Add custom proxy support for OpenAI-compatible APIs

README.md

@@ -85,10 +85,11 @@ Inside interactive mode, switch with `/model`:
 /model ollama/llama3.1:8b
 /model lm_studio/google/gemma-3-4b
 /model llamacpp/llama-3.1-8b-instruct
+/model custom-proxy/Azure AI/gpt-5.5
 ```
 
-Supported local prefixes are `ollama/`, `vllm/`, `lm_studio/`, and
-`llamacpp/`.
+Supported local/custom prefixes are `ollama/`, `vllm/`, `lm_studio/`,
+`llamacpp/`, and `custom-proxy/`.
 
 ```bash
 LOCAL_LLM_BASE_URL=http://localhost:8000
@@ -99,7 +100,17 @@ Set `LOCAL_LLM_BASE_URL` and optional `LOCAL_LLM_API_KEY` to use one shared
 local endpoint, or override a specific provider with its matching `*_BASE_URL`
 / `*_API_KEY` variable, such as `OLLAMA_BASE_URL` or `VLLM_API_KEY`.
 Provider-specific variables take precedence over the shared local variables.
-Base URLs may include or omit `/v1`.
+Base URLs may include or omit `/v1` for the built-in local providers.
+
+For a custom OpenAI-compatible proxy, set `CUSTOM_PROXY_BASE_URL` to the exact
+OpenAI-compatible base URL and optional `CUSTOM_PROXY_API_KEY`, then use
+`custom-proxy/<model>`:
+
+```bash
+CUSTOM_PROXY_BASE_URL="https://proxy.test/api/v1/"
+CUSTOM_PROXY_API_KEY=<optional-proxy-api-key>
+ml-intern --model "custom-proxy/Azure AI/gpt-5.5" "your prompt"
+```
 
 **CLI tool runtime:**
 

agent/core/llm_params.py

@@ -75,14 +75,21 @@ def _resolve_local_model_params(
         or os.environ.get(LOCAL_MODEL_BASE_URL_ENV)
         or provider["base_url_default"]
     )
+    if not raw_base:
+        raise ValueError(
+            f"Set {provider['base_url_env']} or {LOCAL_MODEL_BASE_URL_ENV} "
+            f"to use {model_name}"
+        )
     api_key = (
         os.environ.get(provider["api_key_env"])
         or os.environ.get(LOCAL_MODEL_API_KEY_ENV)
         or LOCAL_MODEL_API_KEY_DEFAULT
     )
     return {
         "model": f"openai/{local_name}",
-        "api_base": _local_api_base(raw_base),
+        "api_base": _local_api_base(raw_base)
+        if provider.get("base_url_mode") != "exact"
+        else raw_base.strip().rstrip("/"),
         "api_key": api_key,
     }
 
@@ -96,11 +103,11 @@ def _resolve_llm_params(
     """
     Build LiteLLM kwargs for a given model id.
 
-    • ``ollama/<model>``, ``vllm/<model>``, ``lm_studio/<model>``, and
-      ``llamacpp/<model>`` — local OpenAI-compatible endpoints. The id prefix
-      selects a configurable localhost base URL, and the model suffix is sent
-      to LiteLLM as ``openai/<model>``. These endpoints don't receive
-      ``reasoning_effort``.
+    • ``ollama/<model>``, ``vllm/<model>``, ``lm_studio/<model>``,
+      ``llamacpp/<model>``, and ``custom-proxy/<model>`` — local or custom
+      OpenAI-compatible endpoints. The id prefix selects a configurable base
+      URL, and the model suffix is sent to LiteLLM as ``openai/<model>``.
+      These endpoints don't receive ``reasoning_effort``.
 
     • Anything else is treated as an HF Router id. We hit the auto-routing
       OpenAI-compatible endpoint at ``https://router.huggingface.co/v1``.

agent/core/local_models.py

@@ -21,6 +21,12 @@
         "base_url_default": "http://localhost:8080",
         "api_key_env": "LLAMACPP_API_KEY",
     },
+    "custom-proxy/": {
+        "base_url_env": "CUSTOM_PROXY_BASE_URL",
+        "base_url_default": "",
+        "api_key_env": "CUSTOM_PROXY_API_KEY",
+        "base_url_mode": "exact",
+    },
 }
 
 LOCAL_MODEL_PREFIXES = tuple(LOCAL_MODEL_PROVIDERS)
@@ -48,10 +54,15 @@ def local_model_name(model_id: str) -> str | None:
 
 
 def is_local_model_id(model_id: str) -> bool:
-    """Return True for non-empty, whitespace-free local model ids."""
-    if not model_id or any(char.isspace() for char in model_id):
+    """Return True for valid local/custom model ids."""
+    if not model_id:
+        return False
+    name = local_model_name(model_id)
+    if name is None:
         return False
-    return local_model_name(model_id) is not None
+    if model_id.startswith("custom-proxy/"):
+        return bool(name.strip())
+    return not any(char.isspace() for char in model_id)
 
 
 def is_reserved_local_model_id(model_id: str) -> bool:

agent/core/model_switcher.py

@@ -59,7 +59,8 @@ def is_valid_model_id(model_id: str) -> bool:
     """Loose format check — lets users pick any model id.
 
