Merge pull request #2091 from dusterbloom/fix/lmstudio-context-length-detection

feat: query local servers for actual context window size

agent/model_metadata.py

@@ -146,6 +146,9 @@ _MAX_COMPLETION_KEYS = (
     "max_tokens",
 )
 
+# Local server hostnames / address patterns
+_LOCAL_HOSTS = ("localhost", "127.0.0.1", "::1", "0.0.0.0")
+
 
 def _normalize_base_url(base_url: str) -> str:
     return (base_url or "").strip().rstrip("/")
@@ -178,6 +181,99 @@ def _is_known_provider_base_url(base_url: str) -> bool:
     return any(known_host in host for known_host in known_hosts)
 
 
+def is_local_endpoint(base_url: str) -> bool:
+    """Return True if base_url points to a local machine (localhost / RFC-1918 / WSL)."""
+    normalized = _normalize_base_url(base_url)
+    if not normalized:
+        return False
+    url = normalized if "://" in normalized else f"http://{normalized}"
+    try:
+        parsed = urlparse(url)
+        host = parsed.hostname or ""
+    except Exception:
+        return False
+    if host in _LOCAL_HOSTS:
+        return True
+    # RFC-1918 private ranges and link-local
+    import ipaddress
+    try:
+        addr = ipaddress.ip_address(host)
+        return addr.is_private or addr.is_loopback or addr.is_link_local
+    except ValueError:
+        pass
+    # Bare IP that looks like a private range (e.g. 172.26.x.x for WSL)
+    parts = host.split(".")
+    if len(parts) == 4:
+        try:
+            first, second = int(parts[0]), int(parts[1])
+            if first == 10:
+                return True
+            if first == 172 and 16 <= second <= 31:
+                return True
+            if first == 192 and second == 168:
+                return True
+        except ValueError:
+            pass
+    return False
+
+
+def detect_local_server_type(base_url: str) -> Optional[str]:
+    """Detect which local server is running at base_url by probing known endpoints.
+
+    Returns one of: "ollama", "lm-studio", "vllm", "llamacpp", or None.
+    """
+    import httpx
+
+    normalized = _normalize_base_url(base_url)
+    server_url = normalized
+    if server_url.endswith("/v1"):
+        server_url = server_url[:-3]
+
+    try:
+        with httpx.Client(timeout=2.0) as client:
+            # LM Studio exposes /api/v1/models — check first (most specific)
+            try:
+                r = client.get(f"{server_url}/api/v1/models")
+                if r.status_code == 200:
+                    return "lm-studio"
+            except Exception:
+                pass
+            # Ollama exposes /api/tags and responds with {"models": [...]}
+            # LM Studio returns {"error": "Unexpected endpoint"} with status 200
+            # on this path, so we must verify the response contains "models".
+            try:
+                r = client.get(f"{server_url}/api/tags")
+                if r.status_code == 200:
+                    try:
+                        data = r.json()
+                        if "models" in data:
+                            return "ollama"
+                    except Exception:
+                        pass
+            except Exception:
+                pass
+            # llama.cpp exposes /props
+            try:
+                r = client.get(f"{server_url}/props")
+                if r.status_code == 200 and "default_generation_settings" in r.text:
+                    return "llamacpp"
+            except Exception:
+                pass
+            # vLLM: /version
+            try:
+                r = client.get(f"{server_url}/version")
+                if r.status_code == 200:
+                    data = r.json()
+                    if "version" in data:
+                        return "vllm"
+            except Exception:
+                pass
+    except Exception:
+        pass
+
+    return None
+
+
 def _iter_nested_dicts(value: Any):
     if isinstance(value, dict):
         yield value
@@ -383,7 +479,7 @@ def _get_context_cache_path() -> Path:
 
 
 def _load_context_cache() -> Dict[str, int]:
-    """Load the model+provider → context_length cache from disk."""
+    """Load the model+provider -> context_length cache from disk."""
     path = _get_context_cache_path()
     if not path.exists():
         return {}
@@ -412,7 +508,7 @@ def save_context_length(model: str, base_url: str, length: int) -> None:
         path.parent.mkdir(parents=True, exist_ok=True)
         with open(path, "w") as f:
             yaml.dump({"context_lengths": cache}, f, default_flow_style=False)
-        logger.info("Cached context length %s → %s tokens", key, f"{length:,}")
+        logger.info("Cached context length %s -> %s tokens", key, f"{length:,}")
     except Exception as e:
         logger.debug("Failed to save context length cache: %s", e)
 
