Add a dedicated OpenAI-compatible LLM adapter

README.md

@@ -176,9 +176,10 @@ Also see [architecture](docs/ARCHITECTURE.md).
 | Provider | Status | Provider | Status |
 |----------|--------|----------|--------|
 | OpenAI | ✅ | Azure OpenAI | ✅ |
-| Anthropic Claude | ✅ | Google Gemini | ✅ |
-| AWS Bedrock | ✅ | Mistral AI | ✅ |
-| Ollama (local) | ✅ | Anyscale | ✅ |
+| OpenAI Compatible | ✅ | Anthropic Claude | ✅ |
+| AWS Bedrock | ✅ | Google Gemini | ✅ |
+| Ollama (local) | ✅ | Mistral AI | ✅ |
+| Anyscale | ✅ | | |
 
 ### Vector Databases
 

unstract/sdk1/src/unstract/sdk1/adapters/base1.py

@@ -4,7 +4,7 @@
 import os
 from abc import ABC, abstractmethod
 from importlib import import_module
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, ClassVar
 
 if TYPE_CHECKING:
     from typing import Any
@@ -76,6 +76,8 @@ def register_adapters(adapters: dict[str, dict[str, "Any"]], adapter_type: str)
 class BaseAdapter(ABC):
     """Adapter base class for compatibility with all SDK v1 providers."""
 
+    SCHEMA_PATH: ClassVar[str | None] = None
+
     @staticmethod
     @abstractmethod
     def get_id() -> str:
@@ -103,11 +105,13 @@ def get_icon() -> str:
 
     @classmethod
     def get_json_schema(cls) -> str:
-        schema_path = (
-            f"{os.path.dirname(__file__)}/"
-            f"{cls.get_adapter_type().name.lower()}1/static/"
-            f"{cls.get_provider()}.json"
-        )
+        schema_path = cls.SCHEMA_PATH
+        if schema_path is None:
+            schema_path = (
+                f"{os.path.dirname(__file__)}/"
+                f"{cls.get_adapter_type().name.lower()}1/static/"
+                f"{cls.get_provider()}.json"
+            )
         with open(schema_path) as f:
             return f.read()
 
@@ -225,6 +229,30 @@ def validate_model(adapter_metadata: dict[str, "Any"]) -> str:
             return f"openai/{model}"
 
 
+class OpenAICompatibleLLMParameters(BaseChatCompletionParameters):
+    """See https://docs.litellm.ai/docs/providers/openai_compatible/."""
+
+    api_key: str | None = None
+    api_base: str
+
+    @staticmethod
+    def validate(adapter_metadata: dict[str, "Any"]) -> dict[str, "Any"]:
+        adapter_metadata["model"] = OpenAICompatibleLLMParameters.validate_model(
+            adapter_metadata
+        )
+        api_key = adapter_metadata.get("api_key")
+        if isinstance(api_key, str) and not api_key.strip():
+            adapter_metadata["api_key"] = None
+        return OpenAICompatibleLLMParameters(**adapter_metadata).model_dump()
+
+    @staticmethod
+    def validate_model(adapter_metadata: dict[str, "Any"]) -> str:
+        model = adapter_metadata.get("model", "")
+        if model.startswith("custom_openai/"):
+            return model
+        return f"custom_openai/{model}"
+
+
 class AzureOpenAILLMParameters(BaseChatCompletionParameters):
     """See https://docs.litellm.ai/docs/providers/azure/#completion---using-azure_ad_token-api_base-api_version."""
 

unstract/sdk1/src/unstract/sdk1/adapters/llm1/__init__.py

@@ -8,6 +8,7 @@
 from unstract.sdk1.adapters.llm1.bedrock import AWSBedrockLLMAdapter
 from unstract.sdk1.adapters.llm1.ollama import OllamaLLMAdapter
 from unstract.sdk1.adapters.llm1.openai import OpenAILLMAdapter
+from unstract.sdk1.adapters.llm1.openai_compatible import OpenAICompatibleLLMAdapter
 from unstract.sdk1.adapters.llm1.vertexai import VertexAILLMAdapter
 
 adapters: dict[str, dict[str, Any]] = {}
@@ -22,5 +23,6 @@
     "AzureOpenAILLMAdapter",
     "OllamaLLMAdapter",
     "OpenAILLMAdapter",
+    "OpenAICompatibleLLMAdapter",
     "VertexAILLMAdapter",
 ]

