feat: Audit Logging Integration

[limaronaldo]

Summary

Integrates the existing audit crate into the production code path, enabling comprehensive audit logging for LLM operations, tool execution, and HTTP requests.

Changes

Providers Crate (audit feature)

• AuditedBackend wrapper for any LLMBackend
  • Logs LlmRequest events before inference
  • Logs LlmResponse events with token usage after inference
  • Logs Error events on failures
  • Supports context propagation (trace_id, session_id, agent_id)

Agents Crate (audit feature)

• AuditedExecutor wrapper for ToolExecutor
  • Logs ToolCall events with parameters and results
  • Logs ApprovalDecision events for denied/skipped tools
  • Logs Error events for approval errors
  • Duration tracking for tool executions

Dashboard Crate (audit feature)

• REST API Endpoints
  • GET /api/audit/events - List events with filtering (event_type, trace_id, session_id, agent_id, pagination)
  • GET /api/audit/stats - Aggregated statistics by event type
• HTTP Audit Middleware - Logs all HTTP requests with method, path, status, duration
• Uses MemoryLogger for queryable in-memory storage

Audit Types Enhancement

• Added AuditEvent::llm_response() convenience constructor

Testing

• Providers: 2 new tests (request/response logging, error logging)
• Agents: 3 new tests (tool calls, denials, multiple tools)
• Dashboard: 4 new tests (state, logging, stats, conversion)
• All 9 new tests passing

Usage

// Wrap any LLM backend with audit logging
use rust_ai_agents_providers::AuditedBackend;
use rust_ai_agents_audit::MemoryLogger;

let backend = AnthropicProvider::claude_35_sonnet(api_key);
let logger = Arc::new(MemoryLogger::new());
let audited = AuditedBackend::new(backend, logger)
    .with_trace_id("req-123")
    .with_agent_id("research-agent");

// All LLM calls are now logged
let output = audited.infer(&messages, &tools, 0.7).await?;

Feature Flags

All integrations are opt-in via feature flags:

• rust-ai-agents-providers/audit
• rust-ai-agents-agents/audit
• rust-ai-agents-dashboard/audit

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.