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

# Codex

> Codex CLI sessions → DuckBrain ingest (Available)

**Status:** Available

Ingests your Codex conversation history from `~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl` — every user and assistant turn becomes a searchable record. Reasoning, tool-call, and `role: developer` lines (injected system-prompt noise) are excluded; text blocks are decoded to prose.

```bash theme={null}
duckbrain sync codex
```

`duckbrain watch` streams new rollout lines live as Codex writes them, with batch backfill over the same files as a safety net.

## Facet mapping

| Facet      | Value                                    |
| ---------- | ---------------------------------------- |
| `--source` | `codex`                                  |
| `--in`     | session uuid (from the rollout filename) |
| observer   | `agent:codex`                            |

```bash theme={null}
duckbrain search "data table expansion" --source codex --since 30d
duckbrain search "data table expansion" --in 019b0f38-10a2-73d3-bf49-e05f649b0f37
```

Rows carry `classification: confidential`.