diff --git a/README.md b/README.md
index a4f188f46..a862707b5 100644
--- a/README.md
+++ b/README.md
@@ -513,7 +513,7 @@ These are only needed for **headless / CI extraction** (`graphify extract`). Whe
 | `AWS_*` / `~/.aws/credentials` | AWS Bedrock — standard credential chain | `--backend bedrock` (no API key, uses IAM) |
 | `GRAPHIFY_MAX_WORKERS` | AST parallelism thread count | optional — also `--max-workers` flag |
 | `GRAPHIFY_MAX_OUTPUT_TOKENS` | Raise output cap for dense corpora | optional — e.g. `32768` for large files |
-| `GRAPHIFY_API_TIMEOUT` | Per-call timeout in seconds for HTTP, claude-cli, and Anthropic SDK backends (default: 600) | optional — also `--api-timeout` flag |
+| `GRAPHIFY_API_TIMEOUT` | Per-call timeout in seconds for HTTP, claude-cli, Anthropic SDK, and Bedrock backends (default: 600) | optional — also `--api-timeout` flag |
 | `GRAPHIFY_MAX_RETRIES` | How many times to retry a rate-limited (429) request before giving up (default: 6; honors `Retry-After`) | optional — raise for strict per-org limits (e.g. kimi); `0` disables |
 | `GRAPHIFY_FORCE` | Force graph rebuild even with fewer nodes | optional — also `--force` flag |
 | `GRAPHIFY_GOOGLE_WORKSPACE` | Auto-enable Google Workspace export | optional — set to `1` |
diff --git a/graphify/llm.py b/graphify/llm.py
index 8c34cde6b..ba26ebcb1 100644
--- a/graphify/llm.py
+++ b/graphify/llm.py
@@ -1609,6 +1609,7 @@ def _call_bedrock(model: str, user_message: str, max_tokens: int = 8192, *, deep
     """Call AWS Bedrock via boto3 Converse API using the standard AWS credential chain."""
     try:
         import boto3
+        import botocore.config
         import botocore.exceptions
     except ImportError as exc:
         raise ImportError(
@@ -1618,7 +1619,19 @@ def _call_bedrock(model: str, user_message: str, max_tokens: int = 8192, *, deep
     region = os.environ.get("AWS_REGION") or os.environ.get("AWS_DEFAULT_REGION") or "us-east-1"
     profile = os.environ.get("AWS_PROFILE")
     session = boto3.Session(profile_name=profile, region_name=region)
-    client = session.client("bedrock-runtime")
+    # Wire GRAPHIFY_API_TIMEOUT into the botocore read timeout. Without an
+    # explicit config, Converse uses botocore's 60s default and a long
+    # generation dies with "Read timeout on endpoint URL" no matter what the
+    # env var / --api-timeout is set to — the same gap #1112/#1442 closed for
+    # the claude-cli and secondary-dispatch paths, on the last cloud backend.
+    client = session.client(
+        "bedrock-runtime",
+        config=botocore.config.Config(
+            read_timeout=_resolve_api_timeout(),
+            connect_timeout=10,
+            retries={"max_attempts": _resolve_max_retries(), "mode": "adaptive"},
+        ),
+    )
 
