Sync

.changeset/aisdk-allow-system-in-messages-act-extract-observe.md

@@ -0,0 +1,5 @@
+---
+"@browserbasehq/stagehand": patch
+---
+
+Remove the noisy AI SDK "system message in messages" warning from `act()`, `extract()`, and `observe()` (including when the agent's own tools call them internally).

.changeset/config.json

@@ -3,6 +3,7 @@
   "commit": false,
   "fixed": [],
   "linked": [],
+  "ignore": ["browse"],
   "baseBranch": "main",
   "updateInternalDependencies": "patch",
   "access": "public",

.changeset/odysseysbench-eval-suite.md

@@ -0,0 +1,5 @@
+---
+"@browserbasehq/stagehand-evals": minor
+---
+
+Add OdysseysBench as a supported agent benchmark in the evals CLI. OdysseysBench is a 200-task web-agent benchmark (45 easy / 46 medium / 109 hard); each task ships a weighted rubric that is baked into the verifier's `precomputed_rubric` format so process + outcome are scored against the published criteria. Run with `--eval-name agent/odysseysbench` (or the `external_agent_benchmarks` category); supports `EVAL_ODYSSEYSBENCH_LIMIT`, `EVAL_ODYSSEYSBENCH_SAMPLE`, `EVAL_ODYSSEYSBENCH_LEVEL`, and `EVAL_ODYSSEYSBENCH_IDS`.

.changeset/stagehand-server-v3-catchup-release.md

@@ -0,0 +1,5 @@
+---
+"@browserbasehq/stagehand-server-v3": patch
+---
+
+Cut a new stagehand-server-v3 SEA binary release to catch up with recent core changes, including Gemini 3.5 Flash computer-use support.

.changeset/wobbly-otters-deprecate.md

@@ -0,0 +1,5 @@
+---
+"browse": patch
+---
+
+Warn when auto-loading variables from `.env` and add `BROWSE_LOAD_DOTENV` to opt out ahead of a future release where this will be off by default.