@agent-inspect/eval 3.5.0 → 3.5.1
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/README.md +63 -0
- package/package.json +6 -5
package/README.md
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
1
|
+
# @agent-inspect/eval
|
|
2
|
+
|
|
3
|
+
Deterministic local eval rules over AgentInspect traces (no LLM judge by default).
|
|
4
|
+
|
|
5
|
+
## When to use
|
|
6
|
+
|
|
7
|
+
- CI checks on trace shape, tool usage, completion status
|
|
8
|
+
- Custom eval rules alongside `agent-inspect check`
|
|
9
|
+
|
|
10
|
+
## When not to use
|
|
11
|
+
|
|
12
|
+
- LLM-as-judge benchmarking platforms
|
|
13
|
+
- Hosted eval datasets
|
|
14
|
+
|
|
15
|
+
## Install
|
|
16
|
+
|
|
17
|
+
```bash
|
|
18
|
+
npm install agent-inspect @agent-inspect/eval
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
## Example
|
|
22
|
+
|
|
23
|
+
```ts
|
|
24
|
+
import { evalRun } from "@agent-inspect/eval";
|
|
25
|
+
|
|
26
|
+
const result = await evalRun(".agent-inspect/run.jsonl", {
|
|
27
|
+
rules: [/* built-in or custom */],
|
|
28
|
+
});
|
|
29
|
+
console.log(result.summary);
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
## Privacy
|
|
33
|
+
|
|
34
|
+
- Reads local files only
|
|
35
|
+
- No network
|
|
36
|
+
|
|
37
|
+
## API
|
|
38
|
+
|
|
39
|
+
| Export | Purpose |
|
|
40
|
+
| ------ | ------- |
|
|
41
|
+
| `evalRun` | Run rules on a trace |
|
|
42
|
+
| `checks` | Built-in check helpers |
|
|
43
|
+
|
|
44
|
+
## CLI
|
|
45
|
+
|
|
46
|
+
`npx agent-inspect eval` · `check`
|
|
47
|
+
|
|
48
|
+
## Docs
|
|
49
|
+
|
|
50
|
+
- [CI starter](https://github.com/rajudandigam/agent-inspect/tree/main/examples/starters/ci-eval-redact)
|
|
51
|
+
|
|
52
|
+
## Troubleshooting
|
|
53
|
+
|
|
54
|
+
- **AI_EVAL_TRACE_UNREADABLE:** Path or format wrong — use `.jsonl` persisted traces
|
|
55
|
+
- **Rule failures:** See `result.findings` for evidence paths
|
|
56
|
+
|
|
57
|
+
## Version
|
|
58
|
+
|
|
59
|
+
`agent-inspect@3.5.x`
|
|
60
|
+
|
|
61
|
+
## License
|
|
62
|
+
|
|
63
|
+
MIT
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@agent-inspect/eval",
|
|
3
|
-
"version": "3.5.
|
|
3
|
+
"version": "3.5.1",
|
|
4
4
|
"license": "MIT",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"description": "Deterministic local eval utilities for AgentInspect traces",
|
|
@@ -12,7 +12,7 @@
|
|
|
12
12
|
"bugs": {
|
|
13
13
|
"url": "https://github.com/rajudandigam/agent-inspect/issues"
|
|
14
14
|
},
|
|
15
|
-
"homepage": "https://github.com/rajudandigam/agent-inspect
|
|
15
|
+
"homepage": "https://github.com/rajudandigam/agent-inspect/tree/main/packages/eval",
|
|
16
16
|
"sideEffects": false,
|
|
17
17
|
"main": "./dist/index.cjs",
|
|
18
18
|
"module": "./dist/index.mjs",
|
|
@@ -30,12 +30,13 @@
|
|
|
30
30
|
}
|
|
31
31
|
},
|
|
32
32
|
"files": [
|
|
33
|
+
"README.md",
|
|
33
34
|
"dist"
|
|
34
35
|
],
|
|
35
36
|
"dependencies": {
|
|
36
|
-
"@agent-inspect/circuit": "3.5.
|
|
37
|
-
"@agent-inspect/guardrails": "3.5.
|
|
38
|
-
"agent-inspect": "3.5.
|
|
37
|
+
"@agent-inspect/circuit": "3.5.1",
|
|
38
|
+
"@agent-inspect/guardrails": "3.5.1",
|
|
39
|
+
"agent-inspect": "3.5.1"
|
|
39
40
|
},
|
|
40
41
|
"publishConfig": {
|
|
41
42
|
"access": "public"
|