base-layer/promptkit
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

Document the bundled OpenRouter configuration

Browse Source
This commit is contained in:
Eric Rakestraw
2026-11-03 15:39:04 +00:00
parent 43f55edb98
commit 2b3d165ce2
12 changed files with 114 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
docs/formats.md
View File

@@ -14,8 +14,8 @@ A configuration document may contain top-level `backends`, `profiles`, and
`prompts` lists, plus a `collections` mapping of namespaced resource groups.
Each resource has an explicit literal `id`, not a filename-derived identity.
IDs and references must be nonblank and remain exact: Promptkit does not trim or
re-key them. Unknown fields are errors. Promptkit installs no provider or
profile catalog; callers select the definitions they need.
re-key them. Unknown fields are errors. Promptkit installs a default-on bundled
OpenRouter underlay; callers still explicitly select a profile.
```yaml
backends:
@@ -332,9 +332,58 @@ For example, a runtime `extra_params["extends"]` key is ordinary provider
data, not a configuration inheritance instruction. An explicit zero timeout
does not disable the caller context or HTTP transport timeout.
### Bundled OpenRouter Configuration
The default bundle installs root backend `openrouter`, endpoint
`https://openrouter.ai/api/v1`, optional credential environment name
`OPENROUTER_API_KEY`, and smoke metadata selecting `gpt-luna-latest` with the
`openrouter_key` authentication adapter. Existing diagnostic endpoint fallback
applies. No prompt, implicit profile selection, or tuning defaults are installed.
Disable the entire bundle with `WithBundledOpenRouter(false)`.
The root profile IDs and their models are:
| Profile | Model |
| --- | --- |
| `gpt-astra-latest` | `~openai/gpt-astra-latest` |
| `gpt-sol-latest` | `~openai/gpt-sol-latest` |
| `gpt-terra-latest` | `~openai/gpt-terra-latest` |
| `gpt-luna-latest` | `~openai/gpt-luna-latest` |
| `claude-fable-latest` | `~anthropic/claude-fable-latest` |
| `claude-opus-latest` | `~anthropic/claude-opus-latest` |
| `claude-sonnet-latest` | `~anthropic/claude-sonnet-latest` |
| `claude-haiku-latest` | `~anthropic/claude-haiku-latest` |
| `gemini-pro-latest` | `~google/gemini-pro-latest` |
| `gemini-flash-latest` | `~google/gemini-flash-latest` |
| `auto` | `openrouter/auto` |
| `free` | `openrouter/free` |
The profiles `auto-low`, `auto-medium`, `auto-high`, `auto-xhigh`, and
`auto-max` extend `auto` and add an `auto-router` plugin with the corresponding
`cost_tier`. Replacing `auto` changes their inherited backend and model while
retaining their plugins. These are root IDs even when a model contains `/`.
For example, a caller can replace `auto` as a complete record and extend another
bundled profile under a new ID:
```yaml
profiles:
- id: auto
backend: openrouter
model: openrouter/free
- id: my-luna
extends: {id: gpt-luna-latest}
timeout_seconds: 60
```
See OpenRouter's [alias examples](https://openrouter.ai/~openai/gpt-luna-latest),
[auto router and cost tiers](https://openrouter.ai/docs/guides/routing/routers/auto-router),
and [free router](https://openrouter.ai/docs/guides/routing/routers/free-router)
for provider behavior and availability.
### Source And Profile Precedence
`NewEngine` composes caller underlays in their given order, followed by an
`NewEngine` composes the bundled OpenRouter underlay first by default, then
caller underlays in their given order, followed by an
optional main input. Later layers overlay earlier ones. Inputs within one
underlay group are strictly composed: conflicting contributions are errors,
not priority signals. The directory helpers discover `.yaml` and `.yml` files recursively