fix(providers): estimate cost for Azure Foundry models via inferred catalog pricing (#11264)

Signed-off-by: Michael Neale <michael.neale@gmail.com> Co-authored-by: Michael Neale <michael.neale@gmail.com>
2026-08-21 01:52:41 +00:00
parent 8e8b674cf0
commit 48d480f911
4 changed files with 111 additions and 5 deletions
@@ -190,6 +190,25 @@ mod tests {
        }
    }

+    /// Both agent loops price a chunk through `estimate_model_cost` and tag the result
+    /// `CostSource::Estimated` when it returns `Some` — `reply_parts::resolve_chunk_cost` for
+    /// the legacy path, `state_machine::usage::enrich` for the new one. Pinning the shared
+    /// helper covers the behaviour both paths inherit.
+    #[test]
+    fn azure_foundry_estimates_from_the_azure_catalog_rate() {
+        let used = usage(Some(1_000_000), Some(1_000_000), None);
+
+        let gpt5 = estimate_model_cost("azure_foundry", "gpt-5", &used)
+            .expect("gpt-5 prices through the Azure catalog");
+        assert!(gpt5 > 0.0);
+
+        // Priced from azure/llama-3.3-70b-instruct ($0.71/M in and out), not from the
+        // meta-llama publisher row that lists the open weights at 0.0/0.0.
+        let llama = estimate_model_cost("azure_foundry", "llama-3.3-70b-instruct", &used)
+            .expect("llama-3.3-70b-instruct prices through the Azure catalog");
+        assert!((llama - 1.42).abs() < 1e-9, "got {llama}");
+    }
+
    #[test]
    fn pricing_from_model_info_converts_per_token_to_per_million_usd() {
        let info = ModelInfo::with_cost("m", 262_144, 0.000002, 0.000006);