> ## Documentation Index
> Fetch the complete documentation index at: https://docs.raptordata.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Evidence Logging

> Full audit trail for every AI request

# Know What Your AI Did

Every request through Raptor is logged with full metadata. Debugging, compliance, analytics—all covered.

## What's Logged

| Field                 | Example                |
| --------------------- | ---------------------- |
| `request_id`          | `abc123-def456-...`    |
| `timestamp`           | `2024-01-15T10:30:00Z` |
| `method`              | `POST`                 |
| `path`                | `/v1/chat/completions` |
| `status`              | `200`                  |
| `cache_hit`           | `true` / `false`       |
| `latency_ms`          | `5`                    |
| `upstream_latency_ms` | `450`                  |
| `firewall_result`     | `null` / `blocked`     |

## Dashboard

View all requests in your dashboard:

1. Go to **Traces** in the sidebar
2. Filter by date, status, cache hit, or agent
3. Click a row for full request details

## Response Headers

Every response includes a request ID:

```
X-Raptor-Request-Id: abc123-def456-...
```

Use this to look up the full evidence record in your dashboard.

## Agent Tracking

Tag requests by agent for multi-agent systems:

```python theme={null}
client = OpenAI(
    base_url="https://proxy.raptordata.dev/v1",
    default_headers={
        "X-Raptor-Api-Key": "rpt_...",
        "X-Raptor-Workspace-Id": "...",
        "X-Raptor-Agent-Id": "support-bot"  # Add this
    }
)
```

Then filter by agent in your dashboard.

## Compliance

Raptor evidence logs support common compliance requirements:

| Requirement         | How Raptor Helps                                 |
| ------------------- | ------------------------------------------------ |
| **Audit trail**     | Every request logged with timestamp and metadata |
| **Data retention**  | Configurable retention periods                   |
| **Access control**  | Workspace isolation, API key scoping             |
| **Request tracing** | Unique request IDs for debugging                 |

## Retention

| Plan       | Retention |
| ---------- | --------- |
| Free       | 7 days    |
| Pro        | 30 days   |
| Enterprise | Custom    |

## Performance

Evidence logging is **completely async**. Logs are sent to a background worker via an in-memory channel. Your requests never wait for logging.

```
Request → Response (immediate)
    └──→ Log channel → Background worker → Database
```

<Note>
  Zero impact on request latency. We buffer up to 10,000 log entries and batch-write to the database.
</Note>