oh-my-promptfoo 1.18.0 → 1.19.0
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/README.md +14 -3
- package/dist/assertions.cjs +2986 -1514
- package/dist/assertions.cjs.map +1 -1
- package/dist/assertions.d.cts +19 -7
- package/dist/assertions.d.ts +19 -7
- package/dist/assertions.js +2979 -1517
- package/dist/assertions.js.map +1 -1
- package/dist/providers.cjs +1694 -0
- package/dist/providers.cjs.map +1 -0
- package/dist/providers.d.cts +51 -0
- package/dist/providers.d.ts +51 -0
- package/dist/providers.js +1660 -0
- package/dist/providers.js.map +1 -0
- package/package.json +6 -1
package/README.md
CHANGED
|
@@ -43,6 +43,9 @@ providers:
|
|
|
43
43
|
path: ./skills # relative to this file
|
|
44
44
|
destination: .agents/skills
|
|
45
45
|
permissions: all
|
|
46
|
+
defaultTest:
|
|
47
|
+
options:
|
|
48
|
+
provider: openai:chat:gpt-6-sol # the llmAssert judge
|
|
46
49
|
tests:
|
|
47
50
|
- vars:
|
|
48
51
|
task: Our checkout page crashes when the price is empty. Explain the likely cause.
|
|
@@ -61,12 +64,19 @@ Run it and open the results:
|
|
|
61
64
|
|
|
62
65
|
```sh
|
|
63
66
|
export GH_TOKEN=... # for the agent
|
|
64
|
-
export OPENAI_API_KEY=... # for the
|
|
65
|
-
export OPENAI_MODEL=... # the judge model
|
|
67
|
+
export OPENAI_API_KEY=... # for the judge's OpenAI provider
|
|
66
68
|
npx promptfoo eval
|
|
67
69
|
npx promptfoo view
|
|
68
70
|
```
|
|
69
71
|
|
|
72
|
+
The judge is an ordinary Promptfoo provider, set the way `llm-rubric`'s is:
|
|
73
|
+
in `defaultTest.options.provider` as above, or with
|
|
74
|
+
`npx promptfoo eval --grader <provider-id>`. For Azure OpenAI, set
|
|
75
|
+
`OPENAI_BASE_URL` to the full `https://<resource>.openai.azure.com/openai/v1`
|
|
76
|
+
address and use `openai:chat:<deployment>`. The
|
|
77
|
+
[grading guide](https://github.com/allagentsdev/oh-my-promptfoo/blob/main/docs/grading-guide.md#choosing-llmasserts-judge)
|
|
78
|
+
covers per-test judges and falling back to another judge during an outage.
|
|
79
|
+
|
|
70
80
|
The [examples](https://github.com/allagentsdev/oh-my-promptfoo/tree/main/examples) include configs for Codex, Claude Code, Copilot,
|
|
71
81
|
pinned Git and OCI sources, and larger runs.
|
|
72
82
|
|
|
@@ -102,7 +112,8 @@ examples.
|
|
|
102
112
|
- [Provider guide](https://github.com/allagentsdev/oh-my-promptfoo/blob/main/docs/provider-guide.md): sources, permissions, each agent's
|
|
103
113
|
settings, limits, cleanup and caching
|
|
104
114
|
- [Grading guide](https://github.com/allagentsdev/oh-my-promptfoo/blob/main/docs/grading-guide.md): `llmAssert` and `agentAssert` in
|
|
105
|
-
full, including
|
|
115
|
+
full, including choosing the judge, fallback judges, scoring modes,
|
|
116
|
+
templates, token usage and prompt size limits
|
|
106
117
|
- [Examples](https://github.com/allagentsdev/oh-my-promptfoo/tree/main/examples): complete Promptfoo configs
|
|
107
118
|
- [Architecture decisions](https://github.com/allagentsdev/oh-my-promptfoo/tree/main/docs/adr): why things work the way they do
|
|
108
119
|
- [Changelog](https://github.com/allagentsdev/oh-my-promptfoo/blob/main/CHANGELOG.md)
|