aient 1.1.15__py3-none-any.whl → 1.1.17__py3-none-any.whl

aient/core/request.py

@@ -4,7 +4,7 @@ import httpx
 import base64
 import urllib.parse
 
-from .models import RequestModel
+from .models import RequestModel, Message
 from .utils import (
     c3s,
     c3o,
@@ -50,7 +50,12 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
     systemInstruction = None
     system_prompt = ""
     function_arguments = None
-    for msg in request.messages:
+
+    try:
+        request_messages = [Message(role="user", content=request.prompt)]
+    except:
+        request_messages = request.messages
+    for msg in request_messages:
         if msg.role == "assistant":
             msg.role = "model"
         tool_calls = None
@@ -104,9 +109,10 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
         elif msg.role == "system":
             content[0]["text"] = re.sub(r"_+", "_", content[0]["text"])
             system_prompt = system_prompt + "\n\n" + content[0]["text"]
-    systemInstruction = {"parts": [{"text": system_prompt}]}
+    if system_prompt.strip():
+        systemInstruction = {"parts": [{"text": system_prompt}]}
 
-    if any(off_model in original_model for off_model in gemini_max_token_65k_models):
+    if any(off_model in original_model for off_model in gemini_max_token_65k_models) or original_model == "gemini-2.0-flash-preview-image-generation":
         safety_settings = "OFF"
     else:
         safety_settings = "BLOCK_NONE"
@@ -160,6 +166,7 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
         'top_logprobs',
         'response_format',
         'stream_options',
+        'prompt',
     ]
     generation_config = {}
 
@@ -214,6 +221,12 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
         else:
             payload["generationConfig"]["maxOutputTokens"] = 8192
 
+        if original_model == "gemini-2.0-flash-preview-image-generation":
+            payload["generationConfig"]["response_modalities"] = [
+                "Text",
+                "Image",
+            ]
+
     if "gemini-2.5" in original_model:
         payload["generationConfig"]["thinkingConfig"] = {
             "includeThoughts": True,
@@ -241,7 +254,7 @@ async def get_gemini_payload(request, engine, provider, api_key=None):
             if key == request.model:
                 for k, v in value.items():
                     payload[k] = v
-            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude"]):
+            elif all(_model not in request.model.lower() for _model in ["gemini", "gpt", "claude", "deepseek"]) and "-" not in key:
                 payload[key] = value
 
     return url, headers, payload

aient/core/response.py

@@ -36,6 +36,7 @@ async def fetch_gemini_response_stream(client, url, headers, payload, model):
         candidatesTokenCount = 0
         totalTokenCount = 0
         parts_json = ""
+        image_base64 = ""
         # line_index = 0
         # last_text_line = 0
         # if "thinking" in model:
@@ -53,6 +54,8 @@ async def fetch_gemini_response_stream(client, url, headers, payload, model):
                 if line and '\"finishReason\": \"' in line:
                     if "stop" not in line.lower():
                         logger.error(f"finishReason: {line}")
+                    sse_string = await generate_sse_response(timestamp, model, stop="stop")
+                    yield sse_string
                     is_finish = True
                 if is_finish and '\"promptTokenCount\": ' in line:
                     json_data = parse_json_safely( "{" + line + "}")
@@ -67,17 +70,25 @@ async def fetch_gemini_response_stream(client, url, headers, payload, model):
                 if (line and '"parts": [' in line or parts_json != "") and is_finish == False:
                     parts_json += line
                 if parts_json != "" and line and '],' == line.strip():
-                    tmp_parts_json =  "{" + parts_json.split("}        ]      },")[0].strip().rstrip("}], ").replace("\n", "\\n").lstrip("{") + "}]}"
+                    # tmp_parts_json =  "{" + parts_json.split("}        ]      },")[0].strip().rstrip("}], ").replace("\n", "\\n").lstrip("{") + "}]}"
+                    tmp_parts_json =  "{" + parts_json.split("}        ]      },")[0].strip().rstrip("}], ").replace("\n", "\\n").lstrip("{")
+                    if "inlineData" in tmp_parts_json:
+                        tmp_parts_json = tmp_parts_json + "}}]}"
+                    else:
+                        tmp_parts_json = tmp_parts_json + "}]}"
                     try:
                         json_data = json.loads(tmp_parts_json)
 
