From 47d205f676f1cba17fb2027523961d5a7ee0b2ad Mon Sep 17 00:00:00 2001 From: Fangyin Cheng Date: Mon, 24 Jun 2024 19:07:10 +0800 Subject: [PATCH] feat(model): Support database model registry (#1656) --- assets/schema/dbgpt.sql | 23 + .../{v_0_5_6 => v0_5_6}/upgrade_to_v0.5.6.sql | 0 .../upgrade/{v_0_5_6 => v0_5_6}/v0.5.5.sql | 0 .../{v_0_5_7 => v0_5_7}/upgrade_to_v0.5.7.sql | 0 .../upgrade/{v_0_5_7 => v0_5_7}/v0.5.6.sql | 0 .../upgrade/v0_5_9/upgrade_to_v0.5.9.sql | 22 + assets/schema/upgrade/v0_5_9/v0.5.8.sql | 396 +++++++++++ .../initialization/db_model_initialization.py | 5 + dbgpt/model/cli.py | 1 + .../model/cluster/apiserver/tests/test_api.py | 30 +- dbgpt/model/cluster/controller/controller.py | 131 +++- .../model/cluster/registry_impl}/__init__.py | 0 .../model/cluster/registry_impl/db_storage.py | 116 +++ dbgpt/model/cluster/registry_impl/storage.py | 374 ++++++++++ .../cluster/tests/registry_impl/__init__.py | 0 .../registry_impl/test_storage_registry.py | 221 ++++++ dbgpt/model/cluster/worker/manager.py | 6 +- dbgpt/model/parameter.py | 78 ++ dbgpt/serve/core/tests/conftest.py | 6 +- dbgpt/serve/prompt/tests/test_endpoints.py | 1 + dbgpt/util/api_utils.py | 176 ++++- dbgpt/util/tests/test_api_utils.py | 105 +++ dbgpt/util/utils.py | 2 +- docker/base/Dockerfile | 19 +- docker/base/build_proxy_image.sh | 17 + .../ha-cluster-docker-compose.yml | 127 ++++ docs/docs/api/datasource.md | 2 +- docs/docs/awel/get_started.ipynb | 28 +- .../installation/model_service/cluster.md | 2 +- .../installation/model_service/cluster_ha.md | 171 +++++ docs/sidebars.js | 4 + .../agent_auto_plan_dialogue_example.ipynb | 670 +----------------- .../agent_awel_layout_dialogue_example.ipynb | 69 +- requirements/dev-requirements.txt | 2 +- setup.py | 2 +- 35 files changed, 2014 insertions(+), 792 deletions(-) rename assets/schema/upgrade/{v_0_5_6 => v0_5_6}/upgrade_to_v0.5.6.sql (100%) rename assets/schema/upgrade/{v_0_5_6 => v0_5_6}/v0.5.5.sql (100%) rename assets/schema/upgrade/{v_0_5_7 => v0_5_7}/upgrade_to_v0.5.7.sql (100%) rename assets/schema/upgrade/{v_0_5_7 => v0_5_7}/v0.5.6.sql (100%) create mode 100644 assets/schema/upgrade/v0_5_9/upgrade_to_v0.5.9.sql create mode 100644 assets/schema/upgrade/v0_5_9/v0.5.8.sql rename {tools => dbgpt/model/cluster/registry_impl}/__init__.py (100%) create mode 100644 dbgpt/model/cluster/registry_impl/db_storage.py create mode 100644 dbgpt/model/cluster/registry_impl/storage.py create mode 100644 dbgpt/model/cluster/tests/registry_impl/__init__.py create mode 100644 dbgpt/model/cluster/tests/registry_impl/test_storage_registry.py create mode 100644 dbgpt/util/tests/test_api_utils.py create mode 100755 docker/base/build_proxy_image.sh create mode 100644 docker/compose_examples/ha-cluster-docker-compose.yml create mode 100644 docs/docs/installation/model_service/cluster_ha.md diff --git a/assets/schema/dbgpt.sql b/assets/schema/dbgpt.sql index 86f7ed740..237091b1e 100644 --- a/assets/schema/dbgpt.sql +++ b/assets/schema/dbgpt.sql @@ -282,6 +282,7 @@ CREATE TABLE `dbgpt_serve_flow` ( `source` varchar(64) DEFAULT NULL COMMENT 'Flow source', `source_url` varchar(512) DEFAULT NULL COMMENT 'Flow source url', `version` varchar(32) DEFAULT NULL COMMENT 'Flow version', + `define_type` varchar(32) null comment 'Flow define type(json or python)', `label` varchar(128) DEFAULT NULL COMMENT 'Flow label', `editable` int DEFAULT NULL COMMENT 'Editable, 0: editable, 1: not editable', PRIMARY KEY (`id`), @@ -340,6 +341,28 @@ CREATE TABLE `gpts_app_detail` ( UNIQUE KEY `uk_gpts_app_agent_node` (`app_name`,`agent_name`,`node_id`) ) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci; + +-- For deploy model cluster of DB-GPT(StorageModelRegistry) +CREATE TABLE IF NOT EXISTS `dbgpt_cluster_registry_instance` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'Auto increment id', + `model_name` varchar(128) NOT NULL COMMENT 'Model name', + `host` varchar(128) NOT NULL COMMENT 'Host of the model', + `port` int(11) NOT NULL COMMENT 'Port of the model', + `weight` float DEFAULT 1.0 COMMENT 'Weight of the model', + `check_healthy` tinyint(1) DEFAULT 1 COMMENT 'Whether to check the health of the model', + `healthy` tinyint(1) DEFAULT 0 COMMENT 'Whether the model is healthy', + `enabled` tinyint(1) DEFAULT 1 COMMENT 'Whether the model is enabled', + `prompt_template` varchar(128) DEFAULT NULL COMMENT 'Prompt template for the model instance', + `last_heartbeat` datetime DEFAULT NULL COMMENT 'Last heartbeat time of the model instance', + `user_name` varchar(128) DEFAULT NULL COMMENT 'User name', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` datetime DEFAULT CURRENT_TIMESTAMP COMMENT 'Record creation time', + `gmt_modified` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'Record update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_model_instance` (`model_name`, `host`, `port`, `sys_code`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='Cluster model instance table, for registering and managing model instances'; + + CREATE DATABASE IF NOT EXISTS EXAMPLE_1; use EXAMPLE_1; diff --git a/assets/schema/upgrade/v_0_5_6/upgrade_to_v0.5.6.sql b/assets/schema/upgrade/v0_5_6/upgrade_to_v0.5.6.sql similarity index 100% rename from assets/schema/upgrade/v_0_5_6/upgrade_to_v0.5.6.sql rename to assets/schema/upgrade/v0_5_6/upgrade_to_v0.5.6.sql diff --git a/assets/schema/upgrade/v_0_5_6/v0.5.5.sql b/assets/schema/upgrade/v0_5_6/v0.5.5.sql similarity index 100% rename from assets/schema/upgrade/v_0_5_6/v0.5.5.sql rename to assets/schema/upgrade/v0_5_6/v0.5.5.sql diff --git a/assets/schema/upgrade/v_0_5_7/upgrade_to_v0.5.7.sql b/assets/schema/upgrade/v0_5_7/upgrade_to_v0.5.7.sql similarity index 100% rename from assets/schema/upgrade/v_0_5_7/upgrade_to_v0.5.7.sql rename to assets/schema/upgrade/v0_5_7/upgrade_to_v0.5.7.sql diff --git a/assets/schema/upgrade/v_0_5_7/v0.5.6.sql b/assets/schema/upgrade/v0_5_7/v0.5.6.sql similarity index 100% rename from assets/schema/upgrade/v_0_5_7/v0.5.6.sql rename to assets/schema/upgrade/v0_5_7/v0.5.6.sql diff --git a/assets/schema/upgrade/v0_5_9/upgrade_to_v0.5.9.sql b/assets/schema/upgrade/v0_5_9/upgrade_to_v0.5.9.sql new file mode 100644 index 000000000..8ee337bf8 --- /dev/null +++ b/assets/schema/upgrade/v0_5_9/upgrade_to_v0.5.9.sql @@ -0,0 +1,22 @@ +USE dbgpt; + +-- For deploy model cluster of DB-GPT(StorageModelRegistry) +CREATE TABLE IF NOT EXISTS `dbgpt_cluster_registry_instance` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'Auto increment id', + `model_name` varchar(128) NOT NULL COMMENT 'Model name', + `host` varchar(128) NOT NULL COMMENT 'Host of the model', + `port` int(11) NOT NULL COMMENT 'Port of the model', + `weight` float DEFAULT 1.0 COMMENT 'Weight of the model', + `check_healthy` tinyint(1) DEFAULT 1 COMMENT 'Whether to check the health of the model', + `healthy` tinyint(1) DEFAULT 0 COMMENT 'Whether the model is healthy', + `enabled` tinyint(1) DEFAULT 1 COMMENT 'Whether the model is enabled', + `prompt_template` varchar(128) DEFAULT NULL COMMENT 'Prompt template for the model instance', + `last_heartbeat` datetime DEFAULT NULL COMMENT 'Last heartbeat time of the model instance', + `user_name` varchar(128) DEFAULT NULL COMMENT 'User name', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` datetime DEFAULT CURRENT_TIMESTAMP COMMENT 'Record creation time', + `gmt_modified` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'Record update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_model_instance` (`model_name`, `host`, `port`, `sys_code`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='Cluster model instance table, for registering and managing model instances'; + diff --git a/assets/schema/upgrade/v0_5_9/v0.5.8.sql b/assets/schema/upgrade/v0_5_9/v0.5.8.sql new file mode 100644 index 000000000..d810d2369 --- /dev/null +++ b/assets/schema/upgrade/v0_5_9/v0.5.8.sql @@ -0,0 +1,396 @@ +-- Full SQL of v0.5.8, please not modify this file(It must be same as the file in the release package) + +CREATE +DATABASE IF NOT EXISTS dbgpt; +use dbgpt; + +-- For alembic migration tool +CREATE TABLE IF NOT EXISTS `alembic_version` +( + version_num VARCHAR(32) NOT NULL, + CONSTRAINT alembic_version_pkc PRIMARY KEY (version_num) +) DEFAULT CHARSET=utf8mb4 ; + +CREATE TABLE IF NOT EXISTS `knowledge_space` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'auto increment id', + `name` varchar(100) NOT NULL COMMENT 'knowledge space name', + `vector_type` varchar(50) NOT NULL COMMENT 'vector type', + `desc` varchar(500) NOT NULL COMMENT 'description', + `owner` varchar(100) DEFAULT NULL COMMENT 'owner', + `context` TEXT DEFAULT NULL COMMENT 'context argument', + `gmt_created` TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + PRIMARY KEY (`id`), + KEY `idx_name` (`name`) COMMENT 'index:idx_name' +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='knowledge space table'; + +CREATE TABLE IF NOT EXISTS `knowledge_document` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'auto increment id', + `doc_name` varchar(100) NOT NULL COMMENT 'document path name', + `doc_type` varchar(50) NOT NULL COMMENT 'doc type', + `space` varchar(50) NOT NULL COMMENT 'knowledge space', + `chunk_size` int NOT NULL COMMENT 'chunk size', + `last_sync` TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT 'last sync time', + `status` varchar(50) NOT NULL COMMENT 'status TODO,RUNNING,FAILED,FINISHED', + `content` LONGTEXT NOT NULL COMMENT 'knowledge embedding sync result', + `result` TEXT NULL COMMENT 'knowledge content', + `vector_ids` LONGTEXT NULL COMMENT 'vector_ids', + `summary` LONGTEXT NULL COMMENT 'knowledge summary', + `gmt_created` TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + PRIMARY KEY (`id`), + KEY `idx_doc_name` (`doc_name`) COMMENT 'index:idx_doc_name' +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='knowledge document table'; + +CREATE TABLE IF NOT EXISTS `document_chunk` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'auto increment id', + `doc_name` varchar(100) NOT NULL COMMENT 'document path name', + `doc_type` varchar(50) NOT NULL COMMENT 'doc type', + `document_id` int NOT NULL COMMENT 'document parent id', + `content` longtext NOT NULL COMMENT 'chunk content', + `meta_info` varchar(200) NOT NULL COMMENT 'metadata info', + `gmt_created` timestamp NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + PRIMARY KEY (`id`), + KEY `idx_document_id` (`document_id`) COMMENT 'index:document_id' +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='knowledge document chunk detail'; + + + +CREATE TABLE IF NOT EXISTS `connect_config` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `db_type` varchar(255) NOT NULL COMMENT 'db type', + `db_name` varchar(255) NOT NULL COMMENT 'db name', + `db_path` varchar(255) DEFAULT NULL COMMENT 'file db path', + `db_host` varchar(255) DEFAULT NULL COMMENT 'db connect host(not file db)', + `db_port` varchar(255) DEFAULT NULL COMMENT 'db cnnect port(not file db)', + `db_user` varchar(255) DEFAULT NULL COMMENT 'db user', + `db_pwd` varchar(255) DEFAULT NULL COMMENT 'db password', + `comment` text COMMENT 'db comment', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_db` (`db_name`), + KEY `idx_q_db_type` (`db_type`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT 'Connection confi'; + +CREATE TABLE IF NOT EXISTS `chat_history` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `conv_uid` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'Conversation record unique id', + `chat_mode` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'Conversation scene mode', + `summary` longtext COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'Conversation record summary', + `user_name` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'interlocutor', + `messages` text COLLATE utf8mb4_unicode_ci COMMENT 'Conversation details', + `message_ids` text COLLATE utf8mb4_unicode_ci COMMENT 'Message id list, split by comma', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` timestamp NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + UNIQUE KEY `conv_uid` (`conv_uid`), + PRIMARY KEY (`id`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT 'Chat history'; + +CREATE TABLE IF NOT EXISTS `chat_history_message` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `conv_uid` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'Conversation record unique id', + `index` int NOT NULL COMMENT 'Message index', + `round_index` int NOT NULL COMMENT 'Round of conversation', + `message_detail` text COLLATE utf8mb4_unicode_ci COMMENT 'Message details, json format', + `gmt_created` timestamp NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + UNIQUE KEY `message_uid_index` (`conv_uid`, `index`), + PRIMARY KEY (`id`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT 'Chat history message'; + +CREATE TABLE IF NOT EXISTS `chat_feed_back` +( + `id` bigint(20) NOT NULL AUTO_INCREMENT, + `conv_uid` varchar(128) DEFAULT NULL COMMENT 'Conversation ID', + `conv_index` int(4) DEFAULT NULL COMMENT 'Round of conversation', + `score` int(1) DEFAULT NULL COMMENT 'Score of user', + `ques_type` varchar(32) DEFAULT NULL COMMENT 'User question category', + `question` longtext DEFAULT NULL COMMENT 'User question', + `knowledge_space` varchar(128) DEFAULT NULL COMMENT 'Knowledge space name', + `messages` longtext DEFAULT NULL COMMENT 'The details of user feedback', + `user_name` varchar(128) DEFAULT NULL COMMENT 'User name', + `gmt_created` timestamp NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_conv` (`conv_uid`,`conv_index`), + KEY `idx_conv` (`conv_uid`,`conv_index`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='User feedback table'; + + +CREATE TABLE IF NOT EXISTS `my_plugin` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `tenant` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'user tenant', + `user_code` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'user code', + `user_name` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'user name', + `name` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'plugin name', + `file_name` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'plugin package file name', + `type` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin type', + `version` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin version', + `use_count` int DEFAULT NULL COMMENT 'plugin total use count', + `succ_count` int DEFAULT NULL COMMENT 'plugin total success count', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT 'plugin install time', + PRIMARY KEY (`id`), + UNIQUE KEY `name` (`name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT='User plugin table'; + +CREATE TABLE IF NOT EXISTS `plugin_hub` +( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `name` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'plugin name', + `description` varchar(255) COLLATE utf8mb4_unicode_ci NOT NULL COMMENT 'plugin description', + `author` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin author', + `email` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin author email', + `type` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin type', + `version` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin version', + `storage_channel` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin storage channel', + `storage_url` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin download url', + `download_param` varchar(255) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'plugin download param', + `gmt_created` TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT 'plugin upload time', + `installed` int DEFAULT NULL COMMENT 'plugin already installed count', + PRIMARY KEY (`id`), + UNIQUE KEY `name` (`name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT='Plugin Hub table'; + + +CREATE TABLE IF NOT EXISTS `prompt_manage` +( + `id` int(11) NOT NULL AUTO_INCREMENT, + `chat_scene` varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Chat scene', + `sub_chat_scene` varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Sub chat scene', + `prompt_type` varchar(100) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Prompt type: common or private', + `prompt_name` varchar(256) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'prompt name', + `content` longtext COLLATE utf8mb4_unicode_ci COMMENT 'Prompt content', + `input_variables` varchar(1024) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Prompt input variables(split by comma))', + `model` varchar(128) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Prompt model name(we can use different models for different prompt)', + `prompt_language` varchar(32) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Prompt language(eg:en, zh-cn)', + `prompt_format` varchar(32) COLLATE utf8mb4_unicode_ci DEFAULT 'f-string' COMMENT 'Prompt format(eg: f-string, jinja2)', + `prompt_desc` varchar(512) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'Prompt description', + `user_name` varchar(128) COLLATE utf8mb4_unicode_ci DEFAULT NULL COMMENT 'User name', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` timestamp NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'created time', + `gmt_modified` timestamp NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'update time', + PRIMARY KEY (`id`), + UNIQUE KEY `prompt_name_uiq` (`prompt_name`, `sys_code`, `prompt_language`, `model`), + KEY `gmt_created_idx` (`gmt_created`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT='Prompt management table'; + + CREATE TABLE IF NOT EXISTS `gpts_conversations` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `conv_id` varchar(255) NOT NULL COMMENT 'The unique id of the conversation record', + `user_goal` text NOT NULL COMMENT 'User''s goals content', + `gpts_name` varchar(255) NOT NULL COMMENT 'The gpts name', + `state` varchar(255) DEFAULT NULL COMMENT 'The gpts state', + `max_auto_reply_round` int(11) NOT NULL COMMENT 'max auto reply round', + `auto_reply_count` int(11) NOT NULL COMMENT 'auto reply count', + `user_code` varchar(255) DEFAULT NULL COMMENT 'user code', + `sys_code` varchar(255) DEFAULT NULL COMMENT 'system app ', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + `team_mode` varchar(255) NULL COMMENT 'agent team work mode', + + PRIMARY KEY (`id`), + UNIQUE KEY `uk_gpts_conversations` (`conv_id`), + KEY `idx_gpts_name` (`gpts_name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT="gpt conversations"; + +CREATE TABLE IF NOT EXISTS `gpts_instance` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `gpts_name` varchar(255) NOT NULL COMMENT 'Current AI assistant name', + `gpts_describe` varchar(2255) NOT NULL COMMENT 'Current AI assistant describe', + `resource_db` text COMMENT 'List of structured database names contained in the current gpts', + `resource_internet` text COMMENT 'Is it possible to retrieve information from the internet', + `resource_knowledge` text COMMENT 'List of unstructured database names contained in the current gpts', + `gpts_agents` varchar(1000) DEFAULT NULL COMMENT 'List of agents names contained in the current gpts', + `gpts_models` varchar(1000) DEFAULT NULL COMMENT 'List of llm model names contained in the current gpts', + `language` varchar(100) DEFAULT NULL COMMENT 'gpts language', + `user_code` varchar(255) NOT NULL COMMENT 'user code', + `sys_code` varchar(255) DEFAULT NULL COMMENT 'system app code', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + `team_mode` varchar(255) NOT NULL COMMENT 'Team work mode', + `is_sustainable` tinyint(1) NOT NULL COMMENT 'Applications for sustainable dialogue', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_gpts` (`gpts_name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT="gpts instance"; + +CREATE TABLE `gpts_messages` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `conv_id` varchar(255) NOT NULL COMMENT 'The unique id of the conversation record', + `sender` varchar(255) NOT NULL COMMENT 'Who speaking in the current conversation turn', + `receiver` varchar(255) NOT NULL COMMENT 'Who receive message in the current conversation turn', + `model_name` varchar(255) DEFAULT NULL COMMENT 'message generate model', + `rounds` int(11) NOT NULL COMMENT 'dialogue turns', + `content` text COMMENT 'Content of the speech', + `current_goal` text COMMENT 'The target corresponding to the current message', + `context` text COMMENT 'Current conversation context', + `review_info` text COMMENT 'Current conversation review info', + `action_report` text COMMENT 'Current conversation action report', + `role` varchar(255) DEFAULT NULL COMMENT 'The role of the current message content', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + PRIMARY KEY (`id`), + KEY `idx_q_messages` (`conv_id`,`rounds`,`sender`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT="gpts message"; + + +CREATE TABLE `gpts_plans` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `conv_id` varchar(255) NOT NULL COMMENT 'The unique id of the conversation record', + `sub_task_num` int(11) NOT NULL COMMENT 'Subtask number', + `sub_task_title` varchar(255) NOT NULL COMMENT 'subtask title', + `sub_task_content` text NOT NULL COMMENT 'subtask content', + `sub_task_agent` varchar(255) DEFAULT NULL COMMENT 'Available agents corresponding to subtasks', + `resource_name` varchar(255) DEFAULT NULL COMMENT 'resource name', + `rely` varchar(255) DEFAULT NULL COMMENT 'Subtask dependencies，like: 1,2,3', + `agent_model` varchar(255) DEFAULT NULL COMMENT 'LLM model used by subtask processing agents', + `retry_times` int(11) DEFAULT NULL COMMENT 'number of retries', + `max_retry_times` int(11) DEFAULT NULL COMMENT 'Maximum number of retries', + `state` varchar(255) DEFAULT NULL COMMENT 'subtask status', + `result` longtext COMMENT 'subtask result', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_sub_task` (`conv_id`,`sub_task_num`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT="gpt plan"; + +-- dbgpt.dbgpt_serve_flow definition +CREATE TABLE `dbgpt_serve_flow` ( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'Auto increment id', + `uid` varchar(128) NOT NULL COMMENT 'Unique id', + `dag_id` varchar(128) DEFAULT NULL COMMENT 'DAG id', + `name` varchar(128) DEFAULT NULL COMMENT 'Flow name', + `flow_data` text COMMENT 'Flow data, JSON format', + `user_name` varchar(128) DEFAULT NULL COMMENT 'User name', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` datetime DEFAULT NULL COMMENT 'Record creation time', + `gmt_modified` datetime DEFAULT NULL COMMENT 'Record update time', + `flow_category` varchar(64) DEFAULT NULL COMMENT 'Flow category', + `description` varchar(512) DEFAULT NULL COMMENT 'Flow description', + `state` varchar(32) DEFAULT NULL COMMENT 'Flow state', + `error_message` varchar(512) NULL comment 'Error message', + `source` varchar(64) DEFAULT NULL COMMENT 'Flow source', + `source_url` varchar(512) DEFAULT NULL COMMENT 'Flow source url', + `version` varchar(32) DEFAULT NULL COMMENT 'Flow version', + `define_type` varchar(32) null comment 'Flow define type(json or python)', + `label` varchar(128) DEFAULT NULL COMMENT 'Flow label', + `editable` int DEFAULT NULL COMMENT 'Editable, 0: editable, 1: not editable', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_uid` (`uid`), + KEY `ix_dbgpt_serve_flow_sys_code` (`sys_code`), + KEY `ix_dbgpt_serve_flow_uid` (`uid`), + KEY `ix_dbgpt_serve_flow_dag_id` (`dag_id`), + KEY `ix_dbgpt_serve_flow_user_name` (`user_name`), + KEY `ix_dbgpt_serve_flow_name` (`name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci; + +-- dbgpt.gpts_app definition +CREATE TABLE `gpts_app` ( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `app_code` varchar(255) NOT NULL COMMENT 'Current AI assistant code', + `app_name` varchar(255) NOT NULL COMMENT 'Current AI assistant name', + `app_describe` varchar(2255) NOT NULL COMMENT 'Current AI assistant describe', + `language` varchar(100) NOT NULL COMMENT 'gpts language', + `team_mode` varchar(255) NOT NULL COMMENT 'Team work mode', + `team_context` text COMMENT 'The execution logic and team member content that teams with different working modes rely on', + `user_code` varchar(255) DEFAULT NULL COMMENT 'user code', + `sys_code` varchar(255) DEFAULT NULL COMMENT 'system app code', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + `icon` varchar(1024) DEFAULT NULL COMMENT 'app icon, url', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_gpts_app` (`app_name`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci; + +CREATE TABLE `gpts_app_collection` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `app_code` varchar(255) NOT NULL COMMENT 'Current AI assistant code', + `user_code` int(11) NOT NULL COMMENT 'user code', + `sys_code` varchar(255) NOT NULL COMMENT 'system app code', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + PRIMARY KEY (`id`), + KEY `idx_app_code` (`app_code`), + KEY `idx_user_code` (`user_code`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT="gpt collections"; + +-- dbgpt.gpts_app_detail definition +CREATE TABLE `gpts_app_detail` ( + `id` int NOT NULL AUTO_INCREMENT COMMENT 'autoincrement id', + `app_code` varchar(255) NOT NULL COMMENT 'Current AI assistant code', + `app_name` varchar(255) NOT NULL COMMENT 'Current AI assistant name', + `agent_name` varchar(255) NOT NULL COMMENT ' Agent name', + `node_id` varchar(255) NOT NULL COMMENT 'Current AI assistant Agent Node id', + `resources` text COMMENT 'Agent bind resource', + `prompt_template` text COMMENT 'Agent bind template', + `llm_strategy` varchar(25) DEFAULT NULL COMMENT 'Agent use llm strategy', + `llm_strategy_value` text COMMENT 'Agent use llm strategy value', + `created_at` datetime DEFAULT NULL COMMENT 'create time', + `updated_at` datetime DEFAULT NULL COMMENT 'last update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_gpts_app_agent_node` (`app_name`,`agent_name`,`node_id`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci; + +CREATE +DATABASE IF NOT EXISTS EXAMPLE_1; +use EXAMPLE_1; +CREATE TABLE IF NOT EXISTS `users` +( + `id` int NOT NULL AUTO_INCREMENT, + `username` varchar(50) NOT NULL COMMENT '用户名', + `password` varchar(50) NOT NULL COMMENT '密码', + `email` varchar(50) NOT NULL COMMENT '邮箱', + `phone` varchar(20) DEFAULT NULL COMMENT '电话', + PRIMARY KEY (`id`), + KEY `idx_username` (`username`) COMMENT '索引：按用户名查询' +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='聊天用户表'; + +INSERT INTO users (username, password, email, phone) +VALUES ('user_1', 'password_1', 'user_1@example.com', '12345678901'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_2', 'password_2', 'user_2@example.com', '12345678902'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_3', 'password_3', 'user_3@example.com', '12345678903'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_4', 'password_4', 'user_4@example.com', '12345678904'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_5', 'password_5', 'user_5@example.com', '12345678905'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_6', 'password_6', 'user_6@example.com', '12345678906'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_7', 'password_7', 'user_7@example.com', '12345678907'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_8', 'password_8', 'user_8@example.com', '12345678908'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_9', 'password_9', 'user_9@example.com', '12345678909'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_10', 'password_10', 'user_10@example.com', '12345678900'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_11', 'password_11', 'user_11@example.com', '12345678901'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_12', 'password_12', 'user_12@example.com', '12345678902'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_13', 'password_13', 'user_13@example.com', '12345678903'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_14', 'password_14', 'user_14@example.com', '12345678904'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_15', 'password_15', 'user_15@example.com', '12345678905'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_16', 'password_16', 'user_16@example.com', '12345678906'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_17', 'password_17', 'user_17@example.com', '12345678907'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_18', 'password_18', 'user_18@example.com', '12345678908'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_19', 'password_19', 'user_19@example.com', '12345678909'); +INSERT INTO users (username, password, email, phone) +VALUES ('user_20', 'password_20', 'user_20@example.com', '12345678900'); \ No newline at end of file diff --git a/dbgpt/app/initialization/db_model_initialization.py b/dbgpt/app/initialization/db_model_initialization.py index 4d6353bcc..0749ccdf0 100644 --- a/dbgpt/app/initialization/db_model_initialization.py +++ b/dbgpt/app/initialization/db_model_initialization.py @@ -1,11 +1,14 @@ """Import all models to make sure they are registered with SQLAlchemy. """ + from dbgpt.app.knowledge.chunk_db import DocumentChunkEntity from dbgpt.app.knowledge.document_db import KnowledgeDocumentEntity from dbgpt.app.openapi.api_v1.feedback.feed_back_db import ChatFeedBackEntity from dbgpt.datasource.manages.connect_config_db import ConnectConfigEntity +from dbgpt.model.cluster.registry_impl.db_storage import ModelInstanceEntity from dbgpt.serve.agent.db.my_plugin_db import MyPluginEntity from dbgpt.serve.agent.db.plugin_hub_db import PluginHubEntity +from dbgpt.serve.flow.models.models import ServeEntity as FlowServeEntity from dbgpt.serve.prompt.models.models import ServeEntity as PromptManageEntity from dbgpt.serve.rag.models.models import KnowledgeSpaceEntity from dbgpt.storage.chat_history.chat_history_db import ( @@ -24,4 +27,6 @@ ConnectConfigEntity, ChatHistoryEntity, ChatHistoryMessageEntity, + ModelInstanceEntity, + FlowServeEntity, ] diff --git a/dbgpt/model/cli.py b/dbgpt/model/cli.py index 24a367fb1..90731ac37 100644 --- a/dbgpt/model/cli.py +++ b/dbgpt/model/cli.py @@ -26,6 +26,7 @@ build_lazy_click_command, ) +# Your can set environment variable CONTROLLER_ADDRESS to set the default address MODEL_CONTROLLER_ADDRESS = "http://127.0.0.1:8000" logger = logging.getLogger("dbgpt_cli") diff --git a/dbgpt/model/cluster/apiserver/tests/test_api.py b/dbgpt/model/cluster/apiserver/tests/test_api.py index d9001ed44..64d880962 100644 --- a/dbgpt/model/cluster/apiserver/tests/test_api.py +++ b/dbgpt/model/cluster/apiserver/tests/test_api.py @@ -1,22 +1,11 @@ -import importlib.metadata as metadata - import pytest import pytest_asyncio -from aioresponses import aioresponses -from fastapi import FastAPI from fastapi.middleware.cors import CORSMiddleware -from httpx import AsyncClient, HTTPError +from httpx import ASGITransport, AsyncClient, HTTPError from dbgpt.component import SystemApp from dbgpt.model.cluster.apiserver.api import ( - ChatCompletionResponse, - ChatCompletionResponseChoice, - ChatCompletionResponseStreamChoice, - ChatCompletionStreamResponse, - ChatMessage, - DeltaMessage, ModelList, - UsageInfo, api_settings, initialize_apiserver, ) @@ -56,12 +45,13 @@ async def client(request, system_app: SystemApp): if api_settings: # Clear global api keys api_settings.api_keys = [] - async with AsyncClient(app=app, base_url="http://test", headers=headers) as client: + async with AsyncClient( + transport=ASGITransport(app), base_url="http://test", headers=headers + ) as client: async with _new_cluster(**param) as cluster: worker_manager, model_registry = cluster system_app.register(_DefaultWorkerManagerFactory, worker_manager) system_app.register_instance(model_registry) - # print(f"Instances {model_registry.registry}") initialize_apiserver(None, app, system_app, api_keys=api_keys) yield client @@ -113,7 +103,11 @@ async def test_chat_completions(client: AsyncClient, expected_messages): "Hello world.", "abc", ), - ({"stream_messags": ["你好，我是", "张三。"], "api_keys": ["abc"]}, "你好，我是张三。", "abc"), + ( + {"stream_messags": ["你好，我是", "张三。"], "api_keys": ["abc"]}, + "你好，我是张三。", + "abc", + ), ], indirect=["client"], ) @@ -160,7 +154,11 @@ async def test_chat_completions_with_openai_lib_async_no_stream( "Hello world.", "abc", ), - ({"stream_messags": ["你好，我是", "张三。"], "api_keys": ["abc"]}, "你好，我是张三。", "abc"), + ( + {"stream_messags": ["你好，我是", "张三。"], "api_keys": ["abc"]}, + "你好，我是张三。", + "abc", + ), ], indirect=["client"], ) diff --git a/dbgpt/model/cluster/controller/controller.py b/dbgpt/model/cluster/controller/controller.py index 64311cd96..a28e4ccb7 100644 --- a/dbgpt/model/cluster/controller/controller.py +++ b/dbgpt/model/cluster/controller/controller.py @@ -1,6 +1,6 @@ import logging from abc import ABC, abstractmethod -from typing import List +from typing import List, Literal, Optional from fastapi import APIRouter @@ -8,6 +8,7 @@ from dbgpt.model.base import ModelInstance from dbgpt.model.cluster.registry import EmbeddedModelRegistry, ModelRegistry from dbgpt.model.parameter import ModelControllerParameters +from dbgpt.util.api_utils import APIMixin from dbgpt.util.api_utils import _api_remote as api_remote from dbgpt.util.api_utils import _sync_api_remote as sync_api_remote from dbgpt.util.fastapi import create_app @@ -46,9 +47,7 @@ async def model_apply(self) -> bool: class LocalModelController(BaseModelController): - def __init__(self, registry: ModelRegistry = None) -> None: - if not registry: - registry = EmbeddedModelRegistry() + def __init__(self, registry: ModelRegistry) -> None: self.registry = registry self.deployment = None @@ -75,9 +74,25 @@ async def send_heartbeat(self, instance: ModelInstance) -> bool: return await self.registry.send_heartbeat(instance) -class _RemoteModelController(BaseModelController): - def __init__(self, base_url: str) -> None: - self.base_url = base_url +class _RemoteModelController(APIMixin, BaseModelController): + def __init__( + self, + urls: str, + health_check_interval_secs: int = 5, + health_check_timeout_secs: int = 30, + check_health: bool = True, + choice_type: Literal["latest_first", "random"] = "latest_first", + ) -> None: + APIMixin.__init__( + self, + urls=urls, + health_check_path="/api/health", + health_check_interval_secs=health_check_interval_secs, + health_check_timeout_secs=health_check_timeout_secs, + check_health=check_health, + choice_type=choice_type, + ) + BaseModelController.