from datetime import datetime from django.shortcuts import render from app.views.ViewsBase import * from app.models import LLMModel from app.utils.Utils import buildPageLabels from app.utils.LLMUtils import LLMUtils from app.utils.Credentials import decrypt_credential, encrypt_credential, mask_credential, redact_mapping def index(request): params = f_parseGetParams(request) page = params.get('p', 1) page_size = params.get('ps', 10) try: page = int(page) if page < 1: page = 1 except Exception: page = 1 try: page_size = int(page_size) if page_size < 1: page_size = 10 elif page_size > 100: page_size = 100 except Exception: page_size = 10 skip = (page - 1) * page_size count_row = g_database.select("select count(id) as count from av_llm") count = int(count_row[0]["count"]) if count_row else 0 data = [] if count > 0: data = g_database.select( "select * from av_llm order by id desc limit %s,%s", [skip, page_size]) for d in data: d["api_key_masked"] = mask_credential(d.pop("api_key", "")) d["has_api_key"] = bool(d["api_key_masked"]) if d.get("last_update_time"): d["last_update_time"] = d["last_update_time"].strftime("%Y/%m/%d %H:%M") page_num = count // page_size if count % page_size > 0: page_num += 1 page_labels = buildPageLabels(page=page, page_num=page_num, lang=f_parseRequestLang(request)) page_data = { "page": page, "page_size": page_size, "page_num": page_num, "count": count, "pageLabels": page_labels, } return render(request, 'app/llm/index.html', {"data": data, "pageData": page_data}) def test(request): return render(request, 'app/llm/test.html', {}) def api_openIndex(request): ret = False msg = LANG_VIEWS_T(request, "msg_unknown_error") data = [] if request.method == "GET": __check_ret, __check_msg = f_checkRequestSafe(request) if __check_ret: try: for d in LLMModel.objects.all().order_by('-id'): data.append({ 'id': d.id, 'code': d.code, 'name': d.name, 'model_name': d.model_name, 'api_url': d.api_url, 'api_key_masked': mask_credential(d.api_key), 'has_api_key': bool(d.api_key), 'timeout': d.timeout, 'inference_tool': d.inference_tool, 'state': d.state, 'remark': d.remark, 'last_update_time': d.last_update_time.strftime("%Y/%m/%d %H:%M") if d.last_update_time else '', }) ret = True msg = LANG_VIEWS_T(request, "msg_success") except Exception as e: msg = str(e) else: msg = __check_msg else: msg = LANG_VIEWS_T(request, "msg_method_not_supported") return f_responseJson({"code": 1000 if ret else 0, "msg": msg, "data": data}) def api_openAdd(request): ret = False msg = LANG_VIEWS_T(request, "msg_unknown_error") if request.method == "POST": __check_ret, __check_msg = f_checkRequestSafe(request) if __check_ret: params = f_parsePostParams(request) g_logger.info("LLMView.openAdd() params:%s" % str(redact_mapping(params))) code = params.get("code", "").strip() name = (params.get("name") or "").strip() model_name = params.get("model_name", "").strip() if not name: name = model_name api_url = params.get("api_url", "").strip() api_key = params.get("api_key", "").strip() timeout = int(params.get("timeout", 30)) inference_tool = params.get("inference_tool", "OpenAI").strip() or "OpenAI" if inference_tool != "OpenAI": raise Exception(LANG_VIEWS_T(request, "llm_inference_tool_unsupported")) state = int(params.get("state", 1)) remark = params.get("remark", "").strip() try: if LLMModel.objects.filter(code=code).first(): raise Exception(LANG_VIEWS_T(request, "msg_code_already_exists")) if api_url and not (api_url.startswith("http://") or api_url.startswith("https://")): raise Exception(LANG_VIEWS_T(request, "llm_api_url_format_error")) if not model_name: raise Exception(LANG_VIEWS_T(request, "llm_input_model_name")) if not api_url: raise Exception(LANG_VIEWS_T(request, "llm_input_api_url")) if not api_key: raise Exception(LANG_VIEWS_T(request, "llm_input_api_key")) llm = LLMModel() llm.user_id = f_sessionReadUserId(request) llm.code = code llm.name = name llm.model_name = model_name llm.api_url = api_url llm.api_key = encrypt_credential(api_key) llm.timeout = timeout llm.inference_tool = inference_tool llm.remark = remark llm.sort = 0 llm.create_time = datetime.now() llm.last_update_time = datetime.now() llm.state = state llm.save() ret = True msg = LANG_VIEWS_T(request, "msg_add_success") except Exception as e: msg = str(e) else: msg = __check_msg else: msg = LANG_VIEWS_T(request, "msg_method_not_supported") return f_responseJson({"code": 1000 if ret else 0, "msg": msg}) def api_openEdit(request): ret = False msg = LANG_VIEWS_T(request, "msg_unknown_error") if request.method == "POST": __check_ret, __check_msg = f_checkRequestSafe(request) if __check_ret: params = f_parsePostParams(request) g_logger.info("LLMView.openEdit() params:%s" % str(redact_mapping(params))) code = params.get("code", "").strip() name = (params.get("name") or "").strip() model_name = params.get("model_name", "").strip() if not name: name = model_name api_url = params.get("api_url", "").strip() api_key = params.get("api_key", "").strip() timeout = int(params.get("timeout", 30)) inference_tool = params.get("inference_tool", "OpenAI").strip() or "OpenAI" if inference_tool != "OpenAI": raise Exception(LANG_VIEWS_T(request, "llm_inference_tool_unsupported")) state = int(params.get("state", 1)) remark = params.get("remark", "").strip() try: if api_url and not (api_url.startswith("http://") or api_url.startswith("https://")): raise Exception(LANG_VIEWS_T(request, "llm_api_url_format_error")) llm = LLMModel.objects.filter(code=code).first() if not llm: raise Exception(LANG_VIEWS_T(request, "msg_data_not_exist")) llm.name = name llm.model_name = model_name llm.api_url = api_url if api_key: llm.api_key = encrypt_credential(api_key) llm.timeout = timeout llm.inference_tool = inference_tool llm.remark = remark llm.last_update_time = datetime.now() llm.state = state llm.save() ret = True msg = LANG_VIEWS_T(request, "msg_edit_success") except Exception as e: msg = str(e) else: msg = __check_msg else: msg = LANG_VIEWS_T(request, "msg_method_not_supported") return f_responseJson({"code": 1000 if ret else 0, "msg": msg}) def api_openInfo(request): ret = False msg = LANG_VIEWS_T(request, "msg_unknown_error") info = {} if request.method == "GET": __check_ret, __check_msg = f_checkRequestSafe(request) if __check_ret: code = f_parseGetParams(request).get("code", "").strip() if not code: msg = LANG_VIEWS_T(request, "msg_invalid_parameter") else: try: llm = LLMModel.objects.filter(code=code).first() if llm: info = { "id": llm.id, "code": llm.code, "name": llm.name, "model_name": llm.model_name, "api_url": llm.api_url or "", "api_key_masked": mask_credential(llm.api_key), "has_api_key": bool(llm.api_key), "timeout": llm.timeout, "inference_tool": llm.inference_tool, "state": llm.state, "remark": llm.remark or "", "create_time": llm.create_time.strftime("%Y-%m-%d %H:%M:%S") if llm.create_time else "", "last_update_time": llm.last_update_time.strftime("%Y-%m-%d %H:%M:%S") if llm.last_update_time else "", } ret = True msg = LANG_VIEWS_T(request, "msg_success") else: msg = LANG_VIEWS_T(request, "llm_config_not_exist") except Exception as e: msg = str(e) else: msg = __check_msg else: msg = LANG_VIEWS_T(request, "msg_method_not_supported") return f_responseJson({"code": 1000 if ret else 0, "msg": msg, "info": info}) def api_openDel(request): ret = False msg = LANG_VIEWS_T(request, "msg_unknown_error") if request.method == "POST": __check_ret, __check_msg = f_checkRequestSafe(request) if __check_ret: code = f_parsePostParams(request).get("code") llm = LLMModel.objects.filter(code=code).first() if llm: if llm.delete(): ret = True msg = LANG_VIEWS_T(request, "msg_success") else: msg = LANG_VIEWS_T(request, "msg_failed_to_delete") else: msg = LANG_VIEWS_T(request, "msg_data_not_exist") else: msg = __check_msg else: msg = LANG_VIEWS_T(request, "msg_method_not_supported") return f_responseJson({"code": 1000 if ret else 0, "msg": msg}) def api_openTest(request): try: if request.method != "POST": raise Exception(LANG_VIEWS_T(request, "msg_method_not_supported")) __check_ret, __check_msg = f_checkRequestSafe(request) if not __check_ret: raise Exception(__check_msg) code = request.POST.get('code', '').strip() prompt = request.POST.get('prompt', '请详细描述图片中的内容？').strip() if code: llm = LLMModel.objects.filter(code=code).first() if not llm: raise Exception(LANG_VIEWS_T(request, "llm_not_found")) if llm.state != 1: raise Exception(LANG_VIEWS_T(request, "llm_is_disabled")) api_url = llm.api_url api_key = decrypt_credential(llm.api_key) timeout = llm.timeout inference_tool = "OpenAI" model = llm.model_name else: api_url = request.POST.get('api_url', 'https://api.openai.com/v1').strip() api_key = request.POST.get('api_key', '').strip() timeout = int(request.POST.get('timeout', 30)) inference_tool = request.POST.get('inferenceTool', 'OpenAI').strip() or 'OpenAI' if inference_tool != 'OpenAI': raise Exception(LANG_VIEWS_T(request, "llm_inference_tool_unsupported")) model = request.POST.get('model', 'gpt-4o').strip() file_content_base64 = request.POST.get('file_content', '').strip() if file_content_base64: image_bytes = base64.b64decode(file_content_base64) elif 'image' in request.FILES: image_bytes = request.FILES['image'].read() else: raise Exception(LANG_VIEWS_T(request, "llm_no_image_provided")) result = LLMUtils(api_url, api_key, timeout, inference_tool, model).infer(prompt, image_bytes) return f_responseJson({"code": 1000, "result": result}) except Exception as e: return f_responseJson({"code": 0, "msg": str(e)})