@verevoir/recipes 0.4.0 → 0.5.0
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 +8 -0
- package/README.md +7 -0
- package/package.json +7 -5
package/CHANGELOG.md
CHANGED
|
@@ -1,5 +1,13 @@
|
|
|
1
1
|
# Changelog
|
|
2
2
|
|
|
3
|
+
## 0.5.0 — 2026-06-12
|
|
4
|
+
|
|
5
|
+
**`@verevoir/llm` is now an optional `peerDependency`** (was a direct dependency) (STDIO-343). As a direct dependency it let npm install a **nested** `@verevoir/llm` under recipes when a consumer pinned a different range — two module instances, two model catalogues, and the host's module mocks missing recipes' copy (the cause of a red deploy). As a peer, the consumer's single `@verevoir/llm` is used.
|
|
6
|
+
|
|
7
|
+
- Only the `/engine` subpath (provisioning / retrieval / plan) needs `@verevoir/llm`; the main entry (the skill / capability parsers) doesn't — so the peer is **optional**. `@anthropic-ai/sdk` stays an optional peer.
|
|
8
|
+
- Added to recipes' own `devDependencies` so its tests + build still resolve it.
|
|
9
|
+
- Consumers (`aigency-web`, `@verevoir/mcp`) already depend on `@verevoir/llm` directly, so there's nothing for them to change.
|
|
10
|
+
|
|
3
11
|
## 0.4.0 — 2026-06-12
|
|
4
12
|
|
|
5
13
|
**Reasoning call de-pinned from Anthropic** (STDIO-340). Practice provisioning's concern-tagging no longer hardcodes the Anthropic adapter — the model call is injectable, so it can run on DeepSeek / Mistral / any `@verevoir/llm` provider without recipes importing every SDK.
|
package/README.md
CHANGED
|
@@ -78,3 +78,10 @@ fence, missing required field, or an `id` that doesn't match the filename).
|
|
|
78
78
|
The function/type names keep the `Skill*` prefix (the descriptors originated as
|
|
79
79
|
Agent-Oriented-Architecture "skills"); the package is named **recipes** to avoid
|
|
80
80
|
collision with the host's _capability_ concept and with model-provider "Skills".
|
|
81
|
+
|
|
82
|
+
## Docs
|
|
83
|
+
|
|
84
|
+
- [Model matrix smoke test](docs/model-matrix-smoke.md) — a periodic, low-cost
|
|
85
|
+
check that the capability/practice mechanism is hooked and behaves consistently
|
|
86
|
+
across providers (Anthropic / Google / Mistral / SambaNova). A smoke test of
|
|
87
|
+
the mechanism, not a model benchmark.
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@verevoir/recipes",
|
|
3
|
-
"version": "0.
|
|
3
|
+
"version": "0.5.0",
|
|
4
4
|
"description": "Recipe-descriptor format + zero-dependency parser: a flat-frontmatter .md describes a parameterised procedure (typed inputs → instructions → result) that a host compiles to a chat-time tool or an MCP prompt.",
|
|
5
5
|
"license": "Apache-2.0",
|
|
6
6
|
"type": "module",
|
|
@@ -36,7 +36,8 @@
|
|
|
36
36
|
"@types/node": "^25.8.0",
|
|
37
37
|
"prettier": "^3",
|
|
38
38
|
"typescript": "^5",
|
|
39
|
-
"vitest": "^4"
|
|
39
|
+
"vitest": "^4",
|
|
40
|
+
"@verevoir/llm": "^0.13.0"
|
|
40
41
|
},
|
|
41
42
|
"publishConfig": {
|
|
42
43
|
"access": "public"
|
|
@@ -48,13 +49,14 @@
|
|
|
48
49
|
"engines": {
|
|
49
50
|
"node": ">=20"
|
|
50
51
|
},
|
|
51
|
-
"dependencies": {
|
|
52
|
-
"@verevoir/llm": "^0.9.0"
|
|
53
|
-
},
|
|
54
52
|
"peerDependencies": {
|
|
53
|
+
"@verevoir/llm": ">=0.9.0",
|
|
55
54
|
"@anthropic-ai/sdk": "*"
|
|
56
55
|
},
|
|
57
56
|
"peerDependenciesMeta": {
|
|
57
|
+
"@verevoir/llm": {
|
|
58
|
+
"optional": true
|
|
59
|
+
},
|
|
58
60
|
"@anthropic-ai/sdk": {
|
|
59
61
|
"optional": true
|
|
60
62
|
}
|