From 2aa3185f2e73a928094d7dbc90c15d50279c33c3 Mon Sep 17 00:00:00 2001
From: yumoqing <yumoqing@gmail.com>
Date: Mon, 30 Mar 2026 16:21:38 +0800
Subject: [PATCH] bugfix

---
 llmage/asyncinference.py | 13 ++++++++-----
 llmage/llmclient.py      |  4 +++-
 llmage/syncinference.py  |  7 ++++---
 3 files changed, 15 insertions(+), 9 deletions(-)

diff --git a/llmage/asyncinference.py b/llmage/asyncinference.py
index e5844df..5c4f51f 100644
--- a/llmage/asyncinference.py
+++ b/llmage/asyncinference.py
@@ -78,9 +78,6 @@ async def async_uapi_request(request, llm, sor,
 			b = b.decode('utf-8')
 		debug(f'task submited:{b}')
 		d = DictObject(**json.loads(b))
-		if d.status == 'FAILED':
-			e = Exception(f'resp={d} FFAILED')
-			raise e
 		responsed_seconds = time.time() - start_timestamp
 		finish_seconds = responsed_seconds
 		llmusage = DictObject()
@@ -90,22 +87,28 @@ async def async_uapi_request(request, llm, sor,
 		llmusage.use_time = timestampstr()
 		llmusage.userid = callerid
 		llmusage.ioinfo = json.dumps({
-			"input": params_kw
+			"input": params_kw,
+			'output': [d]
 		})
 		llmusage.taskid = d.taskid
 		llmusage.transno = params_kw.transno
 		llmusage.responsed_seconds = responsed_seconds
 		llmusage.finish_seconds = finish_seconds
-		llmusage.status = 'CREATED'
+		llmusage.status = d.status
 		llmusage.userorgid = callerorgid
 		llmusage.ownerid = llm.orgid
+		llmusage.accounting_status = 'created'
 		b = json.dumps(d, ensure_ascii=False)
 		yield b
 		await write_llmusage(llmusage)
 		# if llm.callbackurl:
 		#	return
+		if d.status == 'FAILED':
+			e = Exception(f'resp={d} FFAILED')
+			raise e
 		asyncio.create_task(query_task_status(request, llm.upappid, 
 								llm.query_apiname, luid, userid, d.taskid))
+		yield d
 
 	except Exception as e:
 		exception(f'{e=},{format_exc()}')
diff --git a/llmage/llmclient.py b/llmage/llmclient.py
index 0a4f445..6be5ff4 100644
--- a/llmage/llmclient.py
+++ b/llmage/llmclient.py
@@ -105,6 +105,7 @@ async def uapi_request(request, llm, sor, callerid, callerorgid, params_kw=None)
 			except Exception as e:
 				e = Exception(f'{llm.pid} charging error{e}')
 				exception(f'{e}')
+			llmusage.amount = llmusage.cost = 0
 		else:
 			llmusage.amount = 0
 			llmusage.cost = 0
@@ -112,7 +113,8 @@ async def uapi_request(request, llm, sor, callerid, callerorgid, params_kw=None)
 		llmusage.ownerid = llm.orgid
 		llmusage.accounting_status = 'created'
 		await write_llmusage(llmusage)
-		await llm_accounting(request, llmusage)
+		if llmusage.amount > 0.0001:
+			await llm_accounting(request, llmusage)
 
 	except Exception as e:
 		exception(f'{e=},{format_exc()}')
diff --git a/llmage/syncinference.py b/llmage/syncinference.py
index 119f96d..83ed9d1 100644
--- a/llmage/syncinference.py
+++ b/llmage/syncinference.py
@@ -37,7 +37,7 @@ async def sync_uapi_request(request, llm, sor, callerid, callerorgid, params_kw=
 			b = b.decode('utf-8')
 		d = json.loads(b)
 		status = d.get('status')
-		usage = d.get('usage', {})
+		usage = d.get('usage')
 		if status and status != 'SUCCEEDED':
 			raise Exception(d['error'])
 		responsed_seconds = time.time() - start_timestamp
@@ -51,7 +51,7 @@ async def sync_uapi_request(request, llm, sor, callerid, callerorgid, params_kw=
 		llmusage.usage = json.dumps(usage)
 		llmusage.ioinfo = json.dumps({
 			"input": params_kw,
-			"output": d
+			"output": [d]
 		})
 		llmusage.transno = params_kw.transno
 		llmusage.responsed_seconds = responsed_seconds
@@ -77,7 +77,8 @@ async def sync_uapi_request(request, llm, sor, callerid, callerorgid, params_kw=
 		b = json.dumps(d, ensure_ascii=False)
 		yield b
 		await write_llmusage(llmusage)
-		await llm_accounting(request, llmusage)
+		if llmusage.amount > 0.0001:
+			await llm_accounting(request, llmusage)
 	except Exception as e:
 		exception(f'{e=},{format_exc()}')
 		estr = erase_apikey(e)