@@ -460,6 +556,116 @@ def parse_context_limit_from_error(error_msg: str) -> Optional[int]:
     return None
 
 
+def _model_id_matches(candidate_id: str, lookup_model: str) -> bool:
+    """Return True if *candidate_id* (from server) matches *lookup_model* (configured).
+
+    Supports two forms:
+    - Exact match:  "nvidia-nemotron-super-49b-v1" == "nvidia-nemotron-super-49b-v1"
+    - Slug match:   "nvidia/nvidia-nemotron-super-49b-v1" matches "nvidia-nemotron-super-49b-v1"
+                    (the part after the last "/" equals lookup_model)
+
+    This covers LM Studio's native API which stores models as "publisher/slug"
+    while users typically configure only the slug after the "local:" prefix.
+    """
+    if candidate_id == lookup_model:
+        return True
+    # Slug match: basename of candidate equals the lookup name
+    if "/" in candidate_id and candidate_id.rsplit("/", 1)[1] == lookup_model:
+        return True
+    return False
+
+
+def _query_local_context_length(model: str, base_url: str) -> Optional[int]:
+    """Query a local server for the model's context length."""
+    import httpx
+
+    # Strip provider prefix (e.g., "local:model-name" → "model-name").
+    # LM Studio and Ollama don't use provider prefixes in their model IDs.
+    if ":" in model and not model.startswith("http"):
+        model = model.split(":", 1)[1]
+
+    # Strip /v1 suffix to get the server root
+    server_url = base_url.rstrip("/")
+    if server_url.endswith("/v1"):
+        server_url = server_url[:-3]
+
+    try:
+        server_type = detect_local_server_type(base_url)
+    except Exception:
+        server_type = None
+
+    try:
+        with httpx.Client(timeout=3.0) as client:
+            # Ollama: /api/show returns model details with context info
+            if server_type == "ollama":
+                resp = client.post(f"{server_url}/api/show", json={"name": model})
+                if resp.status_code == 200:
+                    data = resp.json()
+                    # Check model_info for context length
+                    model_info = data.get("model_info", {})
+                    for key, value in model_info.items():
+                        if "context_length" in key and isinstance(value, (int, float)):
+                            return int(value)
+                    # Check parameters string for num_ctx
+                    params = data.get("parameters", "")
+                    if "num_ctx" in params:
+                        for line in params.split("\n"):
+                            if "num_ctx" in line:
+                                parts = line.strip().split()
+                                if len(parts) >= 2:
+                                    try:
+                                        return int(parts[-1])
+                                    except ValueError:
+                                        pass
+
+            # LM Studio native API: /api/v1/models returns max_context_length.
+            # This is more reliable than the OpenAI-compat /v1/models which
+            # doesn't include context window information for LM Studio servers.
+            # Use _model_id_matches for fuzzy matching: LM Studio stores models as
+            # "publisher/slug" but users configure only "slug" after "local:" prefix.
+            if server_type == "lm-studio":
+                resp = client.get(f"{server_url}/api/v1/models")
+                if resp.status_code == 200:
+                    data = resp.json()
+                    for m in data.get("models", []):
+                        if _model_id_matches(m.get("key", ""), model) or _model_id_matches(m.get("id", ""), model):
+                            # Prefer loaded instance context (actual runtime value)
+                            for inst in m.get("loaded_instances", []):
+                                cfg = inst.get("config", {})
+                                ctx = cfg.get("context_length")
+                                if ctx and isinstance(ctx, (int, float)):
+                                    return int(ctx)
+                            # Fall back to max_context_length (theoretical model max)
+                            ctx = m.get("max_context_length") or m.get("context_length")
+                            if ctx and isinstance(ctx, (int, float)):
+                                return int(ctx)
+
+            # LM Studio / vLLM / llama.cpp: try /v1/models/{model}
+            resp = client.get(f"{server_url}/v1/models/{model}")
+            if resp.status_code == 200:
+                data = resp.json()
+                # vLLM returns max_model_len
+                ctx = data.get("max_model_len") or data.get("context_length") or data.get("max_tokens")
+                if ctx and isinstance(ctx, (int, float)):
+                    return int(ctx)
+
+            # Try /v1/models and find the model in the list.
+            # Use _model_id_matches to handle "publisher/slug" vs bare "slug".
+            resp = client.get(f"{server_url}/v1/models")
+            if resp.status_code == 200:
+                data = resp.json()
+                models_list = data.get("data", [])
+                for m in models_list:
+                    if _model_id_matches(m.get("id", ""), model):
+                        ctx = m.get("max_model_len") or m.get("context_length") or m.get("max_tokens")
+                        if ctx and isinstance(ctx, (int, float)):
+                            return int(ctx)
+    except Exception:
+        pass
+
+    return None
+
+
 def get_model_context_length(
     model: str,
     base_url: str = "",
@@ -472,14 +678,21 @@ def get_model_context_length(
     0. Explicit config override (model.context_length in config.yaml)
     1. Persistent cache (previously discovered via probing)
     2. Active endpoint metadata (/models for explicit custom endpoints)
-    3. OpenRouter API metadata
-    4. Hardcoded DEFAULT_CONTEXT_LENGTHS (fuzzy match for hosted routes only)
-    5. First probe tier (2M) — will be narrowed on first context error
+    3. Local server query (for local endpoints when model not in /models list)
+    4. OpenRouter API metadata
+    5. Hardcoded DEFAULT_CONTEXT_LENGTHS (fuzzy match for hosted routes only)
+    6. First probe tier (2M) — will be narrowed on first context error
     """
     # 0. Explicit config override — user knows best
     if config_context_length is not None and isinstance(config_context_length, int) and config_context_length > 0:
         return config_context_length
 
