Upgrade CloudLogger to aepipe-sdk v0.1.1 with D1 payload support

[loadchange]

Summary

Upgrade the CloudLogger module from a custom HTTP-based implementation to use the official aepipe-sdk (v0.1.1), introducing D1 payload support for storing complete data without truncation and adding comprehensive event types for the full trading lifecycle.

Key Changes

Core SDK Integration

• Replace custom urllib-based HTTP implementation with official aepipe-sdk client
• Use Aepipe client for all cloud communication with built-in blob size validation
• Import DataPoint, LogEntry, AepipeError, and ValidationError from aepipe SDK
• Remove manual URL construction and Bearer token header management (handled by SDK)

Data Model Improvements

• Change queue types from dict to SDK types: LogEntry for logs, DataPoint for events
• Introduce D1 payload support for storing complete data (AI responses, prompts, market snapshots)
• Implement payload TTL configuration (default 90 days) for automatic cleanup
• Separate concerns: blobs for lightweight indexing, doubles for metrics, payload for full data

Decision Logging Enhancement

• Store complete AI responses and prompts in payload (no truncation)
• Add market data snapshot storage via payload
• Include additional decision context: debate summary, regime, validation result
• Truncate error messages in blobs to 200 chars while preserving full version in payload
• Blobs now contain: [symbol, decision, status, error_msg_truncated, regime]

New Event Types

• send_system_event(): System lifecycle events (startup, shutdown, config changes)
• send_risk_event(): Risk management events (drawdown, daily loss, stop loss triggers)
• send_grid_event(): Grid trading events (sync, rebuild, cancel)
• send_cycle_event(): Trading cycle events (start, end, skip)
• send_account_snapshot(): Periodic account state snapshots with positions
• send_alert(): Market alerts with severity mapping (extreme/high → error, elevated → warn)

Trade Event Expansion

• Add leverage, take_profit_price, stop_loss_price, order_type, and fee fields
• Store complete trade data in payload alongside indexed blobs

Error Handling

• Distinguish between ValidationError (client-side validation, skip without retry) and AepipeError (server errors, retry)
• Gracefully handle network errors without impacting main trading flow

Integration Points

• Update main.py to send system startup event with configuration details
• Add account snapshot logging in trading cycle
• Add market alert logging in volatility detection
• Add risk event logging for stop loss failures and protection triggers
• Add grid event logging for grid rebuilds
• Update grid_main.py and related modules to use new cloud logger API

Configuration

• Add payload_ttl parameter to CloudLogger initialization
• Update config.yaml example with v2 documentation
• Add aepipe-sdk dependency to pyproject.toml

Implementation Details

• Payload TTL defaults to 90 days (7,776,000 seconds), matching Analytics Engine retention
• Blobs limited to 15 items (aepipe-sdk 0.1.1 constraint), doubles to 20 items
• Queue overflow handling: discard oldest message when full (maintains FIFO for important data)
• Batch size capped at 200 items per request (aepipe limit is 250)
• Retry logic: 2 attempts for transient errors, skip on validation errors
• All cloud operations are non-blocking; failures don't affect trading logic

https://claude.ai/code/session_01McaLk3VSxwmNnMVybEsbW9

[gemini-code-assist]

Code Review

This pull request upgrades the cloud logging infrastructure to version 2 by integrating the aepipe-sdk and implementing D1 payload support for storing large datasets like AI responses and prompts. It significantly expands logging coverage across the application, adding specific event types for system lifecycle, risk management, grid operations, and account snapshots. Review feedback recommends enhancing error reporting in main.py by replacing repr(e) with traceback.format_exc() to provide more detailed stack traces for debugging decision and cycle errors.

[gemini-code-assist]

For better debuggability, it's recommended to use traceback.format_exc() instead of repr(e) to capture the full stack trace of the exception. This provides more context when an error occurs. Other parts of the codebase, like in grid_main.py, already use this approach, and maintaining consistency is good practice.

You'll also need to add import traceback at the top of the file.

Suggested change

	details={"traceback": repr(e), "cycle": self.cycle_counter},
	details={"traceback": traceback.format_exc(), "cycle": self.cycle_counter},

[gemini-code-assist]

Similar to my other comment, using traceback.format_exc() here instead of repr(e) would provide a more detailed traceback for easier debugging of cycle errors. This will also require importing the traceback module if it's not already present.

Suggested change

	details={"traceback": repr(e), "cycle": self.cycle_counter},
	details={"traceback": traceback.format_exc(), "cycle": self.cycle_counter},