                         content = safe_get(json_data, "parts", 0, "text", default="")
+                        b64_json = safe_get(json_data, "parts", 0, "inlineData", "data", default="")
+                        if b64_json:
+                            image_base64 = b64_json
 
                         is_thinking = safe_get(json_data, "parts", 0, "thought", default=False)
                         if is_thinking:
                             sse_string = await generate_sse_response(timestamp, model, reasoning_content=content)
                             yield sse_string
-                        else:
+                        elif not image_base64:
                             sse_string = await generate_sse_response(timestamp, model, content=content)
                             yield sse_string
                     except json.JSONDecodeError:
@@ -93,6 +104,10 @@ async def fetch_gemini_response_stream(client, url, headers, payload, model):
 
                     function_full_response += line
 
+        if image_base64:
+            yield await generate_no_stream_response(timestamp, model, content=content, tools_id=None, function_call_name=None, function_call_content=None, role=None, total_tokens=totalTokenCount, prompt_tokens=promptTokenCount, completion_tokens=candidatesTokenCount, image_base64=image_base64)
+            return
+
         if need_function_call:
             function_call = json.loads(function_full_response)
             function_call_name = function_call["functionCall"]["name"]
@@ -535,9 +550,13 @@ async def fetch_response(client, url, headers, payload, engine, model):
         # print("parsed_data", json.dumps(parsed_data, indent=4, ensure_ascii=False))
         content = ""
         reasoning_content = ""
+        image_base64 = ""
         parts_list = safe_get(parsed_data, 0, "candidates", 0, "content", "parts", default=[])
         for item in parts_list:
             chunk = safe_get(item, "text")
+            b64_json = safe_get(item, "inlineData", "data", default="")
+            if b64_json:
+                image_base64 = b64_json
             is_think = safe_get(item, "thought", default=False)
             # logger.info(f"chunk: {repr(chunk)}")
             if chunk:
@@ -571,7 +590,7 @@ async def fetch_response(client, url, headers, payload, engine, model):
         function_call_content = safe_get(parsed_data, -1, "candidates", 0, "content", "parts", 0, "functionCall", "args", default=None)
 
         timestamp = int(datetime.timestamp(datetime.now()))
-        yield await generate_no_stream_response(timestamp, model, content=content, tools_id=None, function_call_name=function_call_name, function_call_content=function_call_content, role=role, total_tokens=total_tokens, prompt_tokens=prompt_tokens, completion_tokens=candidates_tokens, reasoning_content=reasoning_content)
+        yield await generate_no_stream_response(timestamp, model, content=content, tools_id=None, function_call_name=function_call_name, function_call_content=function_call_content, role=role, total_tokens=total_tokens, prompt_tokens=prompt_tokens, completion_tokens=candidates_tokens, reasoning_content=reasoning_content, image_base64=image_base64)
 
     elif engine == "claude":
         response_json = response.json()

aient/core/utils.py

@@ -112,7 +112,7 @@ def get_engine(provider, endpoint=None, original_model=""):
     if provider.get("engine"):
         engine = provider["engine"]
 
-    if endpoint == "/v1/images/generations" or "stable-diffusion" in original_model:
+    if engine != "gemini" and (endpoint == "/v1/images/generations" or "stable-diffusion" in original_model):
         engine = "dalle"
         stream = False
 
@@ -449,7 +449,7 @@ end_of_line = "\n\n"
 
 import random
 import string
-async def generate_sse_response(timestamp, model, content=None, tools_id=None, function_call_name=None, function_call_content=None, role=None, total_tokens=0, prompt_tokens=0, completion_tokens=0, reasoning_content=None):
+async def generate_sse_response(timestamp, model, content=None, tools_id=None, function_call_name=None, function_call_content=None, role=None, total_tokens=0, prompt_tokens=0, completion_tokens=0, reasoning_content=None, stop=None):
     random.seed(timestamp)
     random_str = ''.join(random.choices(string.ascii_letters + string.digits, k=29))
 