+    # Normalise provider-prefixed model names (e.g. "local:model-name" →
+    # "model-name") so cache lookups and server queries use the bare ID that
+    # local servers actually know about.
+    if ":" in model and not model.startswith("http"):
+        model = model.split(":", 1)[1]
+
     # 1. Check persistent cache (model+provider)
     if base_url:
         cached = get_cached_context_length(model, base_url)
@@ -507,6 +720,12 @@ def get_model_context_length(
         if not _is_known_provider_base_url(base_url):
             # Explicit third-party endpoints should not borrow fuzzy global
             # defaults from unrelated providers with similarly named models.
+            # But first try querying the local server directly.
+            if is_local_endpoint(base_url):
+                local_ctx = _query_local_context_length(model, base_url)
+                if local_ctx and local_ctx > 0:
+                    save_context_length(model, base_url, local_ctx)
+                    return local_ctx
             logger.info(
                 "Could not detect context length for model %r at %s — "
                 "defaulting to %s tokens (probe-down). Set model.context_length "
@@ -527,7 +746,14 @@ def get_model_context_length(
         if default_model in model or model in default_model:
             return length
 
-    # 5. Unknown model — start at highest probe tier
+    # 5. Query local server for unknown models before defaulting to 2M
+    if base_url and is_local_endpoint(base_url):
+        local_ctx = _query_local_context_length(model, base_url)
+        if local_ctx and local_ctx > 0:
+            save_context_length(model, base_url, local_ctx)
+            return local_ctx
+
+    # 6. Unknown model — start at highest probe tier
     return CONTEXT_PROBE_TIERS[0]
 
 

run_agent.py

@@ -6569,7 +6569,21 @@ class AIAgent:
                                 self._response_was_previewed = True
                                 break
                             
