access_token=access_token
)
# Execute reset in thread pool (environments may use sync code)
loop = asyncio.get_event_loop()
observation = await loop.run_in_executor(None, self.env.reset)
# Serialize observation manually
result = {
"observation": observation.model_dump() if hasattr(observation, 'model_dump') else observation.__dict__,
"done": getattr(observation, 'done', False),
"reward": getattr(observation, 'reward', 0.0)
}
# Add database reset info to observation metadata
if isinstance(result, dict) and "observation" in result:
if "metadata" not in result["observation"]:
result["observation"]["metadata"] = {}
result["observation"]["metadata"]["database_reset_result"] = db_reset_result
logger.info(f"Environment reset completed for database {database_id}, DB refresh: {db_reset_result['success']}")
return result
# Register custom step endpoint
@app.post("/step")
async def step_with_headers(request: Request, body: Dict[str, Any] = Body(...)) -> Dict[str, Any]:
# Extract headers using dynamic header names from config
headers = dict(request.headers)
database_id = self._get_header_value(headers, "database_id")
access_token = self._get_header_value(headers, "access_token")
# Debug logging to see what headers we're receiving
logger.info(f"Step request - database_id: {database_id}, has_access_token: {bool(access_token)}")
if not access_token:
logger.warning(f"No access token found in headers. Available headers: {list(headers.keys())}")
# Set request context in environment
self.env.set_request_context(database_id=database_id, access_token=access_token)
logger.debug(f"Step request with database_id={database_id}, has_token={bool(access_token)}")
# Support both {"action": {...}} and direct action fields
action_data = body.get("action", body)
# Deserialize action manually using Pydantic
try:
action = self.action_cls(**action_data)
except Exception as e:
logger.error(f"Failed to deserialize action: {e}")
return {
"observation": {
"success": False,
"error_message": f"Invalid action: {str(e)}",
"done": False,
"reward": -1.0,
"metadata": {}
},
"done": False,
"reward": -1.0
}
# Execute step in thread pool (environments may use sync code)
loop = asyncio.get_event_loop()
observation = await loop.run_in_executor(None, self.env.step, action)
# Serialize observation manually
result = {
"observation": observation.model_dump() if hasattr(observation, 'model_dump') else observation.__dict__,
"done": getattr(observation, 'done', False),
"reward": getattr(observation, 'reward', 0.0)
}
return result
# Register state endpoint
@app.get("/state")
async def get_state(
request: Request,
verify_queries: List[str] = Query(default=[])
) -> Dict[str, Any]:
headers = dict(request.headers)
database_id = headers.get("x-database-id", "default")
state = self.env.state
result = {
"episode_id": state.episode_id,
"step_count": state.step_count,
"database_id": database_id
}
db_path = self.session_manager.get_db_path(database_id)
try:
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
cursor = conn.cursor()
if verify_queries:
result["verification_results"] = []
for query in verify_queries:
try:
cursor.execute(query)
rows = cursor.fetchall()
result["verification_results"].append({
"query": query,
"result": [dict(row) for row in rows],
"success": True
})
except Exception as query_error:
result["verification_results"].append({
"query": query,
"error": str(query_error),
"success": False
})
conn.close()
except Exception as e:
result["db_error"] = str(e)
return result
def _reset_database(self, database_id: str, sql_content: Optional[str] = None) -> Dict[str, Any]:
"""
Reset database to clean state with seed data.
This method is generic and works with any MCP that follows
the standard session manager interface.
Args:
database_id: Database identifier for multi-tenancy
sql_content: Optional custom SQL content for seeding. If provided,
this will be used instead of the default seed data.
Returns:
Dictionary with reset status and details
"""
try:
# Dispose any cached engine connections to prevent stale connections
# (Only if the SessionManager has this method - some MCPs may not have it)
if hasattr(self.session_manager, 'dispose_engine'):
self.session_manager.dispose_engine(database_id)
# Get database path using session manager
db_path = self.session_manager.get_db_path(database_id)
# Drop all existing tables using sqlite3
conn = sqlite3.connect(db_path)
cursor = conn.cursor()
cursor.execute("SELECT name FROM sqlite_master WHERE type='table' AND name NOT LIKE 'sqlite_%'")
tables = cursor.fetchall()
for table in tables:
table_name = table[0]
cursor.execute(f"DROP TABLE IF EXISTS {table_name}")
conn.commit()
conn.close()
# Recreate tables using session manager
self.session_manager.init_database(database_id, create_tables=True)
# Use custom SQL content if provided, otherwise use default seed data
if sql_content:
logger.info(f"Using custom SQL content for database {database_id}")
seed_sql = sql_content
used_custom_sql = True
else:
seed_data_fn = get_seed_data_function()
seed_sql = seed_data_fn()
used_custom_sql = False
conn = sqlite3.connect(db_path)
cursor = conn.cursor()
# Parse and execute SQL statements
statements = []
for line in seed_sql.split("\n"):
line = line.strip()
if line and not line.startswith("--"):
statements.append(line)
full_sql = " ".join(statements)
individual_statements = [stmt.strip() for stmt in full_sql.split(";") if stmt.strip()]
executed_count = 0
for statement in individual_statements:
try:
if not statement.strip():
continue
cursor.execute(statement)
executed_count += 1
except Exception as e:
logger.error(f"Error executing statement during seeding: {statement[:100]}...")
logger.error(f"Error details: {e}")
raise e
conn.commit()
conn.close()
seed_source = "custom SQL" if used_custom_sql else "default seed data"
logger.info(f"Database {database_id} reset and seeded with {seed_source} ({executed_count} statements)")
return {
"success": True,
"message": f"Database reset to clean state and seeded with {seed_source}",
"database_id": database_id,
"seeded": True,
"used_custom_sql": used_custom_sql,
"statements_executed": executed_count,
}
except Exception as e:
logger.error(f"Error resetting database {database_id}: {e}", exc_info=True)
return {
"success": False,
"message": f"Database reset failed: {str(e)}",
"database_id": database_id,
"seeded": False,
"used_custom_sql": sql_content is not None,
}
|