__init__(self) @api_remote(path="/api/controller/models", method="POST") async def register_instance(self, instance: ModelInstance) -> bool: @@ -139,13 +154,19 @@ async def model_apply(self) -> bool: def initialize_controller( - app=None, remote_controller_addr: str = None, host: str = None, port: int = None + app=None, + remote_controller_addr: str = None, + host: str = None, + port: int = None, + registry: Optional[ModelRegistry] = None, ): global controller if remote_controller_addr: controller.backend = _RemoteModelController(remote_controller_addr) else: - controller.backend = LocalModelController() + if not registry: + registry = EmbeddedModelRegistry() + controller.backend = LocalModelController(registry=registry) if app: app.include_router(router, prefix="/api", tags=["Model"]) @@ -158,6 +179,12 @@ def initialize_controller( uvicorn.run(app, host=host, port=port, log_level="info") +@router.get("/health") +async def api_health_check(): + """Health check API.""" + return {"status": "ok"} + + @router.post("/controller/models") async def api_register_instance(request: ModelInstance): return await controller.register_instance(request) @@ -179,6 +206,87 @@ async def api_model_heartbeat(request: ModelInstance): return await controller.send_heartbeat(request) +def _create_registry(controller_params: ModelControllerParameters) -> ModelRegistry: + """Create a model registry based on the controller parameters. + + Registry will store the metadata of all model instances, it will be a high + availability service for model instances if you use a database registry now. Also, + we can implement more registry types in the future. + """ + registry_type = controller_params.registry_type.strip() + if controller_params.registry_type == "embedded": + return EmbeddedModelRegistry( + heartbeat_interval_secs=controller_params.heartbeat_interval_secs, + heartbeat_timeout_secs=controller_params.heartbeat_timeout_secs, + ) + elif controller_params.registry_type == "database": + from urllib.parse import quote + from urllib.parse import quote_plus as urlquote + + from dbgpt.model.cluster.registry_impl.storage import StorageModelRegistry + + try_to_create_db = False + + if controller_params.registry_db_type == "mysql": + db_name = controller_params.registry_db_name + db_host = controller_params.registry_db_host + db_port = controller_params.registry_db_port + db_user = controller_params.registry_db_user + db_password = controller_params.registry_db_password + if not db_name: + raise ValueError( + "Registry DB name is required when using MySQL registry." + ) + if not db_host: + raise ValueError( + "Registry DB host is required when using MySQL registry." + ) + if not db_port: + raise ValueError( + "Registry DB port is required when using MySQL registry." + ) + if not db_user: + raise ValueError( + "Registry DB user is required when using MySQL registry." + ) + if not db_password: + raise ValueError( + "Registry DB password is required when using MySQL registry." + ) + db_url = ( + f"mysql+pymysql://{quote(db_user)}:" + f"{urlquote(db_password)}@" + f"{db_host}:" + f"{str(db_port)}/" + f"{db_name}?charset=utf8mb4" + ) + elif controller_params.registry_db_type == "sqlite": + db_name = controller_params.registry_db_name + if not db_name: + raise ValueError( + "Registry DB name is required when using SQLite registry." + ) + db_url = f"sqlite:///{db_name}" + try_to_create_db = True + else: + raise ValueError( + f"Unsupported registry DB type: {controller_params.registry_db_type}" + ) + + registry = StorageModelRegistry.from_url( + db_url, + db_name, + pool_size=controller_params.registry_db_pool_size, + max_overflow=controller_params.registry_db_max_overflow, + try_to_create_db=try_to_create_db, + heartbeat_interval_secs=controller_params.heartbeat_interval_secs, + heartbeat_timeout_secs=controller_params.heartbeat_timeout_secs, + ) + return registry + else: + raise ValueError(f"Unsupported registry type: {registry_type}") + + def run_model_controller(): parser = EnvArgumentParser() env_prefix = "controller_" @@ -192,8 +300,11 @@ def run_model_controller(): logging_level=controller_params.log_level, logger_filename=controller_params.log_file, ) + registry = _create_registry(controller_params) - initialize_controller(host=controller_params.host, port=controller_params.port) + initialize_controller( + host=controller_params.host, port=controller_params.port, registry=registry + ) if __name__ == "__main__": diff --git a/tools/__init__.py b/dbgpt/model/cluster/registry_impl/__init__.py similarity index 100% rename from tools/__init__.py rename to dbgpt/model/cluster/registry_impl/__init__.py diff --git a/dbgpt/model/cluster/registry_impl/db_storage.py b/dbgpt/model/cluster/registry_impl/db_storage.py new file mode 100644 index 000000000..c10f05738 --- /dev/null +++ b/dbgpt/model/cluster/registry_impl/db_storage.py @@ -0,0 +1,116 @@ +from datetime import datetime + +from sqlalchemy import ( + Boolean, + Column, + DateTime, + Float, + Integer, + String, + UniqueConstraint, +) +from sqlalchemy.orm import Session + +from dbgpt.core.interface.storage import ResourceIdentifier, StorageItemAdapter +from dbgpt.storage.metadata import Model + +from .storage import ModelInstanceStorageItem + + +class ModelInstanceEntity(Model): + """Model instance entity. + + Use database as the registry, here is the table schema of the model instance. + """ + + __tablename__ = "dbgpt_cluster_registry_instance" + __table_args__ = ( + UniqueConstraint( + "model_name", + "host", + "port", + "sys_code", + name="uk_model_instance", + ), + ) + id = Column(Integer, primary_key=True, comment="Auto increment id") + model_name = Column(String(128), nullable=False, comment="Model name") + host = Column(String(128), nullable=False, comment="Host of the model") + port = Column(Integer, nullable=False, comment="Port of the model") + weight = Column(Float, nullable=True, default=1.0, comment="Weight of the model") + check_healthy = Column( + Boolean, + nullable=True, + default=True, + comment="Whether to check the health of the model", + ) + healthy = Column( + Boolean, nullable=True, default=False, comment="Whether the model is healthy" + ) + enabled = Column( + Boolean, nullable=True, default=True, comment="Whether the model is enabled" + ) + prompt_template = Column( + String(128), + nullable=True, + comment="Prompt template for the model instance", + ) + last_heartbeat = Column( + DateTime, + nullable=True, + comment="Last heartbeat time of the model instance", + ) + user_name = Column(String(128), nullable=True, comment="User name") + sys_code = Column(String(128), nullable=True, comment="System code") + gmt_created = Column(DateTime, default=datetime.now, comment="Record creation time") + gmt_modified = Column(DateTime, default=datetime.now, comment="Record update time") + + +class ModelInstanceItemAdapter( + StorageItemAdapter[ModelInstanceStorageItem, ModelInstanceEntity] +): + def to_storage_format(self, item: ModelInstanceStorageItem) -> ModelInstanceEntity: + return ModelInstanceEntity( + model_name=item.model_name, + host=item.host, + port=item.port, + weight=item.weight, + check_healthy=item.check_healthy, + healthy=item.healthy, + enabled=item.enabled, + prompt_template=item.prompt_template, + last_heartbeat=item.last_heartbeat, + # user_name=item.user_name, + # sys_code=item.sys_code, + ) + + def from_storage_format( + self, model: ModelInstanceEntity + ) -> ModelInstanceStorageItem: + return ModelInstanceStorageItem( + model_name=model.model_name, + host=model.host, + port=model.port, + weight=model.weight, + check_healthy=model.check_healthy, + healthy=model.healthy, + enabled=model.enabled, + prompt_template=model.prompt_template, + last_heartbeat=model.last_heartbeat, + ) + + def get_query_for_identifier( + self, + storage_format: ModelInstanceEntity, + resource_id: ResourceIdentifier, + **kwargs, + ): + session: Session = kwargs.get("session") + if session is None: + raise Exception("session is None") + query_obj = session.query(ModelInstanceEntity) + for key, value in resource_id.to_dict().items(): + if value is None: + continue + query_obj = query_obj.filter(getattr(ModelInstanceEntity, key) == value) + return query_obj diff --git a/dbgpt/model/cluster/registry_impl/storage.py b/dbgpt/model/cluster/registry_impl/storage.py new file mode 100644 index 000000000..6cafed7cc --- /dev/null +++ b/dbgpt/model/cluster/registry_impl/storage.py @@ -0,0 +1,374 @@ +import threading +import time +from concurrent.futures import Executor, ThreadPoolExecutor +from dataclasses import dataclass, field +from datetime import datetime, timedelta +from typing import Dict, List, Optional, Tuple + +from dbgpt.component import SystemApp +from dbgpt.core.interface.storage import ( + QuerySpec, + ResourceIdentifier, + StorageInterface, + StorageItem, +) +from dbgpt.util.executor_utils import blocking_func_to_async + +from ...base import ModelInstance +from ..registry import ModelRegistry + + +@dataclass +class ModelInstanceIdentifier(ResourceIdentifier): + identifier_split: str = field(default="___$$$$___", init=False) + model_name: str + host: str + port: int + + def __post_init__(self): + """Post init method.""" + if self.model_name is None: + raise ValueError("model_name is required.") + if self.host is None: + raise ValueError("host is required.") + if self.port is None: + raise ValueError("port is required.") + + if any( + self.identifier_split in key + for key in [self.model_name, self.host, str(self.port)] + if key is not None + ): + raise ValueError( + f"identifier_split {self.identifier_split} is not allowed in " + f"model_name, host, port." + ) + + @property + def str_identifier(self) -> str: + """Return the string identifier of the identifier.""" + return self.identifier_split.join( + key + for key in [ + self.model_name, + self.host, + str(self.port), + ] + if key is not None + ) + + def to_dict(self) -> Dict: + """Convert the identifier to a dict. + + Returns: + Dict: The dict of the identifier. + """ + return { + "model_name": self.model_name, + "host": self.host, + "port": self.port, + } + + +@dataclass +class ModelInstanceStorageItem(StorageItem): + + model_name: str + host: str + port: int + weight: Optional[float] = 1.0 + check_healthy: Optional[bool] = True + healthy: Optional[bool] = False + enabled: Optional[bool] = True + prompt_template: Optional[str] = None + last_heartbeat: Optional[datetime] = None + _identifier: ModelInstanceIdentifier = field(init=False) + + def __post_init__(self): + """Post init method.""" + # Convert last_heartbeat to datetime if it's a timestamp + if isinstance(self.last_heartbeat, (int, float)): + self.last_heartbeat = datetime.fromtimestamp(self.last_heartbeat) + + self._identifier = ModelInstanceIdentifier( + model_name=self.model_name, + host=self.host, + port=self.port, + ) + + @property + def identifier(self) -> ModelInstanceIdentifier: + return self._identifier + + def merge(self, other: "StorageItem") -> None: + if not isinstance(other, ModelInstanceStorageItem): + raise ValueError(f"Cannot merge with {type(other)}") + self.from_object(other) + + def to_dict(self) -> Dict: + last_heartbeat = self.last_heartbeat.timestamp() + return { + "model_name": self.model_name, + "host": self.host, + "port": self.port, + "weight": self.weight, + "check_healthy": self.check_healthy, + "healthy": self.healthy, + "enabled": self.enabled, + "prompt_template": self.prompt_template, + "last_heartbeat": last_heartbeat, + } + + def from_object(self, item: "ModelInstanceStorageItem") -> None: + """Build the item from another item.""" + self.model_name = item.model_name + self.host = item.host + self.port = item.port + self.weight = item.weight + self.check_healthy = item.check_healthy + self.healthy = item.healthy + self.enabled = item.enabled + self.prompt_template = item.prompt_template + self.last_heartbeat = item.last_heartbeat + + @classmethod + def from_model_instance(cls, instance: ModelInstance) -> "ModelInstanceStorageItem": + return cls( + model_name=instance.model_name, + host=instance.host, + port=instance.port, + weight=instance.weight, + check_healthy=instance.check_healthy, + healthy=instance.healthy, + enabled=instance.enabled, + prompt_template=instance.prompt_template, + last_heartbeat=instance.last_heartbeat, + ) + + @classmethod + def to_model_instance(cls, item: "ModelInstanceStorageItem") -> ModelInstance: + return ModelInstance( + model_name=item.model_name, + host=item.host, + port=item.port, + weight=item.weight, + check_healthy=item.check_healthy, + healthy=item.healthy, + enabled=item.enabled, + prompt_template=item.prompt_template, + last_heartbeat=item.last_heartbeat, + ) + + +class StorageModelRegistry(ModelRegistry): + def __init__( + self, + storage: StorageInterface, + system_app: SystemApp | None = None, + executor: Optional[Executor] = None, + heartbeat_interval_secs: float | int = 60, + heartbeat_timeout_secs: int = 120, + ): + super().