Merge branch 'main' into alexcheema/trust-remote-code-cli-flag

The constant is the default for built-in models with known model cards,
which are trusted. Custom models added via API already default to
trust_remote_code=False in ModelCard.fetch_from_hf(). The CLI flag
overrides custom models only.

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

src/exo/main.py

@@ -261,6 +261,13 @@ def main():
     if args.offline:
         logger.info("Running in OFFLINE mode — no internet checks, local models only")
 
+    # Set trust_remote_code override env var for runner subprocesses
+    if args.trust_remote_code:
+        os.environ["EXO_TRUST_REMOTE_CODE"] = "1"
+        logger.warning(
+            "--trust-remote-code enabled: models may execute arbitrary code during loading"
+        )
+
     # Set FAST_SYNCH override env var for runner subprocesses
     if args.fast_synch is True:
         os.environ["EXO_FAST_SYNCH"] = "on"
@@ -285,6 +292,7 @@ class Args(CamelCaseModel):
     no_downloads: bool = False
     offline: bool = False
     fast_synch: bool | None = None  # None = auto, True = force on, False = force off
+    trust_remote_code: bool = False
 
     @classmethod
     def parse(cls) -> Self:
@@ -336,6 +344,11 @@ class Args(CamelCaseModel):
             action="store_true",
             help="Run in offline/air-gapped mode: skip internet checks, use only pre-staged local models",
         )
+        parser.add_argument(
+            "--trust-remote-code",
+            action="store_true",
+            help="Allow models to execute custom code during tokenizer loading (security-sensitive, CLI-only)",
+        )
         fast_synch_group = parser.add_mutually_exclusive_group()
         fast_synch_group.add_argument(
             "--fast-synch",

src/exo/worker/engines/mlx/constants.py

@@ -13,5 +13,6 @@ KV_CACHE_BITS: int | None = None
 
 DEFAULT_TOP_LOGPROBS: int = 5
 
-# TODO: We should really make this opt-in, but Kimi requires trust_remote_code=True
+# True for built-in models with known model cards; custom models added via API default to False
+# and can be overridden with the --trust-remote-code CLI flag.
 TRUST_REMOTE_CODE: bool = True

src/exo/worker/engines/mlx/utils_mlx.py

@@ -291,10 +291,14 @@ def shard_and_load(
 
 def get_tokenizer(model_path: Path, shard_metadata: ShardMetadata) -> TokenizerWrapper:
     """Load tokenizer for a model shard. Delegates to load_tokenizer_for_model_id."""
+    trust_remote_code = (
+        shard_metadata.model_card.trust_remote_code
+        or os.environ.get("EXO_TRUST_REMOTE_CODE") == "1"
+    )
     return load_tokenizer_for_model_id(
         shard_metadata.model_card.model_id,
         model_path,
-        trust_remote_code=shard_metadata.model_card.trust_remote_code,
+        trust_remote_code=trust_remote_code,
     )