@@ -467,7 +467,7 @@ async def generate_sse_response(timestamp, model, content=None, tools_id=None, f
                 "index": 0,
                 "delta": delta_content,
                 "logprobs": None,
-                "finish_reason": None if content else "stop"
+                "finish_reason": None if content or reasoning_content else "stop"
             }
         ],
         "usage": None,
@@ -484,14 +484,19 @@ async def generate_sse_response(timestamp, model, content=None, tools_id=None, f
         total_tokens = prompt_tokens + completion_tokens
         sample_data["usage"] = {"prompt_tokens": prompt_tokens, "completion_tokens": completion_tokens, "total_tokens": total_tokens}
         sample_data["choices"] = []
+    if stop:
+        sample_data["choices"][0]["delta"] = {}
+        sample_data["choices"][0]["finish_reason"] = stop
+
     json_data = json.dumps(sample_data, ensure_ascii=False)
+    # print("json_data", json.dumps(sample_data, indent=4, ensure_ascii=False))
 
     # 构建SSE响应
     sse_response = f"data: {json_data}" + end_of_line
 
     return sse_response
 
-async def generate_no_stream_response(timestamp, model, content=None, tools_id=None, function_call_name=None, function_call_content=None, role=None, total_tokens=0, prompt_tokens=0, completion_tokens=0, reasoning_content=None):
+async def generate_no_stream_response(timestamp, model, content=None, tools_id=None, function_call_name=None, function_call_content=None, role=None, total_tokens=0, prompt_tokens=0, completion_tokens=0, reasoning_content=None, image_base64=None):
     random.seed(timestamp)
     random_str = ''.join(random.choices(string.ascii_letters + string.digits, k=29))
     message = {
@@ -554,11 +559,25 @@ async def generate_no_stream_response(timestamp, model, content=None, tools_id=N
             "system_fingerprint": "fp_4691090a87"
         }
 
+    if image_base64:
+        sample_data = {
+            "created": timestamp,
+            "data": [{
+                "b64_json": image_base64
+            }],
+            # "usage": {
+            #     "total_tokens": 100,
+            #     "input_tokens": 50,
+            #     "output_tokens": 50,
+            # }
+        }
+
     if total_tokens:
         total_tokens = prompt_tokens + completion_tokens
         sample_data["usage"] = {"prompt_tokens": prompt_tokens, "completion_tokens": completion_tokens, "total_tokens": total_tokens}
 
     json_data = json.dumps(sample_data, ensure_ascii=False)
+    # print("json_data", json.dumps(sample_data, indent=4, ensure_ascii=False))
 
     return json_data
 

aient/prompt/agent.py

@@ -44,6 +44,7 @@ system_prompt = """
 2. Always respond in 中文。
 3. 尽力满足user的请求，如果 user 要求你使用工具，请自行根据工具的参数要求，组织参数，将工具调用组织成xml格式，即可触发工具执行流程。如果user提供了明确的xml工具调用，请直接复述user的xml工具调用。你必须复述user的xml工具调用才能真正调用工具。
 4. 禁止要求user调用工具，当你需要调用工具时，请自行组织参数，将工具调用组织成xml格式，即可触发工具执行流程。禁止自己没有使用xml格式调用工具就假定工具已经调用。
+5. 仔细阅读并理解 user 的任务描述和当前指令。user 会根据任务描述分步骤让你执行，因此请严格遵循 user 指令按步骤操作，并请勿自行完成 user 未指定的步骤。在 user 明确给出下一步指令前，绝不擅自行动或推测、执行任何未明确要求的后续步骤。完成当前 user 指定的步骤后，必须主动向 user 汇报该步骤的完成情况，并明确请求 user 提供下一步的操作指令。
 </communication>
 
 <making_code_changes>

{aient-1.1.15.dist-info → aient-1.1.17.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aient
-Version: 1.1.15
+Version: 1.1.17
 Summary: Aient: The Awakening of Agent.
 Description-Content-Type: text/markdown
 License-File: LICENSE

