bitfab-cli 0.2.405 → 0.2.406

package/dist/index.js

@@ -24697,10 +24697,10 @@ var semver3 = __toESM(require_semver2(), 1);
 
 // ../bitfab-plugin-lib/dist/bakedSdkVersions.js
 var BAKED_SDK_VERSIONS = {
-  typescript: "0.62.7",
-  python: "0.62.6",
-  ruby: "0.62.6",
-  go: "0.62.6"
+  typescript: "0.62.8",
+  python: "0.62.7",
+  ruby: "0.62.7",
+  go: "0.62.7"
 };
 
 // ../bitfab-plugin-lib/dist/installedSdk.js
@@ -30184,7 +30184,7 @@ The schema is flat, every file object is exactly \`{ path, before, after }\`. Do
 
 **Check the \`supportsCodeChanges\` flag** (from \`detect-replay-capabilities\`). If false, skip writing the code-change JSON file and omit \`--code-change\` from the invocation. The replay itself is unaffected; only the code-change metadata is missing from the experiment viewer.
 
-**Automatic fallback capture.** If you do not pass \`--code-change\`, a recent SDK auto-captures the branch's diff **against trunk** (the merge-base with \`origin/main\`) inside \`replay()\` itself and attaches it, so even a run where you did not hand-write a payload still shows a diff, as long as the SDK is new enough (\`supportsCodeChanges\`), you are in a git repo, and the working tree carries the change. This fallback is **cumulative** (whole branch vs trunk), not per-experiment. The precise per-experiment before/after still comes from writing the \`--code-change\` payload above, which always wins over the fallback, so keep doing that in \`make-change\`; the trunk diff is a last resort, not a reason to skip it.
+**Automatic fallback capture.** If you do not pass \`--code-change\`, a recent SDK auto-captures the **uncommitted** working-tree diff against \`HEAD\` inside \`replay()\` itself and attaches it, so even a run where you did not hand-write a payload still shows a diff, as long as the SDK is new enough (\`supportsCodeChanges\`), you are in a git repo, and the change is not yet committed. Older SDKs diff against the merge-base with trunk instead, which covers the whole branch. The precise per-experiment before/after still comes from writing the \`--code-change\` payload above, which always wins over the fallback, so keep doing that in \`make-change\`; the automatic diff is a last resort, not a reason to skip it.
 
 **Check the \`supportsExperimentGroups\` flag** (from \`detect-replay-capabilities\`). If true and an \`experimentGroupId\` exists, pass \`--experiment-group-id <experimentGroupId>\` (from \`open-experiments-before-replay\`) so the experiment is tagged with the group. This includes the initial \`fixReplayScope = "single-trace"\` pass. If false, skip the flag; the post-replay Bitfab inspection can fall back to \`experimentId\` when available.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bitfab-cli",
-  "version": "0.2.405",
+  "version": "0.2.406",
   "description": "Install and configure the Bitfab plugin in Claude Code, Codex, or Cursor, or the Bitfab skill pack in Amp (Alpha).",
   "type": "module",
   "main": "dist/index.js",