session-brain

1# Session Brain
2 
3Builds a searchable topic graph over your agent session history. The output is a **sidecar**
4at `~/.claude/session-brain/` — the vault is never touched.
5 
6All the heavy lifting is deterministic Python in the `obsidian-wiki` CLI. Your only job is to
7name the clusters, which takes exactly one turn and requires reading no transcripts.
8 
9## When to use which skill
10 
11| Goal | Skill |
12|---|---|
13| Build or refresh the graph; survey topics | `session-brain` (this one) |
14| Find and load a specific past session | `session-search` |
15| Distil sessions into permanent vault pages | `wiki-history-ingest` / `claude-history-ingest` |
16 
17## Step 1: Build
18 
19```bash
20obsidian-wiki sessions-build --json
21```
22 
23Roughly 3 seconds cold on ~1000 sessions, well under a second incrementally — it re-reads only
24transcripts whose size or mtime changed. Useful flags:
25 
26| Flag | When |
27|---|---|
28| `--full` | Ignore all caches and re-read everything |
29| `--mutual` | Tighter, smaller clusters (mutual-kNN edges only) |
30| `--half-life N` | Change the recency half-life (default 90 days) |
31| `--min-sim 0.15` | Fewer, stronger edges — use if the graph is too dense to read |
32| `--skip name` | Exclude a project. Match is substring-based; pass the bare name, because cache dirs start with `-` and argparse reads that as a flag |
33 
34Report the headline numbers: total sessions, how many have transcripts vs. are history-only,
35edges, and cluster count.
36 
37## Step 2: Name the unnamed clusters
38 
39```bash
40…