@geohar/pi-cribsheet 0.23.0-dev1 → 0.23.0-dev3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +69 -0
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -9,6 +9,75 @@ This file ships in the Pi package, where Pi shows the entries newer than the
|
|
|
9
9
|
version you last ran. `scripts/bump-version.sh` refuses a release whose
|
|
10
10
|
version has no heading here.
|
|
11
11
|
|
|
12
|
+
## [0.23.0-dev3] - 2026-11-13
|
|
13
|
+
|
|
14
|
+
### Added
|
|
15
|
+
|
|
16
|
+
- `crib generate setup`: finds the LLM providers this machine has (API keys
|
|
17
|
+
in the environment, a local Ollama or llama-swap, the `claude` CLI,
|
|
18
|
+
providers already defined for zsh-ai), asks which to use and which model,
|
|
19
|
+
and writes `~/.config/crib/models.toml` and the `[generate]` section of
|
|
20
|
+
`config.toml` after showing the diff. Both files are backed up first; a
|
|
21
|
+
symlinked config is edited where it lives. Ollama Cloud gets
|
|
22
|
+
`structured = "tool"`, and providers get room for crib's bulk replies.
|
|
23
|
+
- `crib generate test`: one small call per purpose, reporting whether the key
|
|
24
|
+
works and structured output comes back.
|
|
25
|
+
- The README quick start has a generation step, and
|
|
26
|
+
`docs/generate.toml.example` shows current practice.
|
|
27
|
+
|
|
28
|
+
## [0.23.0-dev2] - 2026-11-13
|
|
29
|
+
|
|
30
|
+
### Added
|
|
31
|
+
|
|
32
|
+
- `crib project doctor` (`project_doctor`): what in a project needs
|
|
33
|
+
attention, report-only, each finding naming the command that fixes it.
|
|
34
|
+
It checks dangling dependencies; dependencies on a superseded decision
|
|
35
|
+
(naming its replacement) or a declined plan; cited sections that changed
|
|
36
|
+
or vanished; moved or orphaned learnings; one id in several files;
|
|
37
|
+
near-duplicate notes, decisions and plans; and pending descriptions.
|
|
38
|
+
- `crib project related <ref>` (`project_related`): one item's links both
|
|
39
|
+
ways and its nearest notes, decisions, plans, learnings and symbols.
|
|
40
|
+
- `crib project lookup` (`project_lookup`): one search across notes,
|
|
41
|
+
decisions, plans and learnings, with `--kind` and `--status` filters.
|
|
42
|
+
- `[generate] mode = "on" | "explicit" | "off"`. Without a models
|
|
43
|
+
config it is `explicit`: crib generates descriptions, keywords and
|
|
44
|
+
summaries only when you run a verb that asks for them, and never drives
|
|
45
|
+
the Claude CLI in the background on its own. Background indexing still
|
|
46
|
+
saves code structure; `status` counts the descriptions waiting.
|
|
47
|
+
- `status` reports the daemon: uptime, startup milestones, the background
|
|
48
|
+
work running now, queue depths, and whether file events are arriving.
|
|
49
|
+
The daemon writes a log, by default `~/.cache/crib/daemon.log`.
|
|
50
|
+
- When file events are delayed (macOS fseventsd under load), the daemon
|
|
51
|
+
says so and sweeps for changes every two minutes until they recover.
|
|
52
|
+
|
|
53
|
+
### Changed
|
|
54
|
+
|
|
55
|
+
- After a note, decision or plan write, `similar` compares the whole item
|
|
56
|
+
with every note, decision and plan in the project, not just its first
|
|
57
|
+
line against one pillar. Each match says whether it is a duplicate, a
|
|
58
|
+
confusingly similar title, already settled (done, declined, parked,
|
|
59
|
+
superseded) or just related, and what to do. Plan adds return it too.
|
|
60
|
+
- Decision and plan reads list `suggested_deps`: active decisions the item
|
|
61
|
+
is close to but doesn't depend on, with the command to add the link.
|
|
62
|
+
- `design_import` and `plan_import` name, for each section, the existing
|
|
63
|
+
decisions or plans closest to it.
|
|
64
|
+
- The learnings report names where an orphaned learning probably went.
|
|
65
|
+
- `dep_add` with a different kind changes an existing dependency's kind
|
|
66
|
+
and keeps it verified.
|
|
67
|
+
- Summary aliases are embedded once and kept, instead of re-embedded after
|
|
68
|
+
every daemon start; the keyword index is read from the store.
|
|
69
|
+
|
|
70
|
+
### Fixed
|
|
71
|
+
|
|
72
|
+
- `status` no longer shows a git remote's password or token.
|
|
73
|
+
|
|
74
|
+
### Upgrading
|
|
75
|
+
|
|
76
|
+
- The store's schema changes again, so the first daemon on this version
|
|
77
|
+
rebuilds its cache from your files once.
|
|
78
|
+
- If you relied on background generation without a models config, set
|
|
79
|
+
`[generate] mode = "on"`.
|
|
80
|
+
|
|
12
81
|
## [0.23.0-dev1] - 2026-11-12
|
|
13
82
|
|
|
14
83
|
The DuckDB cache now has one uniform shape: an item per note, decision, plan,
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@geohar/pi-cribsheet",
|
|
3
|
-
"version": "0.23.0-
|
|
3
|
+
"version": "0.23.0-dev3",
|
|
4
4
|
"description": "Pi extension: run the crib (cribsheet) memory + code-index MCP server via sharedserver, inject the reach-for-crib directive, and ship the /crib recall command. Pairs with pi-mcp-adapter.",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "dist/index.js",
|