-                            # No fallback -- append the empty message as-is
+                            # No fallback -- if reasoning_text exists, the model put its
+                            # entire response inside <think> tags; use that as the content.
+                            if reasoning_text:
+                                self._vprint(f"{self.log_prefix}Using reasoning as response content (model wrapped entire response in think tags).", force=True)
+                                final_response = reasoning_text
+                                empty_msg = {
+                                    "role": "assistant",
+                                    "content": final_response,
+                                    "reasoning": reasoning_text,
+                                    "finish_reason": finish_reason,
+                                }
+                                messages.append(empty_msg)
+                                break
+
+                            # Truly empty -- no reasoning and no content
                             empty_msg = {
                                 "role": "assistant",
                                 "content": final_response,
@@ -6577,10 +6591,10 @@ class AIAgent:
                                 "finish_reason": finish_reason,
                             }
                             messages.append(empty_msg)
-                            
+
                             self._cleanup_task_resources(effective_task_id)
                             self._persist_session(messages, conversation_history)
-                            
+
                             return {
                                 "final_response": final_response or None,
                                 "messages": messages,

tests/test_model_metadata_local_ctx.py

@@ -0,0 +1,493 @@
+"""Tests for _query_local_context_length and the local server fallback in
+get_model_context_length.
+
+All tests use synthetic inputs — no filesystem or live server required.
+"""
+
+import sys
+import os
+import json
+from unittest.mock import MagicMock, patch
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
+
+import pytest
+
+
+# ---------------------------------------------------------------------------
+# _query_local_context_length — unit tests with mocked httpx
+# ---------------------------------------------------------------------------
+
+class TestQueryLocalContextLengthOllama:
+    """_query_local_context_length with server_type == 'ollama'."""
+
+    def _make_resp(self, status_code, body):
+        resp = MagicMock()
+        resp.status_code = status_code
+        resp.json.return_value = body
+        return resp
+
+    def test_ollama_model_info_context_length(self):
+        """Reads context length from model_info dict in /api/show response."""
+        from agent.model_metadata import _query_local_context_length
+
+        show_resp = self._make_resp(200, {
+            "model_info": {"llama.context_length": 131072}
+        })
+        models_resp = self._make_resp(404, {})
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = show_resp
+        client_mock.get.return_value = models_resp
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="ollama"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        assert result == 131072
+
+    def test_ollama_parameters_num_ctx(self):
+        """Falls back to num_ctx in parameters string when model_info lacks context_length."""
+        from agent.model_metadata import _query_local_context_length
+
+        show_resp = self._make_resp(200, {
+            "model_info": {},
+            "parameters": "num_ctx 32768\ntemperature 0.7\n"
+        })
+        models_resp = self._make_resp(404, {})
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = show_resp
+        client_mock.get.return_value = models_resp
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="ollama"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("some-model", "http://localhost:11434/v1")
+
+        assert result == 32768
+
+    def test_ollama_show_404_falls_through(self):
+        """When /api/show returns 404, falls through to /v1/models/{model}."""
+        from agent.model_metadata import _query_local_context_length
+
+        show_resp = self._make_resp(404, {})
+        model_detail_resp = self._make_resp(200, {"max_model_len": 65536})
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = show_resp
+        client_mock.get.return_value = model_detail_resp
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="ollama"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("some-model", "http://localhost:11434/v1")
+
+        assert result == 65536
+
+
+class TestQueryLocalContextLengthVllm:
+    """_query_local_context_length with vLLM-style /v1/models/{model} response."""
+
+    def _make_resp(self, status_code, body):
+        resp = MagicMock()
+        resp.status_code = status_code
+        resp.json.return_value = body
+        return resp
+
+    def test_vllm_max_model_len(self):
+        """Reads max_model_len from /v1/models/{model} response."""
+        from agent.model_metadata import _query_local_context_length
