Hermes Agent commited on
Commit ·
4e5e577
1
Parent(s): 18e0d8b
Add append-only chat.log + /logs endpoint
Browse files
app.py
CHANGED
|
@@ -16,6 +16,7 @@ if STATIC_DIR.exists():
|
|
| 16 |
DATA_DIR = Path("/data")
|
| 17 |
LLAMA_DIR = Path("/home/user/llama.cpp")
|
| 18 |
MODELS_DIR = DATA_DIR / "models"
|
|
|
|
| 19 |
LLAMA_PORT = 8080
|
| 20 |
LLAMA_URL = f"http://127.0.0.1:{LLAMA_PORT}"
|
| 21 |
|
|
@@ -155,6 +156,16 @@ async def chat_completions(request: Request):
|
|
| 155 |
if not server_running():
|
| 156 |
raise HTTPException(status_code=503, detail="llama-server not running. POST /start first.")
|
| 157 |
body = await request.json()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 158 |
# inject default max_tokens if missing
|
| 159 |
if "max_tokens" not in body and "max_completion_tokens" not in body:
|
| 160 |
body["max_tokens"] = 512
|
|
@@ -165,10 +176,35 @@ async def chat_completions(request: Request):
|
|
| 165 |
async with httpx.AsyncClient(timeout=300.0) as client:
|
| 166 |
try:
|
| 167 |
response = await client.post(f"{LLAMA_URL}/v1/chat/completions", json=body)
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
except httpx.ConnectError:
|
| 170 |
raise HTTPException(status_code=503, detail="llama-server not reachable")
|
| 171 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 172 |
@app.get("/v1/models")
|
| 173 |
async def list_models():
|
| 174 |
models = []
|
|
|
|
| 16 |
DATA_DIR = Path("/data")
|
| 17 |
LLAMA_DIR = Path("/home/user/llama.cpp")
|
| 18 |
MODELS_DIR = DATA_DIR / "models"
|
| 19 |
+
CHAT_LOG = DATA_DIR / "chat.log"
|
| 20 |
LLAMA_PORT = 8080
|
| 21 |
LLAMA_URL = f"http://127.0.0.1:{LLAMA_PORT}"
|
| 22 |
|
|
|
|
| 156 |
if not server_running():
|
| 157 |
raise HTTPException(status_code=503, detail="llama-server not running. POST /start first.")
|
| 158 |
body = await request.json()
|
| 159 |
+
# Log user messages
|
| 160 |
+
try:
|
| 161 |
+
with open(CHAT_LOG, "a", encoding="utf-8") as lf:
|
| 162 |
+
ts = time.strftime("%Y-%m-%d %H:%M:%S")
|
| 163 |
+
for msg in body.get("messages", []):
|
| 164 |
+
if msg.get("role") == "user":
|
| 165 |
+
lf.write(f"[{ts}] USER: {msg.get('content','')}\n")
|
| 166 |
+
except Exception:
|
| 167 |
+
pass
|
| 168 |
+
|
| 169 |
# inject default max_tokens if missing
|
| 170 |
if "max_tokens" not in body and "max_completion_tokens" not in body:
|
| 171 |
body["max_tokens"] = 512
|
|
|
|
| 176 |
async with httpx.AsyncClient(timeout=300.0) as client:
|
| 177 |
try:
|
| 178 |
response = await client.post(f"{LLAMA_URL}/v1/chat/completions", json=body)
|
| 179 |
+
content = response.content
|
| 180 |
+
# Log assistant response
|
| 181 |
+
try:
|
| 182 |
+
with open(CHAT_LOG, "a", encoding="utf-8") as lf:
|
| 183 |
+
ts = time.strftime("%Y-%m-%d %H:%M:%S")
|
| 184 |
+
resp_json = json.loads(content)
|
| 185 |
+
for choice in resp_json.get("choices", []):
|
| 186 |
+
msg = choice.get("message", {})
|
| 187 |
+
if msg.get("content"):
|
| 188 |
+
lf.write(f"[{ts}] ASSISTANT: {msg['content']}\n")
|
| 189 |
+
elif msg.get("tool_calls"):
|
| 190 |
+
lf.write(f"[{ts}] ASSISTANT-TOOL: {json.dumps(msg['tool_calls'], ensure_ascii=False)}\n")
|
| 191 |
+
except Exception:
|
| 192 |
+
pass
|
| 193 |
+
return Response(content=content, status_code=response.status_code, headers=dict(response.headers))
|
| 194 |
except httpx.ConnectError:
|
| 195 |
raise HTTPException(status_code=503, detail="llama-server not reachable")
|
| 196 |
|
| 197 |
+
@app.get("/logs")
|
| 198 |
+
def get_logs(lines: int = 100):
|
| 199 |
+
try:
|
| 200 |
+
if not CHAT_LOG.exists():
|
| 201 |
+
return {"lines": []}
|
| 202 |
+
with open(CHAT_LOG, "r", encoding="utf-8") as f:
|
| 203 |
+
all_lines = f.readlines()
|
| 204 |
+
return {"lines": [line.rstrip("\n") for line in all_lines[-lines:]]}
|
| 205 |
+
except Exception as e:
|
| 206 |
+
raise HTTPException(status_code=500, detail=str(e))
|
| 207 |
+
|
| 208 |
@app.get("/v1/models")
|
| 209 |
async def list_models():
|
| 210 |
models = []
|