initial commit

.gitignore

@@ -0,0 +1,2 @@
+/.venv
+__pycache__

README.md

@@ -0,0 +1,7 @@
+# llm-ui
+
+```shell
+$ uv venv
+$ uv pip install -r requirements.txt
+$ uv run uvicorn server:app
+```

server/__init__.py

@@ -0,0 +1,11 @@
+from server.http import Starlette, Route, Request, Response, JSONResponse, WebSocketRoute
+from server.inference import list_conversations, connect_to_conversation
+
+async def status(request: Request) -> Response:
+  return JSONResponse({"status": "ok"})
+
+app = Starlette(debug=True, routes=[
+  Route("/api/", status),
+  Route("/api/conversation", list_conversations, methods=["GET"]),
+  WebSocketRoute("/api/conversation/:conversation/connect", connect_to_conversation)
+])

server/http.py

@@ -0,0 +1,5 @@
+from starlette.applications import Starlette
+from starlette.routing import Route, WebSocketRoute
+from starlette.responses import *
+from starlette.requests import *
+from starlette.websockets import WebSocket

server/inference.py

@@ -0,0 +1,33 @@
+import llm, llm.cli, sqlite_utils
+from .http import Request, JSONResponse, WebSocket, RedirectResponse
+import json
+
+db = sqlite_utils.Database(llm.cli.logs_db_path())
+
+async def list_conversations(request: Request):
+  conversations = []
+  for row in db["conversations"].rows:
+    conversations.append({ "id": row["id"], "name": row["name"] })
+  return JSONResponse(conversations)
+
+girlypop_prompt = llm.cli.load_template("girlypop").system
+
+async def connect_to_conversation(ws: WebSocket):
+  conversation_id = ws.path_params["conversation"]
+  if conversation_id == "new":
+    conversation = llm.AsyncConversation(llm.get_async_model())
+  else:
+    try:
+      conversation: llm.AsyncConversation = llm.cli.load_conversation(conversation_id, async_=True)
+    except:
+      await ws.send_denial_response(JSONResponse({
+        "error": "unable to load conversation {}".format(conversation_id)
+      }))
+      return
+
+  await ws.accept()
+  async for message in ws.iter_text():
+    response = conversation.prompt(message, system=girlypop_prompt)
+    async for chunk in response:
+      ws.send_text(json.dumps({"c": chunk}))
+    ws.send_text(json.dumps({"d": True})) # done