+
+        detail_resp = self._make_resp(200, {"id": "omnicoder-9b", "max_model_len": 100000})
+        list_resp = self._make_resp(404, {})
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = self._make_resp(404, {})
+        client_mock.get.return_value = detail_resp
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="vllm"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("omnicoder-9b", "http://localhost:8000/v1")
+
+        assert result == 100000
+
+    def test_vllm_context_length_key(self):
+        """Reads context_length from /v1/models/{model} response."""
+        from agent.model_metadata import _query_local_context_length
+
+        detail_resp = self._make_resp(200, {"id": "some-model", "context_length": 32768})
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = self._make_resp(404, {})
+        client_mock.get.return_value = detail_resp
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="vllm"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("some-model", "http://localhost:8000/v1")
+
+        assert result == 32768
+
+
+class TestQueryLocalContextLengthModelsList:
+    """_query_local_context_length: falls back to /v1/models list."""
+
+    def _make_resp(self, status_code, body):
+        resp = MagicMock()
+        resp.status_code = status_code
+        resp.json.return_value = body
+        return resp
+
+    def test_models_list_max_model_len(self):
+        """Finds context length for model in /v1/models list."""
+        from agent.model_metadata import _query_local_context_length
+
+        detail_resp = self._make_resp(404, {})
+        list_resp = self._make_resp(200, {
+            "data": [
+                {"id": "other-model", "max_model_len": 4096},
+                {"id": "omnicoder-9b", "max_model_len": 131072},
+            ]
+        })
+
+        call_count = [0]
+        def side_effect(url, **kwargs):
+            call_count[0] += 1
+            if call_count[0] == 1:
+                return detail_resp  # /v1/models/omnicoder-9b
+            return list_resp  # /v1/models
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = self._make_resp(404, {})
+        client_mock.get.side_effect = side_effect
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value=None), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("omnicoder-9b", "http://localhost:1234")
+
+        assert result == 131072
+
+    def test_models_list_model_not_found_returns_none(self):
+        """Returns None when model is not in the /v1/models list."""
+        from agent.model_metadata import _query_local_context_length
+
+        detail_resp = self._make_resp(404, {})
+        list_resp = self._make_resp(200, {
+            "data": [{"id": "other-model", "max_model_len": 4096}]
+        })
+
+        call_count = [0]
+        def side_effect(url, **kwargs):
+            call_count[0] += 1
+            if call_count[0] == 1:
+                return detail_resp
+            return list_resp
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = self._make_resp(404, {})
+        client_mock.get.side_effect = side_effect
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value=None), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("omnicoder-9b", "http://localhost:1234")
+
+        assert result is None
+
+
+class TestQueryLocalContextLengthLmStudio:
+    """_query_local_context_length with LM Studio native /api/v1/models response."""
+
+    def _make_resp(self, status_code, body):
+        resp = MagicMock()
+        resp.status_code = status_code
+        resp.json.return_value = body
+        return resp
+
+    def _make_client(self, native_resp, detail_resp, list_resp):
+        """Build a mock httpx.Client with sequenced GET responses."""
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.return_value = self._make_resp(404, {})
+
+        responses = [native_resp, detail_resp, list_resp]
+        call_idx = [0]
+
+        def get_side_effect(url, **kwargs):
+            idx = call_idx[0]
+            call_idx[0] += 1
+            if idx < len(responses):
+                return responses[idx]
+            return self._make_resp(404, {})
+
+        client_mock.get.side_effect = get_side_effect
+        return client_mock
+
+    def test_lmstudio_exact_key_match(self):
+        """Reads max_context_length when key matches exactly."""
+        from agent.model_metadata import _query_local_context_length
+
+        native_resp = self._make_resp(200, {
+            "models": [
+                {"key": "nvidia/nvidia-nemotron-super-49b-v1", "id": "nvidia/nvidia-nemotron-super-49b-v1",
+                 "max_context_length": 131072},
+            ]
+        })
+        client_mock = self._make_client(
+            native_resp,
+            self._make_resp(404, {}),
+            self._make_resp(404, {}),
+        )
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="lm-studio"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length(
+                "nvidia/nvidia-nemotron-super-49b-v1", "http://192.168.1.22:1234/v1"
+            )
+
+        assert result == 131072
+
+    def test_lmstudio_slug_only_matches_key_with_publisher_prefix(self):