{aient-1.1.15.dist-info → aient-1.1.17.dist-info}/RECORD

@@ -4,9 +4,9 @@ aient/core/.gitignore,sha256=5JRRlYYsqt_yt6iFvvzhbqh2FTUQMqwo6WwIuFzlGR8,13
 aient/core/__init__.py,sha256=NxjebTlku35S4Dzr16rdSqSTWUvvwEeACe8KvHJnjPg,34
 aient/core/log_config.py,sha256=kz2_yJv1p-o3lUQOwA3qh-LSc3wMHv13iCQclw44W9c,274
 aient/core/models.py,sha256=kF-HLi1I2k_G5r153ZHuiGH8_NmpTlFMfK0_myB28YQ,7366
-aient/core/request.py,sha256=2u_vPJjoIqRsKOPOkPilGzM4iZXUpsx5xtouQy87Z4E,65693
-aient/core/response.py,sha256=Z0Bjl_QvpUguyky1LIcsVks4BKKqT0eYEpDmKa_cwpQ,31978
-aient/core/utils.py,sha256=-naFCv8V-qhnqvDUd8BNbW1HR9CVAPxISrXoAz464Qg,26580
+aient/core/request.py,sha256=nvF_V71svezQ0-UbnC9RB_pXo_wV6QC7WE_SANwQzxE,66195
+aient/core/response.py,sha256=3OASBI-Y_8VlAusY-ch3oPwBHhn1rMDhhMyDBsK1FcY,33224
+aient/core/utils.py,sha256=VQ9uutGRR_JOvECOrjeoRBO2aA6w-pGwoXnnS2UvfPU,27263
 aient/core/test/test_base_api.py,sha256=pWnycRJbuPSXKKU9AQjWrMAX1wiLC_014Qc9hh5C2Pw,524
 aient/core/test/test_geminimask.py,sha256=HFX8jDbNg_FjjgPNxfYaR-0-roUrOO-ND-FVsuxSoiw,13254
 aient/core/test/test_image.py,sha256=_T4peNGdXKBHHxyQNx12u-NTyFE8TlYI6NvvagsG2LE,319
@@ -34,12 +34,12 @@ aient/plugins/run_python.py,sha256=dgcUwBunMuDkaSKR5bToudVzSdrXVewktDDFUz_iIOQ,4
 aient/plugins/websearch.py,sha256=llxy1U0vJiNMiKvamMr4p7IruLb3nnDR4YErz8TYimc,15215
 aient/plugins/write_file.py,sha256=YRvQKMvV-5lwohxlvwt9hjfxz2dRJP85AJWAMUIqbBY,3804
 aient/prompt/__init__.py,sha256=GBtn6-JDT8KHFCcuPpfSNE_aGddg5p4FEyMCy4BfwGs,20
-aient/prompt/agent.py,sha256=y2GETN6ScC5yQVs75VFfzm4YUWzblbqLYz0Sy6JnPRw,24950
+aient/prompt/agent.py,sha256=ZNsbgXRyvYzAFTRRziAnNVqcTyAnxrGcsGfGrt72j6k,25427
 aient/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 aient/utils/prompt.py,sha256=UcSzKkFE4-h_1b6NofI6xgk3GoleqALRKY8VBaXLjmI,11311
 aient/utils/scripts.py,sha256=LD8adnfuRrJoY2tWKseXOPJXaxbrUmz4czsnUvHswNY,29096
-aient-1.1.15.dist-info/licenses/LICENSE,sha256=XNdbcWldt0yaNXXWB_Bakoqnxb3OVhUft4MgMA_71ds,1051
-aient-1.1.15.dist-info/METADATA,sha256=MTwSePn_DF0VVK3i0k9yW8E6Y2dOEn76NwsF7BGx_E0,4968
-aient-1.1.15.dist-info/WHEEL,sha256=zaaOINJESkSfm_4HQVc5ssNzHCPXhJm0kEUakpsEHaU,91
-aient-1.1.15.dist-info/top_level.txt,sha256=3oXzrP5sAVvyyqabpeq8A2_vfMtY554r4bVE-OHBrZk,6
-aient-1.1.15.dist-info/RECORD,,
+aient-1.1.17.dist-info/licenses/LICENSE,sha256=XNdbcWldt0yaNXXWB_Bakoqnxb3OVhUft4MgMA_71ds,1051
+aient-1.1.17.dist-info/METADATA,sha256=4rxyF7eATHamHyS_kV-HyEtyw1vtn5BE5cKMVsgDHKg,4968
+aient-1.1.17.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+aient-1.1.17.dist-info/top_level.txt,sha256=3oXzrP5sAVvyyqabpeq8A2_vfMtY554r4bVE-OHBrZk,6
+aient-1.1.17.dist-info/RECORD,,

{aient-1.1.15.dist-info → aient-1.1.17.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.8.0)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any