Add verbose Python diagnostics and fix braintrust version for canary bt

- Add bt --version, PYTHONUNBUFFERED=1, and python3 version check to
  help diagnose Python eval failures in CI
- Update test-eval-py to require braintrust>=0.8.0 (was 0.0.160) so
  bt's canary eval-runner.py can import braintrust.parameters
- Regenerate uv.lock with braintrust 0.8.0

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: ibolmo

eval/dist/index.js

@@ -134,6 +134,18 @@ export async function runEval(args: Params, onSummary: OnSummaryFn) {
 
   process.chdir(path.resolve(root));
 
+  await runCommand("bt --version", () => {});
+
+  // Ensure Python output is not buffered so bt captures it in real time.
+  process.env.PYTHONUNBUFFERED = "1";
+
+  if (args.runtime === "python" || args.runner?.includes("python")) {
+    await runCommand(
+      "python3 -c \"import sys; print('python:', sys.executable, sys.version)\"",
+      () => {},
+    );
+  }
+
   // Build bt eval flags
   const flags: string[] = ["--jsonl", "--verbose"];
 

eval/dist/index.js.map

@@ -6,5 +6,5 @@ readme = "README.md"
 requires-python = ">=3.12"
 dependencies = [
     "autoevals>=0.0.92",
-    "braintrust>=0.0.160",
+    "braintrust>=0.8.0",
 ]

eval/src/braintrust.ts

@@ -1,24 +1,2 @@
-attrs==24.2.0
-autoevals==0.0.92
-braintrust==0.0.160
-braintrust-core==0.0.54
-certifi==2024.8.30
-charset-normalizer==3.3.2
-chevron==0.14.0
-exceptiongroup==1.2.0
-gitdb==4.0.11
-gitpython==3.1.43
-idna==3.10
-jsonschema==4.23.0
-jsonschema-specifications==2023.12.1
-levenshtein==0.26.0
-python-dotenv==1.0.1
-pyyaml==6.0.2
-rapidfuzz==3.10.0
-referencing==0.35.1
-requests==2.32.4
-rpds-py==0.20.0
-smmap==5.0.1
-sseclient-py==1.8.0
-tqdm==4.66.5
-urllib3==2.5.0
+braintrust>=0.8.0
+autoevals