__init__(system_app) + self._storage = storage + self._executor = executor or ThreadPoolExecutor(max_workers=2) + self.heartbeat_interval_secs = heartbeat_interval_secs + self.heartbeat_timeout_secs = heartbeat_timeout_secs + self.heartbeat_thread = threading.Thread(target=self._heartbeat_checker) + self.heartbeat_thread.daemon = True + self.heartbeat_thread.start() + + @classmethod + def from_url( + cls, + db_url: str, + db_name: str, + pool_size: int = 5, + max_overflow: int = 10, + try_to_create_db: bool = False, + **kwargs, + ) -> "StorageModelRegistry": + from dbgpt.storage.metadata.db_manager import DatabaseManager, initialize_db + from dbgpt.storage.metadata.db_storage import SQLAlchemyStorage + from dbgpt.util.serialization.json_serialization import JsonSerializer + + from .db_storage import ModelInstanceEntity, ModelInstanceItemAdapter + + engine_args = { + "pool_size": pool_size, + "max_overflow": max_overflow, + "pool_timeout": 30, + "pool_recycle": 3600, + "pool_pre_ping": True, + } + + db: DatabaseManager = initialize_db( + db_url, db_name, engine_args, try_to_create_db=try_to_create_db + ) + storage_adapter = ModelInstanceItemAdapter() + serializer = JsonSerializer() + storage = SQLAlchemyStorage( + db, + ModelInstanceEntity, + storage_adapter, + serializer, + ) + return cls(storage, **kwargs) + + async def _get_instances_by_model( + self, model_name: str, host: str, port: int, healthy_only: bool = False + ) -> Tuple[List[ModelInstanceStorageItem], List[ModelInstanceStorageItem]]: + query_spec = QuerySpec(conditions={"model_name": model_name}) + # Query all instances of the model + instances = await blocking_func_to_async( + self._executor, self._storage.query, query_spec, ModelInstanceStorageItem + ) + if healthy_only: + instances = [ins for ins in instances if ins.healthy is True] + exist_ins = [ins for ins in instances if ins.host == host and ins.port == port] + return instances, exist_ins + + def _heartbeat_checker(self): + while True: + all_instances: List[ModelInstanceStorageItem] = self._storage.query( + QuerySpec(conditions={}), ModelInstanceStorageItem + ) + for instance in all_instances: + if ( + instance.check_healthy + and datetime.now() - instance.last_heartbeat + > timedelta(seconds=self.heartbeat_timeout_secs) + ): + instance.healthy = False + self._storage.update(instance) + time.sleep(self.heartbeat_interval_secs) + + async def register_instance(self, instance: ModelInstance) -> bool: + model_name = instance.model_name.strip() + host = instance.host.strip() + port = instance.port + _, exist_ins = await self._get_instances_by_model( + model_name, host, port, healthy_only=False + ) + if exist_ins: + # Exist instances, just update the instance + # One exist instance at most + ins: ModelInstanceStorageItem = exist_ins[0] + # Update instance + ins.weight = instance.weight + ins.healthy = True + ins.prompt_template = instance.prompt_template + ins.last_heartbeat = datetime.now() + await blocking_func_to_async(self._executor, self._storage.update, ins) + else: + # No exist instance, save the new instance + new_inst = ModelInstanceStorageItem.from_model_instance(instance) + new_inst.healthy = True + new_inst.last_heartbeat = datetime.now() + await blocking_func_to_async(self._executor, self._storage.save, new_inst) + return True + + async def deregister_instance(self, instance: ModelInstance) -> bool: + """Deregister a model instance. + + If the instance exists, set the instance as unhealthy, nothing to do if the + instance does not exist. + + Args: + instance (ModelInstance): The instance to deregister. + """ + model_name = instance.model_name.strip() + host = instance.host.strip() + port = instance.port + _, exist_ins = await self._get_instances_by_model( + model_name, host, port, healthy_only=False + ) + if exist_ins: + ins = exist_ins[0] + ins.healthy = False + await blocking_func_to_async(self._executor, self._storage.update, ins) + return True + + async def get_all_instances( + self, model_name: str, healthy_only: bool = False + ) -> List[ModelInstance]: + """Get all instances of a model(Async). + + Args: + model_name (str): The model name. + healthy_only (bool): Whether only get healthy instances. Defaults to False. + """ + return await blocking_func_to_async( + self._executor, self.sync_get_all_instances, model_name, healthy_only + ) + + def sync_get_all_instances( + self, model_name: str, healthy_only: bool = False + ) -> List[ModelInstance]: + """Get all instances of a model. + + Args: + model_name (str): The model name. + healthy_only (bool): Whether only get healthy instances. Defaults to False. + + Returns: + List[ModelInstance]: The list of instances. + """ + instances = self._storage.query( + QuerySpec(conditions={"model_name": model_name}), ModelInstanceStorageItem + ) + if healthy_only: + instances = [ins for ins in instances if ins.healthy is True] + return [ModelInstanceStorageItem.to_model_instance(ins) for ins in instances] + + async def get_all_model_instances( + self, healthy_only: bool = False + ) -> List[ModelInstance]: + """Get all model instances. + + Args: + healthy_only (bool): Whether only get healthy instances. Defaults to False. + + Returns: + List[ModelInstance]: The list of instances. + """ + all_instances = await blocking_func_to_async( + self._executor, + self._storage.query, + QuerySpec(conditions={}), + ModelInstanceStorageItem, + ) + if healthy_only: + all_instances = [ins for ins in all_instances if ins.healthy is True] + return [ + ModelInstanceStorageItem.to_model_instance(ins) for ins in all_instances + ] + + async def send_heartbeat(self, instance: ModelInstance) -> bool: + """Receive heartbeat from model instance. + + Update the last heartbeat time of the instance. If the instance does not exist, + register the instance. + + Args: + instance (ModelInstance): The instance to send heartbeat. + + Returns: + bool: True if the heartbeat is received successfully. + """ + model_name = instance.model_name.strip() + host = instance.host.strip() + port = instance.port + _, exist_ins = await self._get_instances_by_model( + model_name, host, port, healthy_only=False + ) + if not exist_ins: + # register new instance from heartbeat + await self.register_instance(instance) + return True + else: + ins = exist_ins[0] + ins.last_heartbeat = datetime.now() + ins.healthy = True + await blocking_func_to_async(self._executor, self._storage.update, ins) + return True diff --git a/dbgpt/model/cluster/tests/registry_impl/__init__.py b/dbgpt/model/cluster/tests/registry_impl/__init__.py new file mode 100644 index 000000000..e69de29bb diff --git a/dbgpt/model/cluster/tests/registry_impl/test_storage_registry.py b/dbgpt/model/cluster/tests/registry_impl/test_storage_registry.py new file mode 100644 index 000000000..6158a8c25 --- /dev/null +++ b/dbgpt/model/cluster/tests/registry_impl/test_storage_registry.py @@ -0,0 +1,221 @@ +import asyncio +from concurrent.futures import ThreadPoolExecutor +from datetime import datetime, timedelta +from unittest.mock import AsyncMock, MagicMock + +import pytest + +from dbgpt.core.interface.storage import InMemoryStorage, QuerySpec +from dbgpt.util.serialization.json_serialization import JsonSerializer + +from ...registry_impl.storage import ( + ModelInstance, + ModelInstanceStorageItem, + StorageModelRegistry, +) + + +@pytest.fixture +def in_memory_storage(): + return InMemoryStorage(serializer=JsonSerializer()) + + +@pytest.fixture +def thread_pool_executor(): + return ThreadPoolExecutor(max_workers=2) + + +@pytest.fixture +def registry(in_memory_storage, thread_pool_executor): + return StorageModelRegistry( + storage=in_memory_storage, + executor=thread_pool_executor, + heartbeat_interval_secs=1, + heartbeat_timeout_secs=2, + ) + + +@pytest.fixture +def model_instance(): + return ModelInstance( + model_name="test_model", + host="localhost", + port=8080, + weight=1.0, + check_healthy=True, + healthy=True, + enabled=True, + prompt_template=None, + last_heartbeat=datetime.now(), + ) + + +@pytest.fixture +def model_instance_2(): + return ModelInstance( + model_name="test_model", + host="localhost", + port=8081, + weight=1.0, + check_healthy=True, + healthy=True, + enabled=True, + prompt_template=None, + last_heartbeat=datetime.now(), + ) + + +@pytest.fixture +def model_instance_3(): + return ModelInstance( + model_name="test_model_2", + host="localhost", + port=8082, + weight=1.0, + check_healthy=True, + healthy=True, + enabled=True, + prompt_template=None, + last_heartbeat=datetime.now(), + ) + + +@pytest.fixture +def model_instance_storage_item(model_instance): + return ModelInstanceStorageItem.from_model_instance(model_instance) + + +@pytest.mark.asyncio +async def test_register_instance_new(registry, model_instance): + """Test registering a new model instance.""" + result = await registry.register_instance(model_instance) + + assert result is True + instances = await registry.get_all_instances(model_instance.model_name) + assert len(instances) == 1 + saved_instance = instances[0] + assert saved_instance.model_name == model_instance.model_name + assert saved_instance.host == model_instance.host + assert saved_instance.port == model_instance.port + assert saved_instance.healthy is True + assert saved_instance.last_heartbeat is not None + + +@pytest.mark.asyncio +async def test_register_instance_existing( + registry, model_instance, model_instance_storage_item +): + """Test registering an existing model instance and updating it.""" + await registry.register_instance(model_instance) + + # Register the instance again with updated heartbeat + result = await registry.register_instance(model_instance) + + assert result is True + instances = await registry.get_all_instances(model_instance.model_name) + assert len(instances) == 1 + updated_instance = instances[0] + assert updated_instance.model_name == model_instance.model_name + assert updated_instance.host == model_instance.host + assert updated_instance.port == model_instance.port + assert updated_instance.healthy is True + assert updated_instance.last_heartbeat is not None + + +@pytest.mark.asyncio +async def test_deregister_instance(registry, model_instance): + """Test deregistering a model instance.""" + await registry.register_instance(model_instance) + + result = await registry.deregister_instance(model_instance) + + assert result is True + instances = await registry.get_all_instances(model_instance.model_name) + assert len(instances) == 1 + deregistered_instance = instances[0] + assert deregistered_instance.healthy is False + + +@pytest.mark.asyncio +async def test_get_all_instances(registry, model_instance): + """Test retrieving all model instances.""" + await registry.register_instance(model_instance) + + result = await registry.get_all_instances( + model_instance.model_name, healthy_only=True + ) + + assert len(result) == 1 + assert result[0].model_name == model_instance.model_name + + +def test_sync_get_all_instances(registry, model_instance): + """Test synchronously retrieving all model instances.""" + registry.sync_get_all_instances(model_instance.model_name, healthy_only=True) + registry._storage.save(ModelInstanceStorageItem.from_model_instance(model_instance)) + + result = registry.sync_get_all_instances( + model_instance.model_name, healthy_only=True + ) + + assert len(result) == 1 + assert result[0].model_name == model_instance.model_name + + +@pytest.mark.asyncio +async def test_send_heartbeat_new_instance(registry, model_instance): + """Test sending a heartbeat for a new instance.""" + result = await registry.send_heartbeat(model_instance) + + assert result is True + instances = await registry.get_all_instances(model_instance.model_name) + assert len(instances) == 1 + saved_instance = instances[0] + assert saved_instance.model_name == model_instance.model_name + + +@pytest.mark.asyncio +async def test_send_heartbeat_existing_instance(registry, model_instance): + """Test sending a heartbeat for an existing instance.""" + await registry.register_instance(model_instance) + + # Send heartbeat to update the instance + result = await registry.send_heartbeat(model_instance) + + assert result is True + instances = await registry.get_all_instances(model_instance.model_name) + assert len(instances) == 1 + updated_instance = instances[0] + assert updated_instance.last_heartbeat > model_instance.last_heartbeat + + +@pytest.mark.asyncio +async def test_heartbeat_checker( + in_memory_storage, thread_pool_executor, model_instance +): + """Test the heartbeat checker mechanism.""" + heartbeat_timeout_secs = 1 + registry = StorageModelRegistry( + storage=in_memory_storage, + executor=thread_pool_executor, + heartbeat_interval_secs=0.1, + heartbeat_timeout_secs=heartbeat_timeout_secs, + ) + + async def check_heartbeat(model_name: str, expected_healthy: bool): + instances = await registry.get_all_instances(model_name) + assert len(instances) == 1 + updated_instance = instances[0] + assert updated_instance.healthy == expected_healthy + + await registry.register_instance(model_instance) + # First heartbeat should be successful + await check_heartbeat(model_instance.model_name, True) + # Wait heartbeat timeout + await asyncio.sleep(heartbeat_timeout_secs + 0.5) + await check_heartbeat(model_instance.model_name, False) + + # Send heartbeat again + await registry.send_heartbeat(model_instance) + # Should be healthy again + await check_heartbeat(model_instance.model_name, True) diff --git a/dbgpt/model/cluster/worker/manager.py b/dbgpt/model/cluster/worker/manager.py index 9bc2a80c7..449f52ac6 100644 --- a/dbgpt/model/cluster/worker/manager.py +++ b/dbgpt/model/cluster/worker/manager.py @@ -1059,6 +1059,10 @@ def initialize_worker_manager_in_client( if not app: raise Exception("app can't be None") + if system_app: + logger.info(f"Register WorkerManager {_DefaultWorkerManagerFactory.name}") + system_app.register(_DefaultWorkerManagerFactory, worker_manager) + worker_params: ModelWorkerParameters = _parse_worker_params( model_name=model_name, model_path=model_path, controller_addr=controller_addr ) @@ -1104,8 +1108,6 @@ def initialize_worker_manager_in_client( if include_router and app: # mount WorkerManager router app.include_router(router, prefix="/api") - if system_app: - system_app.register(_DefaultWorkerManagerFactory, worker_manager) def run_worker_manager( diff --git a/dbgpt/model/parameter.py b/dbgpt/model/parameter.py index f3396d564..5aad75a50 100644 --- a/dbgpt/model/parameter.py +++ b/dbgpt/model/parameter.py @@ -55,6 +55,84 @@ class ModelControllerParameters(BaseParameters): port: Optional[int] = field( default=8000, metadata={"help": "Model Controller deploy port"} ) + registry_type: Optional[str] = field( + default="embedded", + metadata={ + "help": "Registry type: embedded, database...", + "valid_values": ["embedded", "database"], + }, + ) + registry_db_type: Optional[str] = field( + default="mysql", + metadata={ + "help": "Registry database type, now only support sqlite and mysql, it is " + "valid when registry_type is database", + "valid_values": ["mysql", "sqlite"], + }, + ) + registry_db_name: Optional[str] = field( + default="dbgpt", + metadata={ + "help": "Registry database name, just for database, it is valid when " + "registry_type is database, please set to full database path for sqlite" + }, + ) + registry_db_host: Optional[str] = field( + default=None, + metadata={ + "help": "Registry database host, just for database, it is valid when " + "registry_type is database" + }, + ) + registry_db_port: Optional[int] = field( + default=None, + metadata={ + "help": "Registry database port, just for database, it is valid when " + "registry_type is database" + }, + ) + registry_db_user: Optional[str] = field( + default=None, + metadata={ + "help": "Registry database user, just for database, it is valid when " + "registry_type is database" + }, + ) + registry_db_password: Optional[str] = field( + default=None, + metadata={ + "help": "Registry database password, just for database, it is valid when " + "registry_type is database. We recommend to use environment variable to " + "store password, you can set it in your environment variable like " + "export CONTROLLER_REGISTRY_DB_PASSWORD='your_password'" + }, + ) + registry_db_pool_size: Optional[int] = field( + default=5, + metadata={ + "help": "Registry database pool size, just for database, it is valid when " + "registry_type is database" + }, + ) + registry_db_max_overflow: Optional[int] = field( + default=10, + metadata={ + "help": "Registry database max overflow, just for database, it is valid " + "when registry_type is database" + }, + ) + + heartbeat_interval_secs: Optional[int] = field( + default=20, metadata={"help": "The interval for checking heartbeats (seconds)"} + ) + heartbeat_timeout_secs: Optional[int] = field( + default=60, + metadata={ + "help": "The timeout for checking heartbeats (seconds), it will be set " + "unhealthy if the worker is not responding in this time" + }, + ) + daemon: Optional[bool] = field( default=False, metadata={"help": "Run Model Controller in background"} ) diff --git a/dbgpt/serve/core/tests/conftest.py b/dbgpt/serve/core/tests/conftest.py index efc02cf47..b4a0d1c5e 100644 --- a/dbgpt/serve/core/tests/conftest.py +++ b/dbgpt/serve/core/tests/conftest.py @@ -3,7 +3,7 @@ import pytest import pytest_asyncio from fastapi.middleware.cors import CORSMiddleware -from httpx import AsyncClient +from httpx import ASGITransport, AsyncClient from dbgpt.component import SystemApp from dbgpt.util import AppConfig @@ -56,7 +56,9 @@ async def client(request, asystem_app: SystemApp): test_app = asystem_app.app - async with AsyncClient(app=test_app, base_url=base_url, headers=headers) as client: + async with AsyncClient( + transport=ASGITransport(test_app), base_url=base_url, headers=headers + ) as client: for router in routers: test_app.include_router(router) if app_caller: diff --git a/dbgpt/serve/prompt/tests/test_endpoints.py b/dbgpt/serve/prompt/tests/test_endpoints.py index f9fd1db74..e73a456c4 100644 --- a/dbgpt/serve/prompt/tests/test_endpoints.py +++ b/dbgpt/serve/prompt/tests/test_endpoints.py @@ -86,6 +86,7 @@ async def test_api_health(client: AsyncClient, asystem_app, has_auth: bool): ) async def test_api_auth(client: AsyncClient): response = await client.get("/health") + response.raise_for_status() assert response.status_code == 200 assert response.json() == {"status": "ok"} diff --git a/dbgpt/util/api_utils.py b/dbgpt/util/api_utils.py index 175cd55de..3312aaafc 100644 --- a/dbgpt/util/api_utils.py +++ b/dbgpt/util/api_utils.py @@ -1,13 +1,156 @@ +import asyncio import logging +import threading +import time +from abc import ABC +from concurrent.futures import Executor, ThreadPoolExecutor from dataclasses import asdict, is_dataclass +from datetime import datetime, timedelta from inspect import signature -from typing import List, Optional, Tuple, Type, TypeVar, Union, get_type_hints +from typing import List, Literal, Optional, Tuple, Type, TypeVar, Union, get_type_hints T = TypeVar("T") logger = logging.getLogger(__name__) +class APIMixin(ABC): + """API mixin class.""" + + def __init__( + self, + urls: Union[str, List[str]], + health_check_path: str, + health_check_interval_secs: int = 5, + health_check_timeout_secs: int = 30, + check_health: bool = True, + choice_type: Literal["latest_first", "random"] = "latest_first", + executor: Optional[Executor] = None, + ): + if isinstance(urls, str): + # Split by "," + urls = urls.split(",") + urls = [url.strip() for url in urls] + self._remote_urls = urls + self._health_check_path = health_check_path + self._health_urls = [] + self._health_check_interval_secs = health_check_interval_secs + self._health_check_timeout_secs = health_check_timeout_secs + self._heartbeat_map = {} + self._choice_type = choice_type + self._heartbeat_thread = threading.Thread(target=self._heartbeat_checker) + self._heartbeat_executor = executor or ThreadPoolExecutor(max_workers=3) + self._heartbeat_stop_event = threading.Event() + + if check_health: + self._heartbeat_thread.daemon = True + self._heartbeat_thread.start() + + def _heartbeat_checker(self): + logger.debug("Running health check") + while not self._heartbeat_stop_event.is_set(): + try: + healthy_urls = self._check_and_update_health() + logger.debug(f"Healthy urls: {healthy_urls}") + except Exception as e: + logger.warning(f"Health check failed, error: {e}") + time.sleep(self._health_check_interval_secs) + + def __del__(self): + + self._heartbeat_stop_event.set() + + def _check_health(self, url: str) -> Tuple[bool, str]: + try: + import requests + + logger.debug(f"Checking health for {url}") + req_url = url + self._health_check_path + response = requests.get(req_url, timeout=10) + return response.status_code == 200, url + except Exception as e: + logger.warning(f"Health check failed for {url}, error: {e}") + return False, url + + def _check_and_update_health(self) -> List[str]: + """Check health of all remote urls and update the health urls list.""" + check_tasks = [] + check_results = [] + for url in self._remote_urls: + check_tasks.append(self._heartbeat_executor.submit(self._check_health, url)) + for task in check_tasks: + check_results.append(task.result()) + now = datetime.now() + for is_healthy, url in check_results: + if is_healthy: + self._heartbeat_map[url] = now + healthy_urls = [] + for url, last_heartbeat in self._heartbeat_map.items(): + if now - last_heartbeat < timedelta( + seconds=self._health_check_interval_secs + ): + healthy_urls.append((url, last_heartbeat)) + # Sort by last heartbeat time, latest first + healthy_urls.sort(key=lambda x: x[1], reverse=True) + + self._health_urls = [url for url, _ in healthy_urls] + return self._health_urls + + async def select_url(self, max_wait_health_timeout_secs: int = 2) -> str: + """Select a healthy url to send request. + + If no healthy urls found, select randomly. + """ + import random + + def _select(urls: List[str]): + if self._choice_type == "latest_first": + return urls[0] + elif self._choice_type == "random": + return random.choice(urls) + else: + raise ValueError(f"Invalid choice type: {self._choice_type}") + + if self._health_urls: + return _select(self._health_urls) + elif max_wait_health_timeout_secs > 0: + start_time = datetime.now() + while datetime.now() - start_time < timedelta( + seconds=max_wait_health_timeout_secs + ): + if self._health_urls: + return _select(self._health_urls) + await asyncio.sleep(0.1) + logger.warning("No healthy urls found, selecting