Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions docs/decision-retrieval.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,10 @@ additive evaluation path over native lexical/vector retrieval. When the
provider is disabled, unavailable, or returns an invalid response, Noema keeps
native retrieval results.

The default local endpoint is `http://127.0.0.1:18788/v1/decision`. Providers
that still expose the older TypeSafe-compatible path `/v1/systemone` can be
selected with an explicit `endpoint:`.

## Provider configuration

Configure under `decision` in the Noema config file:
Expand Down
4 changes: 4 additions & 0 deletions docs/testing/native-decision.md
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,10 @@ evaluate retrieval candidates. Inference uses a configured endpoint; model
weights are not bundled. When evaluation is disabled or fails closed, native
retrieval results are preserved.

The default local endpoint is `http://127.0.0.1:18788/v1/decision`. Providers
that still expose the older TypeSafe-compatible path `/v1/systemone` can be
selected with an explicit `endpoint:`.

## Configuration

```yaml
Expand Down
8 changes: 4 additions & 4 deletions src/decision.rs
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ use std::{
};
use tokio::sync::Semaphore;

const DEFAULT_ENDPOINT: &str = "http://127.0.0.1:18788/v1/systemone";
const DEFAULT_ENDPOINT: &str = "http://127.0.0.1:18788/v1/decision";
const MAX_REQUEST_BYTES: usize = 30_000;
pub(crate) const RESCUE_MARKER: &str = " [retrieval rescue; verify date and scope]";
static CONCURRENCY: std::sync::OnceLock<Arc<Semaphore>> = std::sync::OnceLock::new();
Expand Down Expand Up @@ -979,7 +979,7 @@ mod tests {
let client = Client {
token: "local".into(),
model: "clef-flash".into(),
endpoint: "http://127.0.0.1:18788/v1/systemone".into(),
endpoint: "http://127.0.0.1:18788/v1/decision".into(),
timeout: Duration::from_secs(30),
};
let (state, questions) = batch_fixture(16);
Expand Down Expand Up @@ -1298,12 +1298,12 @@ mod tests {
enabled: Enablement::Explicit(true),
provider: Provider::Local,
model: "any-local-model".into(),
endpoint: "http://127.0.0.1:9/v1/systemone".into(),
endpoint: "http://127.0.0.1:9/v1/decision".into(),
..DecisionConfig::default()
};
let client = config.client().unwrap().unwrap();
assert_eq!(client.token, "local-clef");
assert_eq!(client.endpoint, "http://127.0.0.1:9/v1/systemone");
assert_eq!(client.endpoint, "http://127.0.0.1:9/v1/decision");
assert_eq!(config.credential().0, "local-clef");
}

Expand Down
2 changes: 1 addition & 1 deletion tests/decision_enablement.rs
Original file line number Diff line number Diff line change
Expand Up @@ -186,7 +186,7 @@ fn local_clef_auto_does_not_use_cloud_key() {
#[test]
fn explicit_local_provider_activates_without_cloud_key() {
let (_dir, mut command) = command(
"decision:\n provider: local\n enabled: true\n model: local-eval\n endpoint: http://127.0.0.1:18788/v1/systemone\n",
"decision:\n provider: local\n enabled: true\n model: local-eval\n endpoint: http://127.0.0.1:18788/v1/decision\n",
);
command
.env("TYPESAFE_API_KEY", "synthetic-cloud-key-must-not-be-used")
Expand Down
Loading