+        """Fuzzy match: bare model slug matches key that includes publisher prefix.
+
+        When the user configures the model as "local:nvidia-nemotron-super-49b-v1"
+        (slug only, no publisher), but LM Studio's native API stores it as
+        "nvidia/nvidia-nemotron-super-49b-v1", the lookup must still succeed.
+        """
+        from agent.model_metadata import _query_local_context_length
+
+        native_resp = self._make_resp(200, {
+            "models": [
+                {"key": "nvidia/nvidia-nemotron-super-49b-v1",
+                 "id": "nvidia/nvidia-nemotron-super-49b-v1",
+                 "max_context_length": 131072},
+            ]
+        })
+        client_mock = self._make_client(
+            native_resp,
+            self._make_resp(404, {}),
+            self._make_resp(404, {}),
+        )
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="lm-studio"), \
+             patch("httpx.Client", return_value=client_mock):
+            # Model passed in is just the slug after stripping "local:" prefix
+            result = _query_local_context_length(
+                "nvidia-nemotron-super-49b-v1", "http://192.168.1.22:1234/v1"
+            )
+
+        assert result == 131072
+
+    def test_lmstudio_v1_models_list_slug_fuzzy_match(self):
+        """Fuzzy match also works for /v1/models list when exact match fails.
+
+        LM Studio's OpenAI-compat /v1/models returns id like
+        "nvidia/nvidia-nemotron-super-49b-v1" — must match bare slug.
+        """
+        from agent.model_metadata import _query_local_context_length
+
+        # native /api/v1/models: no match
+        native_resp = self._make_resp(404, {})
+        # /v1/models/{model}: no match
+        detail_resp = self._make_resp(404, {})
+        # /v1/models list: model found with publisher prefix, includes context_length
+        list_resp = self._make_resp(200, {
+            "data": [
+                {"id": "nvidia/nvidia-nemotron-super-49b-v1", "context_length": 131072},
+            ]
+        })
+        client_mock = self._make_client(native_resp, detail_resp, list_resp)
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="lm-studio"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length(
+                "nvidia-nemotron-super-49b-v1", "http://192.168.1.22:1234/v1"
+            )
+
+        assert result == 131072
+
+    def test_lmstudio_loaded_instances_context_length(self):
+        """Reads active context_length from loaded_instances when max_context_length absent."""
+        from agent.model_metadata import _query_local_context_length
+
+        native_resp = self._make_resp(200, {
+            "models": [
+                {
+                    "key": "nvidia/nvidia-nemotron-super-49b-v1",
+                    "id": "nvidia/nvidia-nemotron-super-49b-v1",
+                    "loaded_instances": [
+                        {"config": {"context_length": 65536}},
+                    ],
+                },
+            ]
+        })
+        client_mock = self._make_client(
+            native_resp,
+            self._make_resp(404, {}),
+            self._make_resp(404, {}),
+        )
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="lm-studio"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length(
+                "nvidia-nemotron-super-49b-v1", "http://192.168.1.22:1234/v1"
+            )
+
+        assert result == 65536
+
+    def test_lmstudio_loaded_instance_beats_max_context_length(self):
+        """loaded_instances context_length takes priority over max_context_length.
+
+        LM Studio may show max_context_length=1_048_576 (theoretical model max)
+        while the actual loaded context is 122_651 (runtime setting). The loaded
+        value is the real constraint and must be preferred.
+        """
+        from agent.model_metadata import _query_local_context_length
+
+        native_resp = self._make_resp(200, {
+            "models": [
+                {
+                    "key": "nvidia/nvidia-nemotron-3-nano-4b",
+                    "id": "nvidia/nvidia-nemotron-3-nano-4b",
+                    "max_context_length": 1_048_576,
+                    "loaded_instances": [
+                        {"config": {"context_length": 122_651}},
+                    ],
+                },
+            ]
+        })
+        client_mock = self._make_client(
+            native_resp,
+            self._make_resp(404, {}),
+            self._make_resp(404, {}),
+        )
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value="lm-studio"), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length(
+                "nvidia-nemotron-3-nano-4b", "http://192.168.1.22:1234/v1"
+            )
+
+        assert result == 122_651, (
+            f"Expected loaded instance context (122651) but got {result}. "
+            "max_context_length (1048576) must not win over loaded_instances."
+        )
+
+
+class TestQueryLocalContextLengthNetworkError:
+    """_query_local_context_length handles network failures gracefully."""
+
+    def test_connection_error_returns_none(self):
