@bivelio/savings-layer 0.4.16 → 0.4.19
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/dist/gateway/bin.js +273 -217
- package/dist/index.d.ts +32 -29
- package/dist/index.js +429 -428
- package/docs/terminales/claude-code.md +11 -1
- package/package.json +2 -1
|
@@ -35,11 +35,21 @@ terminales a la vez):
|
|
|
35
35
|
|
|
36
36
|
```bash
|
|
37
37
|
bvsala gateway # ventana 1: http://127.0.0.1:8402
|
|
38
|
-
ANTHROPIC_BASE_URL=http://127.0.0.1:8402 claude
|
|
38
|
+
ENABLE_TOOL_SEARCH=true ANTHROPIC_BASE_URL=http://127.0.0.1:8402 claude # ventana 2 — SIN /v1
|
|
39
39
|
```
|
|
40
40
|
|
|
41
41
|
Claude Code añade `/v1/messages` por su cuenta. Sin barra final.
|
|
42
42
|
|
|
43
|
+
**`ENABLE_TOOL_SEARCH=true` no es opcional.** Al ver un `ANTHROPIC_BASE_URL`
|
|
44
|
+
que no es el oficial, Claude Code apaga por defecto la búsqueda de herramientas
|
|
45
|
+
y mete TODAS las definiciones (las de tus servidores MCP incluidas) en el
|
|
46
|
+
prefijo cacheado de cada petición. Medido el 2026-09-09 en un montaje con 226
|
|
47
|
+
herramientas MCP: 450 KB por petición, el 66 % de los bytes y el 25 % de la
|
|
48
|
+
factura del día. El gateway reenvía la beta y los bloques `tool_reference`
|
|
49
|
+
intactos, así que la búsqueda funciona a través de él. `bvsala claude` la
|
|
50
|
+
activa solo; si ya tienes la variable fijada (`false`, `auto`, `auto:N`), se
|
|
51
|
+
respeta tu valor.
|
|
52
|
+
|
|
43
53
|
## Comprobar que funciona
|
|
44
54
|
|
|
45
55
|
**Con `bvsala claude` (el envoltorio), la comprobación es tu dashboard**: el
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@bivelio/savings-layer",
|
|
3
|
-
"version": "0.4.
|
|
3
|
+
"version": "0.4.19",
|
|
4
4
|
"description": "BV-SALA — provider-neutral LLM execution optimizer. A decision layer in front of any LLM provider that cuts the cost of every request while preserving required quality, and reports the saving per call with token-level accounting you can check against your bill. Proprietary and confidential — property of BiVelio Inc.",
|
|
5
5
|
"keywords": [
|
|
6
6
|
"llm",
|
|
@@ -83,6 +83,7 @@
|
|
|
83
83
|
"anillo1": "tsx --env-file-if-exists=.env.bench benchmarks/anillo1/arnes.ts",
|
|
84
84
|
"anillo2": "tsx benchmarks/anillo2/arnes.ts",
|
|
85
85
|
"anillo2:conciliar": "tsx benchmarks/anillo2/conciliar.ts",
|
|
86
|
+
"anillo2:fugas": "zsh scripts/anillo2-vigia-fugas.zsh",
|
|
86
87
|
"salida": "tsx --env-file-if-exists=.env.bench benchmarks/salida/arnes.ts",
|
|
87
88
|
"salida:anatomia": "tsx benchmarks/salida/anatomia.ts"
|
|
88
89
|
},
|