unstract/sdk1/src/unstract/sdk1/adapters/llm1/openai_compatible.py

@@ -0,0 +1,51 @@
+import os
+from typing import Any, ClassVar
+
+from unstract.sdk1.adapters.base1 import BaseAdapter, OpenAICompatibleLLMParameters
+from unstract.sdk1.adapters.enums import AdapterTypes
+
+DESCRIPTION = (
+    "Adapter for servers that implement the OpenAI Chat Completions API "
+    "(vLLM, LM Studio, self-hosted gateways, and third-party providers). "
+    "Use OpenAI for the official OpenAI service."
+)
+
+
+class OpenAICompatibleLLMAdapter(OpenAICompatibleLLMParameters, BaseAdapter):
+    SCHEMA_PATH: ClassVar[str] = os.path.join(
+        os.path.dirname(__file__), "static", "openai_compatible.json"
+    )
+
+    @staticmethod
+    def get_id() -> str:
+        return "openaicompatible|b6d10f33-2c41-49fc-a8c2-58d2b247fc09"
+
+    @staticmethod
+    def get_metadata() -> dict[str, Any]:
+        return {
+            "name": "OpenAI Compatible",
+            "version": "1.0.0",
+            "adapter": OpenAICompatibleLLMAdapter,
+            "description": DESCRIPTION,
+            "is_active": True,
+        }
+
+    @staticmethod
+    def get_name() -> str:
+        return "OpenAI Compatible"
+
+    @staticmethod
+    def get_description() -> str:
+        return DESCRIPTION
+
+    @staticmethod
+    def get_provider() -> str:
+        return "custom_openai"
+
+    @staticmethod
+    def get_icon() -> str:
+        return "/icons/adapter-icons/OpenAICompatible.png"
+
+    @staticmethod
+    def get_adapter_type() -> AdapterTypes:
+        return AdapterTypes.LLM

unstract/sdk1/src/unstract/sdk1/adapters/llm1/static/openai_compatible.json

@@ -0,0 +1,60 @@
+{
+  "title": "OpenAI Compatible",
+  "type": "object",
+  "required": [
+    "adapter_name",
+    "api_base"
+  ],
+  "properties": {
+    "adapter_name": {
+      "type": "string",
+      "title": "Name",
+      "default": "",
+      "description": "Provide a unique name for this adapter instance. Example: compatible-gateway-1"
+    },
+    "api_key": {
+      "type": [
+        "string",
+        "null"
+      ],
+      "title": "API Key",
+      "format": "password",
+      "description": "API key for your OpenAI-compatible endpoint. Leave empty if the endpoint does not require one."
+    },
+    "model": {
+      "type": "string",
+      "title": "Model",
+      "description": "The model name expected by your OpenAI-compatible endpoint. Examples: gateway-model, gpt-4o-mini, openai/gpt-4o"
+    },
+    "api_base": {
+      "type": "string",
+      "format": "url",
+      "title": "API Base",
+      "description": "Base URL for the OpenAI-compatible endpoint. Examples: https://gateway.example.com/v1, https://llm.example.net/openai/v1"
+    },
+    "max_tokens": {
+      "type": "number",
+      "minimum": 0,
+      "multipleOf": 1,
+      "title": "Maximum Output Tokens",
+      "default": 4096,
+      "description": "Maximum number of output tokens to limit LLM replies. Leave it empty to use the provider default."
+    },
+    "max_retries": {
+      "type": "number",
+      "minimum": 0,
+      "multipleOf": 1,
+      "title": "Max Retries",
+      "default": 5,
+      "description": "The maximum number of times to retry a request if it fails."
+    },
+    "timeout": {
+      "type": "number",
+      "minimum": 0,
+      "multipleOf": 1,
+      "title": "Timeout",
+      "default": 900,
+      "description": "Timeout in seconds."
+    }
+  }
+}

unstract/sdk1/src/unstract/sdk1/llm.py

@@ -652,10 +652,22 @@ def _record_usage(
         usage: Mapping[str, int] | None,
         llm_api: str,
     ) -> None:
-        prompt_tokens = token_counter(model=model, messages=messages)
         usage_data: Mapping[str, int] = usage or {}
+        prompt_tokens = usage_data.get("prompt_tokens")
+        if prompt_tokens is None:
+            try:
+                prompt_tokens = token_counter(model=model, messages=messages)
+            except Exception as e:
+                prompt_tokens = 0
+                logger.warning(
+                    "[sdk1][LLM][%s][%s] Failed to estimate prompt tokens; "
+                    "recording 0 prompt tokens for usage audit: %s",
+                    model,
+                    llm_api,
+                    e,
+                )
         all_tokens = TokenCounterCompat(
-            prompt_tokens=usage_data.get("prompt_tokens", 0),
+            prompt_tokens=prompt_tokens or 0,
             completion_tokens=usage_data.get("completion_tokens", 0),
             total_tokens=usage_data.get("total_tokens", 0),
         )