Chat glm with fast api

[hamletzhang]

增加了Fastapi接口 post /stream 可以得到服务端结果

[ypwhs]

相关 Issue：#17

[yuri2peter]

多谢，已经用上。按照我自己的习惯稍作修改了些：

    @app.post("/stream")
    def continue_question_stream(arg_dict: dict):

        def decorate(generator):
            for item in generator:
                yield ServerSentEvent(
                    json.dumps(item, ensure_ascii=False), event='delta')

        # inputs = [query, answer_prefix, max_length, top_p, temperature, history]
        try:
            query = arg_dict["query"]
            answer_prefix = arg_dict.get("answer_prefix", "")
            max_length = arg_dict.get("max_length", 256)
            top_p = arg_dict.get("top_p", 0.7)
            temperature = arg_dict.get("temperature", 1.0)
            history = arg_dict.get("history", [])
            allow_generate[0] = True
            logger.info("Query - {}".format(query))
            if answer_prefix:
                logger.info(f"answer_prefix - {answer_prefix}")
            history = history[-MAX_HISTORY:]
            if len(history) > 0:
                logger.info("History - {}".format(history))
            history = [tuple(h) for h in history]
            inputs = [
                query, answer_prefix, max_length, top_p, temperature,
                allow_generate, history
            ]
            return EventSourceResponse(decorate(bot.predict_continue(*inputs)))
            # return EventSourceResponse(bot.predict_continue(*inputs))
        except Exception as e:
            logger.error(f"error: {e}")
            return EventSourceResponse(
                decorate(bot.predict_continue(None, None)))
        
    
    @app.post("/interrupt")
    def interrupt():
        allow_generate[0] = False
        logger.error("Interrupted.")
        return {"message": "OK"}