+        """Returns None when the server is unreachable."""
+        from agent.model_metadata import _query_local_context_length
+
+        client_mock = MagicMock()
+        client_mock.__enter__ = lambda s: client_mock
+        client_mock.__exit__ = MagicMock(return_value=False)
+        client_mock.post.side_effect = Exception("Connection refused")
+        client_mock.get.side_effect = Exception("Connection refused")
+
+        with patch("agent.model_metadata.detect_local_server_type", return_value=None), \
+             patch("httpx.Client", return_value=client_mock):
+            result = _query_local_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        assert result is None
+
+
+# ---------------------------------------------------------------------------
+# get_model_context_length — integration-style tests with mocked helpers
+# ---------------------------------------------------------------------------
+
+class TestGetModelContextLengthLocalFallback:
+    """get_model_context_length uses local server query before falling back to 2M."""
+
+    def test_local_endpoint_unknown_model_queries_server(self):
+        """Unknown model on local endpoint gets ctx from server, not 2M default."""
+        from agent.model_metadata import get_model_context_length
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=None), \
+             patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             patch("agent.model_metadata.is_local_endpoint", return_value=True), \
+             patch("agent.model_metadata._query_local_context_length", return_value=131072), \
+             patch("agent.model_metadata.save_context_length") as mock_save:
+            result = get_model_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        assert result == 131072
+
+    def test_local_endpoint_unknown_model_result_is_cached(self):
+        """Context length returned from local server is persisted to cache."""
+        from agent.model_metadata import get_model_context_length
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=None), \
+             patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             patch("agent.model_metadata.is_local_endpoint", return_value=True), \
+             patch("agent.model_metadata._query_local_context_length", return_value=131072), \
+             patch("agent.model_metadata.save_context_length") as mock_save:
+            get_model_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        mock_save.assert_called_once_with("omnicoder-9b", "http://localhost:11434/v1", 131072)
+
+    def test_local_endpoint_server_returns_none_falls_back_to_2m(self):
+        """When local server returns None, still falls back to 2M probe tier."""
+        from agent.model_metadata import get_model_context_length, CONTEXT_PROBE_TIERS
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=None), \
+             patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             patch("agent.model_metadata.is_local_endpoint", return_value=True), \
+             patch("agent.model_metadata._query_local_context_length", return_value=None):
+            result = get_model_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        assert result == CONTEXT_PROBE_TIERS[0]
+
+    def test_non_local_endpoint_does_not_query_local_server(self):
+        """For non-local endpoints, _query_local_context_length is not called."""
+        from agent.model_metadata import get_model_context_length, CONTEXT_PROBE_TIERS
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=None), \
+             patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             patch("agent.model_metadata.is_local_endpoint", return_value=False), \
+             patch("agent.model_metadata._query_local_context_length") as mock_query:
+            result = get_model_context_length(
+                "unknown-model", "https://some-cloud-api.example.com/v1"
+            )
+
+        mock_query.assert_not_called()
+
+    def test_cached_result_skips_local_query(self):
+        """Cached context length is returned without querying the local server."""
+        from agent.model_metadata import get_model_context_length
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=65536), \
+             patch("agent.model_metadata._query_local_context_length") as mock_query:
+            result = get_model_context_length("omnicoder-9b", "http://localhost:11434/v1")
+
+        assert result == 65536
+        mock_query.assert_not_called()
+
+    def test_no_base_url_does_not_query_local_server(self):
+        """When base_url is empty, local server is not queried."""
+        from agent.model_metadata import get_model_context_length
+
+        with patch("agent.model_metadata.get_cached_context_length", return_value=None), \
+             patch("agent.model_metadata.fetch_endpoint_model_metadata", return_value={}), \
+             patch("agent.model_metadata.fetch_model_metadata", return_value={}), \
+             patch("agent.model_metadata._query_local_context_length") as mock_query:
+            result = get_model_context_length("unknown-xyz-model", "")
+
+        mock_query.assert_not_called()