     try:
         resp = client.converse(
@@ -2565,12 +2578,20 @@ def _rec(inp, out) -> None:
     if backend == "bedrock":
         try:
             import boto3
+            import botocore.config
         except ImportError as exc:
             raise ImportError(_backend_pkg_hint("boto3", "bedrock")) from exc
         region = os.environ.get("AWS_REGION") or os.environ.get("AWS_DEFAULT_REGION") or "us-east-1"
         profile = os.environ.get("AWS_PROFILE")
         session = boto3.Session(profile_name=profile, region_name=region)
-        client = session.client("bedrock-runtime")
+        client = session.client(
+            "bedrock-runtime",
+            config=botocore.config.Config(
+                read_timeout=_resolve_api_timeout(),
+                connect_timeout=10,
+                retries={"max_attempts": _resolve_max_retries(), "mode": "adaptive"},
+            ),
+        )
         resp = client.converse(
             modelId=mdl,
             messages=[{"role": "user", "content": [{"text": prompt}]}],
diff --git a/tests/test_image_vision.py b/tests/test_image_vision.py
index c73b1141e..f9c733a70 100644
--- a/tests/test_image_vision.py
+++ b/tests/test_image_vision.py
@@ -279,15 +279,37 @@ def converse(self, **kw):
                 "usage": {"inputTokens": 1, "outputTokens": 2},
                 "stopReason": "end_turn",
             }
+
+    def _make_client(svc, **kw):
+        # Record the client-construction kwargs (notably config=) separately so
+        # they don't collide with the converse() call kwargs captured above.
+        captured["_client_service"] = svc
+        captured["_client_config"] = kw.get("config")
+        return _Client()
+
     boto3 = types.ModuleType("boto3")
-    boto3.Session = lambda **kw: SimpleNamespace(client=lambda svc: _Client())
+    boto3.Session = lambda **kw: SimpleNamespace(client=_make_client)
     monkeypatch.setitem(sys.modules, "boto3", boto3)
+
     botocore = types.ModuleType("botocore")
     exc = types.ModuleType("botocore.exceptions")
     exc.ClientError = type("ClientError", (Exception,), {})
+    config_mod = types.ModuleType("botocore.config")
+
+    class _Config:
+        def __init__(self, **kw):
+            # Mirror botocore.config.Config: expose the kwargs as attributes so
+            # tests can assert read_timeout/connect_timeout/retries were wired.
+            self.read_timeout = kw.get("read_timeout")
+            self.connect_timeout = kw.get("connect_timeout")
+            self.retries = kw.get("retries")
+
+    config_mod.Config = _Config
     botocore.exceptions = exc
+    botocore.config = config_mod
     monkeypatch.setitem(sys.modules, "botocore", botocore)
     monkeypatch.setitem(sys.modules, "botocore.exceptions", exc)
+    monkeypatch.setitem(sys.modules, "botocore.config", config_mod)
 
 
 # ── backend payload shape (mocked) ────────────────────────────────────────────
@@ -330,6 +352,32 @@ def test_call_bedrock_sends_raw_image_bytes(tmp_path, monkeypatch):
     assert img_block["image"]["source"]["bytes"] == _PNG_BYTES
 
 
+def test_call_bedrock_honors_api_timeout(monkeypatch):
+    # GRAPHIFY_API_TIMEOUT must reach the botocore client's read_timeout; else
+    # Converse falls back to botocore's 60s default and a long generation dies
+    # with "Read timeout on endpoint URL" regardless of the env var.
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "1800")
+    monkeypatch.delenv("GRAPHIFY_MAX_RETRIES", raising=False)
+    captured: dict = {}
+    _fake_boto3(monkeypatch, captured)
+    llm._call_bedrock("model", "CORPUS")
+    cfg = captured["_client_config"]
+    assert cfg is not None, "bedrock client built without a botocore config"
+    assert cfg.read_timeout == 1800.0
+    assert cfg.connect_timeout == 10
+    assert cfg.retries == {"max_attempts": 6, "mode": "adaptive"}
+
+
+def test_call_bedrock_api_timeout_defaults_when_unset(monkeypatch):
+    # With no override the client still gets an explicit 600s read timeout,
+    # not botocore's silent 60s default.
+    monkeypatch.delenv("GRAPHIFY_API_TIMEOUT", raising=False)
+    captured: dict = {}
+    _fake_boto3(monkeypatch, captured)
+    llm._call_bedrock("model", "CORPUS")
+    assert captured["_client_config"].read_timeout == 600.0
+
+
 # ── CLI backends (mocked subprocess) ──────────────────────────────────────────