     Accepts:
-      • ollama/<model>, vllm/<model>, lm_studio/<model>, llamacpp/<model>
+      • ollama/<model>, vllm/<model>, lm_studio/<model>, llamacpp/<model>,
+        custom-proxy/<model>
       • <org>/<model>[:<tag>]            (HF router; tag = provider or policy)
       • huggingface/<org>/<model>[:<tag>] (same, optional LiteLLM prefix)
 
@@ -163,8 +164,9 @@ def print_model_listing(config, console) -> None:
     console.print(
         "\n[dim]Paste any HF model id (e.g. 'MiniMaxAI/MiniMax-M2.7').\n"
         "Add ':fastest', ':cheapest', ':preferred', or ':<provider>' to override routing.\n"
-        "Use 'ollama/<model>', 'vllm/<model>', 'lm_studio/<model>', or "
-        "'llamacpp/<model>' for local OpenAI-compatible endpoints.[/dim]"
+        "Use 'ollama/<model>', 'vllm/<model>', 'lm_studio/<model>', "
+        "'llamacpp/<model>', or 'custom-proxy/<model>' for local/custom "
+        "OpenAI-compatible endpoints.[/dim]"
     )
 
 
@@ -173,7 +175,8 @@ def print_invalid_id(arg: str, console) -> None:
     console.print(
         "[dim]Expected:\n"
         "  • <org>/<model>[:tag]    (HF router — paste from huggingface.co)\n"
-        "  • ollama/<model> | vllm/<model> | lm_studio/<model> | llamacpp/<model>[/dim]"
+        "  • ollama/<model> | vllm/<model> | lm_studio/<model> | "
+        "llamacpp/<model> | custom-proxy/<model>[/dim]"
     )
 
 

tests/unit/test_cli_local_models.py

@@ -11,20 +11,23 @@ def test_local_model_helper_accepts_supported_prefixes():
     assert is_local_model_id("vllm/meta-llama/Llama-3.1-8B-Instruct")
     assert is_local_model_id("lm_studio/google/gemma-3-4b")
     assert is_local_model_id("llamacpp/unsloth/Qwen3.5-2B")
+    assert is_local_model_id("custom-proxy/Azure AI/gpt-5.5")
 
 
 def test_model_switcher_accepts_supported_local_prefixes():
     assert model_switcher.is_valid_model_id("ollama/llama3.1:8b")
     assert model_switcher.is_valid_model_id("vllm/meta-llama/Llama-3.1-8B")
     assert model_switcher.is_valid_model_id("lm_studio/google/gemma-3-4b")
     assert model_switcher.is_valid_model_id("llamacpp/llama-3.1-8b")
+    assert model_switcher.is_valid_model_id("custom-proxy/Azure AI/gpt-5.5")
 
 
 def test_model_switcher_rejects_empty_or_whitespace_local_ids():
     assert not model_switcher.is_valid_model_id("ollama/")
     assert not model_switcher.is_valid_model_id("vllm/")
     assert not model_switcher.is_valid_model_id("lm_studio/")
     assert not model_switcher.is_valid_model_id("llamacpp/")
+    assert not model_switcher.is_valid_model_id("custom-proxy/")
     assert not model_switcher.is_valid_model_id("ollama/llama 3.1")
 
 

tests/unit/test_llm_params.py

@@ -137,6 +137,28 @@ def test_resolve_llamacpp_params_strips_provider_prefix(monkeypatch):
     assert params["api_base"] == "http://localhost:8080/v1"
 
 
+def test_resolve_custom_proxy_params_uses_explicit_endpoint(monkeypatch):
+    monkeypatch.setenv(
+        "CUSTOM_PROXY_BASE_URL",
+        "https://proxy.test/api/v1/",
+    )
+    monkeypatch.setenv("CUSTOM_PROXY_API_KEY", "proxy-secret")
+
+    params = _resolve_llm_params("custom-proxy/Azure AI/gpt-5.5")
+
+    assert params["model"] == "openai/Azure AI/gpt-5.5"
+    assert params["api_base"] == "https://proxy.test/api/v1"
+    assert params["api_key"] == "proxy-secret"
+
+
+def test_resolve_custom_proxy_requires_base_url(monkeypatch):
+    monkeypatch.delenv("CUSTOM_PROXY_BASE_URL", raising=False)
+    monkeypatch.delenv("LOCAL_LLM_BASE_URL", raising=False)
+
+    with pytest.raises(ValueError, match="CUSTOM_PROXY_BASE_URL"):
+        _resolve_llm_params("custom-proxy/Azure AI/gpt-5.5")
+
+
 def test_local_params_reject_reasoning_effort_in_strict_mode():
     with pytest.raises(UnsupportedEffortError, match="reasoning_effort"):
         _resolve_llm_params("ollama/llama3.1", reasoning_effort="high", strict=True)