Update main.py

main.py

@@ -79,16 +79,28 @@ async def demo():
     return HTMLResponse(content=html_content, status_code=200)
 
 
+@app.get("/v1")
+async def flow(prompt = ""):
+    completion = llm(prompt)
+    async def server_sent_events(chat_chunks):
+        yield prompt
+        for chat_chunk in chat_chunks:
+            yield chat_chunk
+        yield ""
+
+    return EventSourceResponse(server_sent_events(completion))
+
+
 @app.get("/stream")
 async def chat(prompt = "Once upon a time there was a "):
-    tokens = llm(prompt)
-    async def server_sent_events(chat_chunks):
+    tokens = llm.tokenize(prompt)
+    async def server_sent_events(chat_chunks, llm):
         yield prompt
         for chat_chunk in llm.generate(chat_chunks):
             yield llm.detokenize(chat_chunk)
         yield ""
 
-    return EventSourceResponse(server_sent_events(tokens))
+    return EventSourceResponse(server_sent_events(tokens, llm))
 
 @app.post("/v1/chat/completions")
 async def chat(request: ChatCompletionRequest, response_mode=None):