| 1 | # Copyright (C) 2025 AIDC-AI |
| 2 | # |
| 3 | # Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | # you may not use this file except in compliance with the License. |
| 5 | # You may obtain a copy of the License at |
| 6 | # http://www.apache.org/licenses/LICENSE-2.0 |
| 7 | # Unless required by applicable law or agreed to in writing, software |
| 8 | # distributed under the License is distributed on an "AS IS" BASIS, |
| 9 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 10 | # See the License for the specific language governing permissions and |
| 11 | # limitations under the License. |
| 12 | |
| 13 | """ |
| 14 | LLM utility functions for model discovery and connection testing. |
| 15 | |
| 16 | Uses the OpenAI-compatible models endpoint. |
| 17 | """ |
| 18 | |
| 19 | import re |
| 20 | from typing import List, Tuple |
| 21 | import httpx |
| 22 | from loguru import logger |
| 23 | |
| 24 | |
| 25 | def _build_models_url(base_url: str) -> str: |
| 26 | """Build a provider models endpoint from a user-entered API base URL.""" |
| 27 | raw = (base_url or "").strip().rstrip("/") |
| 28 | if raw.endswith("/models"): |
| 29 | return raw |
| 30 | |
| 31 | normalized = normalize_openai_base_url(base_url) |
| 32 | |
| 33 | if re.search(r"/v\d+(?:\.\d+)?$", normalized): |
| 34 | return f"{normalized}/models" |
| 35 | |
| 36 | return f"{normalized}/v1/models" |
| 37 | |
| 38 | |
| 39 | def normalize_openai_base_url(base_url: str) -> str: |
| 40 | """Normalize a user-entered OpenAI-compatible Base URL for SDK calls. |
| 41 | |
| 42 | Users sometimes paste a concrete endpoint such as /chat/completions or |
| 43 | /models. The OpenAI SDK expects the API root, so concrete endpoint suffixes |
| 44 | must be stripped before real model calls. |
| 45 | """ |
| 46 | normalized = (base_url or "").strip().rstrip("/") |
| 47 | for suffix in ("/chat/completions", "/completions", "/responses", "/models"): |
| 48 | if normalized.endswith(suffix): |
| 49 | normalized = normalized[: -len(suffix)].rstrip("/") |
| 50 | break |
| 51 | return normalized |
| 52 | |
| 53 | |
| 54 | def fetch_available_models(api_key: str, base_url: str, timeout: float = 10.0) -> List[str]: |
| 55 | """ |
| 56 | Fetch available models from an OpenAI-compatible API endpoint. |
| 57 | |
| 58 | Uses the provider models endpoint with Bearer token authentication. |
| 59 | |
| 60 | Args: |
| 61 | api_key: The API key for authentication |
| 62 | base_url: The base URL of the API (e.g., https://api.openai.com/v1). |
| 63 | If a chat endpoint is pasted by mistake, it will be normalized. |
| 64 | timeout: Request timeout in seconds |
| 65 | |
| 66 | Returns: |
| 67 | List of model IDs available from the API |
| 68 | |
| 69 | Raises: |
| 70 | httpx.HTTPStatusError: If the API returns an error status code |
| 71 | httpx.RequestError: If there's a network error |
| 72 | """ |
| 73 | models_url = _build_models_url(base_url) |
| 74 | |
| 75 | headers = { |
| 76 | "Authorization": f"Bearer {api_key}", |
| 77 | "Content-Type": "application/json", |
| 78 | } |
| 79 | |
| 80 | logger.debug(f"Fetching models from: {models_url}") |
| 81 | |
| 82 | with httpx.Client(timeout=timeout) as client: |
| 83 | response = client.get(models_url, headers=headers) |
| 84 | response.raise_for_status() |
| 85 | |
| 86 | data = response.json() |
| 87 | models = [model["id"] for model in data.get("data", [])] |
| 88 | |
| 89 | # Sort models alphabetically for better UX |
| 90 | models.sort() |
| 91 | |
| 92 | logger.debug(f"Fetched {len(models)} models") |
| 93 | return models |
| 94 | |
| 95 | |
| 96 | def test_llm_connection(api_key: str, base_url: str, timeout: float = 10.0) -> Tuple[bool, str, int]: |
| 97 | """ |
| 98 | Test the LLM API connection by attempting to fetch the models list. |
| 99 | |
| 100 | Args: |
| 101 | api_key: The API key for authentication |
| 102 | base_url: The base URL of the API |
| 103 | timeout: Request timeout in seconds |
| 104 | |
| 105 | Returns: |
| 106 | Tuple of (success: bool, message: str, model_count: int) |
| 107 | - success: True if connection succeeded |
| 108 | - message: Human-readable status message |
| 109 | - model_count: Number of models available (0 if failed) |
| 110 | """ |
| 111 | try: |
| 112 | models = fetch_available_models(api_key, base_url, timeout) |
| 113 | return True, f"Connection successful! {len(models)} models available.", len(models) |
| 114 | except httpx.HTTPStatusError as e: |
| 115 | status_code = e.response.status_code |
| 116 | if status_code == 401: |
| 117 | return False, "Authentication failed: Invalid API Key", 0 |
| 118 | elif status_code == 403: |
| 119 | return False, "Access forbidden: Check your API Key permissions", 0 |
| 120 | elif status_code == 404: |
| 121 | return False, "API endpoint not found: Check your Base URL", 0 |
| 122 | else: |
| 123 | return False, f"API error: HTTP {status_code}", 0 |
| 124 | except httpx.ConnectError: |
| 125 | return False, "Connection failed: Cannot reach the server", 0 |
| 126 | except httpx.TimeoutException: |
| 127 | return False, "Connection timeout: Server did not respond in time", 0 |
| 128 | except Exception as e: |
| 129 | logger.error(f"LLM connection test error: {e}") |
| 130 | return False, f"Error: {str(e)}", 0 |
| 131 |