randomly") + return _select(self._remote_urls) + + def sync_select_url(self, max_wait_health_timeout_secs: int = 2) -> str: + """Synchronous version of select_url.""" + import random + import time + + def _select(urls: List[str]): + if self._choice_type == "latest_first": + return urls[0] + elif self._choice_type == "random": + return random.choice(urls) + else: + raise ValueError(f"Invalid choice type: {self._choice_type}") + + if self._health_urls: + return _select(self._health_urls) + elif max_wait_health_timeout_secs > 0: + start_time = datetime.now() + while datetime.now() - start_time < timedelta( + seconds=max_wait_health_timeout_secs + ): + if self._health_urls: + return _select(self._health_urls) + time.sleep(0.1) + logger.warning("No healthy urls found, selecting randomly") + return _select(self._remote_urls) + + def _extract_dataclass_from_generic(type_hint: Type[T]) -> Union[Type[T], None]: import typing_inspect @@ -17,7 +160,7 @@ def _extract_dataclass_from_generic(type_hint: Type[T]) -> Union[Type[T], None]: return None -def _build_request(self, func, path, method, *args, **kwargs): +def _build_request(self, base_url, func, path, method, *args, **kwargs): return_type = get_type_hints(func).get("return") if return_type is None: raise TypeError("Return type must be annotated in the decorated function.") @@ -27,7 +170,6 @@ def _build_request(self, func, path, method, *args, **kwargs): if not actual_dataclass: actual_dataclass = return_type sig = signature(func) - base_url = self.base_url # Get base_url from class instance bound = sig.bind(self, *args, **kwargs) bound.apply_defaults() @@ -61,13 +203,22 @@ def _build_request(self, func, path, method, *args, **kwargs): return return_type, actual_dataclass, request_params -def _api_remote(path, method="GET"): +def _api_remote(path: str, method: str = "GET", max_wait_health_timeout_secs: int = 2): def decorator(func): async def wrapper(self, *args, **kwargs): import httpx + if not isinstance(self, APIMixin): + raise TypeError( + "The class must inherit from APIMixin to use the @_api_remote " + "decorator." + ) + # Found a healthy url to send request + base_url = await self.select_url( + max_wait_health_timeout_secs=max_wait_health_timeout_secs + ) return_type, actual_dataclass, request_params = _build_request( - self, func, path, method, *args, **kwargs + self, base_url, func, path, method, *args, **kwargs ) async with httpx.AsyncClient() as client: response = await client.request(**request_params) @@ -84,13 +235,24 @@ async def wrapper(self, *args, **kwargs): return decorator -def _sync_api_remote(path, method="GET"): +def _sync_api_remote( + path: str, method: str = "GET", max_wait_health_timeout_secs: int = 2 +): def decorator(func): def wrapper(self, *args, **kwargs): import requests + if not isinstance(self, APIMixin): + raise TypeError( + "The class must inherit from APIMixin to use the @_sync_api_remote " + "decorator." + ) + base_url = self.sync_select_url( + max_wait_health_timeout_secs=max_wait_health_timeout_secs + ) + return_type, actual_dataclass, request_params = _build_request( - self, func, path, method, *args, **kwargs + self, base_url, func, path, method, *args, **kwargs ) response = requests.request(**request_params) diff --git a/dbgpt/util/tests/test_api_utils.py b/dbgpt/util/tests/test_api_utils.py new file mode 100644 index 000000000..2d0695dc3 --- /dev/null +++ b/dbgpt/util/tests/test_api_utils.py @@ -0,0 +1,105 @@ +import time +from concurrent.futures import ThreadPoolExecutor +from datetime import datetime, timedelta +from unittest.mock import MagicMock, patch + +import pytest + +from ..api_utils import APIMixin + + +# Mock requests.get +@pytest.fixture +def mock_requests_get(): + with patch("requests.get") as mock_get: + yield mock_get + + +@pytest.fixture +def apimixin(): + urls = "http://example.com,http://example2.com" + health_check_path = "/health" + apimixin = APIMixin(urls, health_check_path) + yield apimixin + # Ensure the executor is properly shut down after tests + apimixin._heartbeat_executor.shutdown(wait=False) + + +def test_apimixin_initialization(apimixin): + """Test APIMixin initialization with various parameters.""" + assert apimixin._remote_urls == ["http://example.com", "http://example2.com"] + assert apimixin._health_check_path == "/health" + assert apimixin._health_check_interval_secs == 5 + assert apimixin._health_check_timeout_secs == 30 + assert apimixin._choice_type == "latest_first" + assert isinstance(apimixin._heartbeat_executor, ThreadPoolExecutor) + + +def test_health_check(apimixin, mock_requests_get): + """Test the _check_health method.""" + url = "http://example.com" + + # Mocking a successful response + mock_response = MagicMock() + mock_response.status_code = 200 + mock_requests_get.return_value = mock_response + + is_healthy, checked_url = apimixin._check_health(url) + assert is_healthy + assert checked_url == url + + # Mocking a failed response + mock_requests_get.side_effect = Exception("Connection error") + is_healthy, checked_url = apimixin._check_health(url) + assert not is_healthy + assert checked_url == url + + +def test_check_and_update_health(apimixin, mock_requests_get): + """Test the _check_and_update_health method.""" + apimixin._heartbeat_map = { + "http://example.com": datetime.now() - timedelta(seconds=3), + "http://example2.com": datetime.now() - timedelta(seconds=10), + } + + # Mocking responses + def side_effect(url, timeout): + mock_response = MagicMock() + if url == "http://example.com/health": + mock_response.status_code = 200 + elif url == "http://example2.com/health": + mock_response.status_code = 500 + return mock_response + + mock_requests_get.side_effect = side_effect + + health_urls = apimixin._check_and_update_health() + assert "http://example.com" in health_urls + assert "http://example2.com" not in health_urls + + +@pytest.mark.asyncio +async def test_select_url(apimixin, mock_requests_get): + """Test the async select_url method.""" + apimixin._health_urls = ["http://example.com"] + + selected_url = await apimixin.select_url() + assert selected_url == "http://example.com" + + # Test with no healthy URLs + apimixin._health_urls = [] + selected_url = await apimixin.select_url(max_wait_health_timeout_secs=1) + assert selected_url in ["http://example.com", "http://example2.com"] + + +def test_sync_select_url(apimixin, mock_requests_get): + """Test the synchronous sync_select_url method.""" + apimixin._health_urls = ["http://example.com"] + + selected_url = apimixin.sync_select_url() + assert selected_url == "http://example.com" + + # Test with no healthy URLs + apimixin._health_urls = [] + selected_url = apimixin.sync_select_url(max_wait_health_timeout_secs=1) + assert selected_url in ["http://example.com", "http://example2.com"] diff --git a/dbgpt/util/utils.py b/dbgpt/util/utils.py index 03ae1473d..0dd87543c 100644 --- a/dbgpt/util/utils.py +++ b/dbgpt/util/utils.py @@ -172,7 +172,7 @@ def setup_http_service_logging(exclude_paths: Optional[List[str]] = None): """ if not exclude_paths: # Not show heartbeat log - exclude_paths = ["/api/controller/heartbeat"] + exclude_paths = ["/api/controller/heartbeat", "/api/health"] uvicorn_logger = logging.getLogger("uvicorn.access") if uvicorn_logger: for path in exclude_paths: diff --git a/docker/base/Dockerfile b/docker/base/Dockerfile index 268913683..c30fa7981 100644 --- a/docker/base/Dockerfile +++ b/docker/base/Dockerfile @@ -34,15 +34,16 @@ RUN pip3 install --upgrade pip -i $PIP_INDEX_URL \ # install openai for proxyllm && pip3 install -i $PIP_INDEX_URL ".[openai]" -RUN (if [ "${LANGUAGE}" = "zh" ]; \ - # language is zh, download zh_core_web_sm from github - then wget https://github.com/explosion/spacy-models/releases/download/zh_core_web_sm-3.7.0/zh_core_web_sm-3.7.0-py3-none-any.whl -O /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl \ - && pip3 install /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl -i $PIP_INDEX_URL \ - && rm /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl; \ - # not zh, download directly - else python3 -m spacy download zh_core_web_sm; \ - fi;) \ - && rm -rf `pip3 cache dir` +# Not install spacy model for now +#RUN (if [ "${LANGUAGE}" = "zh" ]; \ +# # language is zh, download zh_core_web_sm from github +# then wget https://github.com/explosion/spacy-models/releases/download/zh_core_web_sm-3.7.0/zh_core_web_sm-3.7.0-py3-none-any.whl -O /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl \ +# && pip3 install /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl -i $PIP_INDEX_URL \ +# && rm /tmp/zh_core_web_sm-3.7.0-py3-none-any.whl; \ +# # not zh, download directly +# else python3 -m spacy download zh_core_web_sm; \ +# fi;) \ +# && rm -rf `pip3 cache dir` ARG BUILD_LOCAL_CODE="false" # COPY the rest of the app diff --git a/docker/base/build_proxy_image.sh b/docker/base/build_proxy_image.sh new file mode 100755 index 000000000..c3306f1db --- /dev/null +++ b/docker/base/build_proxy_image.sh @@ -0,0 +1,17 @@ +#!/bin/bash + +SCRIPT_LOCATION=$0 +cd "$(dirname "$SCRIPT_LOCATION")" +WORK_DIR=$(pwd) + +if [[ " $* " == *" --help "* ]] || [[ " $* " == *" -h "* ]]; then + bash $WORK_DIR/build_image.sh "$@" + exit 0 +fi + +bash $WORK_DIR/build_image.sh --install-mode openai "$@" + +if [ 0 -ne $? ]; then + echo "Error: build base image failed" + exit 1 +fi \ No newline at end of file diff --git a/docker/compose_examples/ha-cluster-docker-compose.yml b/docker/compose_examples/ha-cluster-docker-compose.yml new file mode 100644 index 000000000..412a7ab7a --- /dev/null +++ b/docker/compose_examples/ha-cluster-docker-compose.yml @@ -0,0 +1,127 @@ +# An example of using docker-compose to start a HA model serving cluster with two controllers and one worker. +# For simplicity, we use chatgpt_proxyllm as the model for the worker, and we build a new docker image named eosphorosai/dbgpt-openai:latest. +# How to build the image: +# run `bash ./docker/base/build_proxy_image.sh` in the root directory of the project. +# If you want to use other pip index url, you can run command with `--pip-index-url` option. +# For example, `bash ./docker/base/build_proxy_image.sh --pip-index-url https://pypi.tuna.tsinghua.edu.cn/simple` +# +# How to start the cluster: +# 1. run `cd docker/compose_examples` +# 2. run `OPENAI_API_KEY="{your api key}" OPENAI_API_BASE="https://api.openai.com/v1" docker compose -f ha-cluster-docker-compose.yml up -d` +# Note: Make sure you have set the environment variables OPENAI_API_KEY. +version: '3.10' + +services: + init: + image: busybox + volumes: + - ../examples/sqls:/sqls + - ../../assets/schema/dbgpt.sql:/dbgpt.sql + - dbgpt-init-scripts:/docker-entrypoint-initdb.d + command: /bin/sh -c "cp /dbgpt.sql /docker-entrypoint-initdb.d/ && cp /sqls/* /docker-entrypoint-initdb.d/ && ls /docker-entrypoint-initdb.d/" + + db: + image: mysql/mysql-server + environment: + MYSQL_USER: 'user' + MYSQL_PASSWORD: 'password' + MYSQL_ROOT_PASSWORD: 'aa123456' + ports: + - 3306:3306 + volumes: + - dbgpt-myql-db:/var/lib/mysql + - ../examples/my.cnf:/etc/my.cnf + - dbgpt-init-scripts:/docker-entrypoint-initdb.d + restart: unless-stopped + networks: + - dbgptnet + depends_on: + - init + controller-1: + image: eosphorosai/dbgpt-openai:latest + command: dbgpt start controller --registry_type database --registry_db_type mysql --registry_db_name dbgpt --registry_db_host db --registry_db_port 3306 --registry_db_user root --registry_db_password aa123456 + volumes: + - ../../:/app + restart: unless-stopped + networks: + - dbgptnet + depends_on: + - db + controller-2: + image: eosphorosai/dbgpt-openai:latest + command: dbgpt start controller --registry_type database --registry_db_type mysql --registry_db_name dbgpt --registry_db_host db --registry_db_port 3306 --registry_db_user root --registry_db_password aa123456 + volumes: + - ../../:/app + restart: unless-stopped + networks: + - dbgptnet + depends_on: + - db + llm-worker: + image: eosphorosai/dbgpt-openai:latest + command: dbgpt start worker --model_type proxy --model_name chatgpt_proxyllm --model_path chatgpt_proxyllm --proxy_server_url ${OPENAI_API_BASE}/chat/completions --proxy_api_key ${OPENAI_API_KEY} --controller_addr "http://controller-1:8000,http://controller-2:8000" + environment: + - DBGPT_LOG_LEVEL=DEBUG + # Your real openai model name, e.g. gpt-3.5-turbo, gpt-4o + - PROXYLLM_BACKEND=gpt-3.5-turbo + depends_on: + - controller-1 + - controller-2 + volumes: + - ../../:/app + restart: unless-stopped + networks: + - dbgptnet + ipc: host + embedding-worker: + image: eosphorosai/dbgpt-openai:latest + command: dbgpt start worker --worker_type text2vec --model_name proxy_http_openapi --model_path proxy_http_openapi --proxy_server_url ${OPENAI_API_BASE}/embeddings --proxy_api_key ${OPENAI_API_KEY} --controller_addr "http://controller-1:8000,http://controller-2:8000" + environment: + - DBGPT_LOG_LEVEL=DEBUG + - proxy_http_openapi_proxy_backend=text-embedding-3-small + depends_on: + - controller-1 + - controller-2 + volumes: + - ../../:/app + restart: unless-stopped + networks: + - dbgptnet + ipc: host + webserver: + image: eosphorosai/dbgpt-openai:latest + command: dbgpt start webserver --light --remote_embedding --controller_addr "http://controller-1:8000,http://controller-2:8000" + environment: + - DBGPT_LOG_LEVEL=DEBUG + - LOCAL_DB_TYPE=mysql + - LOCAL_DB_HOST=db + - LOCAL_DB_USER=root + - LOCAL_DB_PASSWORD=aa123456 + - LLM_MODEL=chatgpt_proxyllm + - EMBEDDING_MODEL=proxy_http_openapi + depends_on: + - controller-1 + - controller-2 + - llm-worker + - embedding-worker + volumes: + - ../../:/app + - dbgpt-data:/app/pilot/data + - dbgpt-message:/app/pilot/message + # env_file: + # - .env.template + ports: + - 5670:5670/tcp + # webserver may be failed, it must wait all sqls in /docker-entrypoint-initdb.d execute finish. + restart: unless-stopped + networks: + - dbgptnet +volumes: + dbgpt-init-scripts: + dbgpt-myql-db: + dbgpt-data: + dbgpt-message: +networks: + dbgptnet: + driver: bridge + name: dbgptnet \ No newline at end of file diff --git a/docs/docs/api/datasource.md b/docs/docs/api/datasource.md index 7553bb673..e6e6cff23 100644 --- a/docs/docs/api/datasource.md +++ b/docs/docs/api/datasource.md @@ -34,7 +34,7 @@ curl -X POST "http://localhost:5670/api/v2/chat/completions" \ -H "Authorization: Bearer $DBGPT_API_KEY" \ -H "accept: application/json" \ -H "Content-Type: application/json" \ - -d "{\"messages\":\"show space datas limit 5\",\"model\":\"chatgpt_proxyllm\", \"chat_mode\": \"chat_datasource\", \"chat_param\": \"$DB_NAME\"}" + -d "{\"messages\":\"show space datas limit 5\",\"model\":\"chatgpt_proxyllm\", \"chat_mode\": \"chat_data\", \"chat_param\": \"$DB_NAME\"}" ``` diff --git a/docs/docs/awel/get_started.ipynb b/docs/docs/awel/get_started.ipynb index 4b46adc2a..063cde67c 100644 --- a/docs/docs/awel/get_started.ipynb +++ b/docs/docs/awel/get_started.ipynb @@ -47,11 +47,11 @@ "cell_type": "code", "execution_count": null, "metadata": {}, - "outputs": [], "source": [ "from dbgpt._private.pydantic import BaseModel, Field\n", "from dbgpt.core.awel import DAG, HttpTrigger, MapOperator" - ] + ], + "outputs": [] }, { "attachments": {}, @@ -67,12 +67,12 @@ "cell_type": "code", "execution_count": null, "metadata": {}, - "outputs": [], "source": [ "class TriggerReqBody(BaseModel):\n", " name: str = Field(..., description=\"User name\")\n", " age: int = Field(18, description=\"User age\")" - ] + ], + "outputs": [] }, { "attachments": {}, @@ -87,7 +87,6 @@ "cell_type": "code", "execution_count": null, "metadata": {}, - "outputs": [], "source": [ "class RequestHandleOperator(MapOperator[TriggerReqBody, str]):\n", " def __init__(self, **kwargs):\n", @@ -96,7 +95,8 @@ " async def map(self, input_value: TriggerReqBody) -> str:\n", " print(f\"Receive input value: {input_value}\")\n", " return f\"Hello, {input_value.name}, your age is {input_value.age}\"" - ] + ], + "outputs": [] }, { "attachments": {}, @@ -112,13 +112,13 @@ "cell_type": "code", "execution_count": null, "metadata": {}, - "outputs": [], "source": [ "with DAG(\"simple_dag_example\") as dag:\n", " trigger = HttpTrigger(\"/examples/hello\", request_body=TriggerReqBody)\n", " map_node = RequestHandleOperator()\n", " trigger >> map_node" - ] + ], + "outputs": [] }, { "attachments": {}, @@ -138,12 +138,12 @@ "languageId": "powershell" } }, - "outputs": [], "source": [ "\n", "% curl -X GET http://127.0.0.1:5000/api/v1/awel/trigger/examples/hello\\?name\\=zhangsan\n", "\"Hello, zhangsan, your age is 18\"" - ] + ], + "outputs": [] }, { "attachments": {}, @@ -161,7 +161,6 @@ "languageId": "powershell" } }, - "outputs": [], "source": [ "if __name__ == \"__main__\":\n", " if dag.leaf_nodes[0].dev_mode:\n", @@ -171,7 +170,8 @@ " else:\n", " # Production mode, DB-GPT will automatically load and execute the current file after startup.\n", " pass" - ] + ], + "outputs": [] }, { "cell_type": "code", @@ -181,11 +181,11 @@ "languageId": "powershell" } }, - "outputs": [], "source": [ "curl -X GET http://127.0.0.1:5555/api/v1/awel/trigger/examples/hello\\?name\\=zhangsan\n", "\"Hello, zhangsan, your age is 18\"" - ] + ], + "outputs": [] }, { "attachments": {}, diff --git a/docs/docs/installation/model_service/cluster.md b/docs/docs/installation/model_service/cluster.md index fea91a792..ade43c48a 100644 --- a/docs/docs/installation/model_service/cluster.md +++ b/docs/docs/installation/model_service/cluster.md @@ -124,7 +124,7 @@ MODEL_SERVER=http://127.0.0.1:8000 Or it can be started directly by command to formulate the model. ```shell -LLM_MODEL=glm-4-9b-chat dbgpt start webserver --light +LLM_MODEL=glm-4-9b-chat dbgpt start webserver --light --remote_embedding ``` ## Command line usage diff --git a/docs/docs/installation/model_service/cluster_ha.md b/docs/docs/installation/model_service/cluster_ha.md new file mode 100644 index 000000000..8ca7363b1 --- /dev/null +++ b/docs/docs/installation/model_service/cluster_ha.md @@ -0,0 +1,171 @@ +# High Availability + + +## Architecture + +Here is the architecture of the high availability cluster, more details can be found in +the [cluster deployment](/docs/latest/installation/model_service/cluster) mode and [SMMF](/docs/latest/modules/smmf) module. + +

+ +

+ +The model worker and API server can be deployed on different machines, and the model +worker and API server can be deployed with multiple instances. +But the model controller has only one instance by default, because it is a stateful +service and stores all metadata of the model service, specifically, all metadata are +stored in the component named **Model Registry**. + +The default model registry is `EmbeddedModelRegistry`, which is a simple in-memory component. +To support high availability, we can use `StorageModelRegistry` as the model registry, +it can use a database as the storage backend, such as MySQL, SQLite, etc. + +So we can deploy the model controller with multiple instances, and they can share the metadata by connecting to the same database. + +Now let's see how to deploy the high availability cluster. + +## Deploy High Availability Cluster +For simplicity, we will deploy two model controllers on two machines(`server1` and `server2`), +and deploy a model worker, an embedding model worker, and a web server on another machine(`server3`). + +(Of course, you can deploy all of them on the same machine with different ports.) + +### Prepare A MySQL Database + +1. Install MySQL, create a database and a user for the model controller. +2. Create a table for the model controller, you can use the following SQL script to create the table. + +```sql + +-- For deploy model cluster of DB-GPT(StorageModelRegistry) +CREATE TABLE IF NOT EXISTS `dbgpt_cluster_registry_instance` ( + `id` int(11) NOT NULL AUTO_INCREMENT COMMENT 'Auto increment id', + `model_name` varchar(128) NOT NULL COMMENT 'Model name', + `host` varchar(128) NOT NULL COMMENT 'Host of the model', + `port` int(11) NOT NULL COMMENT 'Port of the model', + `weight` float DEFAULT 1.0 COMMENT 'Weight of the model', + `check_healthy` tinyint(1) DEFAULT 1 COMMENT 'Whether to check the health of the model', + `healthy` tinyint(1) DEFAULT 0 COMMENT 'Whether the model is healthy', + `enabled` tinyint(1) DEFAULT 1 COMMENT 'Whether the model is enabled', + `prompt_template` varchar(128) DEFAULT NULL COMMENT 'Prompt template for the model instance', + `last_heartbeat` datetime DEFAULT NULL COMMENT 'Last heartbeat time of the model instance', + `user_name` varchar(128) DEFAULT NULL COMMENT 'User name', + `sys_code` varchar(128) DEFAULT NULL COMMENT 'System code', + `gmt_created` datetime DEFAULT CURRENT_TIMESTAMP COMMENT 'Record creation time', + `gmt_modified` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT 'Record update time', + PRIMARY KEY (`id`), + UNIQUE KEY `uk_model_instance` (`model_name`, `host`, `port`, `sys_code`) +) ENGINE=InnoDB AUTO_INCREMENT=1 DEFAULT CHARSET=utf8mb4 COMMENT='Cluster model instance table, for registering and managing model instances'; + +``` + +### Start Model Controller With Storage Model Registry + +We need to start the model controllers on two machines(`server1` and `server2`), and +they will share the metadata by connecting to the same database. + +1. Start the model controller on `server1`: + +```bash +dbgpt start controller \ +--port 8000 \ +--registry_type database \ +--registry_db_type mysql \ +--registry_db_name dbgpt \ +--registry_db_host 127.0.0.1 \ +--registry_db_port 3306 \ +--registry_db_user root \ +--registry_db_password aa123456 +``` +2. Start the model controller on `server2`: + +```bash +dbgpt start controller \ +--port 8000 \ +--registry_type database \ +--registry_db_type mysql \ +--registry_db_name dbgpt \ +--registry_db_host 127.0.0.1 \ +--registry_db_port 3306 \ +--registry_db_user root \ +--registry_db_password aa123456 +``` + +Note: please modify the parameters according to your actual situation. + +### Start Model Worker + +:::tip +Start `glm-4-9b-chat` model Worker +::: + +```shell +dbgpt start worker --model_name glm-4-9b-chat \ +--model_path /app/models/glm-4-9b-chat \ +--port 8001 \ +--controller_addr "http://server1:8000,http://server2:8000" +``` +Here we use `server1` and `server2` as the controller address, so the model worker can +register to any healthy controller. + +### Start Embedding Model Worker + +```shell +dbgpt start worker --model_name text2vec \ +--model_path /app/models/text2vec-large-chinese \ +--worker_type text2vec \ +--port 8003 \ +--controller_addr "http://server1:8000,http://server2:8000" +``` +:::info note +⚠️ Make sure to use your own model name and model path. + +::: + +### Deploy Web Server + +```shell +LLM_MODEL=glm-4-9b-chat EMBEDDING_MODEL=text2vec \ +dbgpt start webserver \ +--light \ +--remote_embedding \ +--controller_addr "http://server1:8000,http://server2:8000" +``` + +### Show Your Model Instances + +```bash +CONTROLLER_ADDRESS="http://server1:8000,http://server2:8000" dbgpt model list +``` + +Congratulations! You have successfully deployed a high availability cluster of DB-GPT. + + +## Deploy High Availability Cluster With Docker Compose + +If your want know more about deploying a high availability DB-GPT cluster, you can see +the example of docker compose in `docker/compose_examples/ha-cluster-docker-compose.yml`. +It uses OpenAI LLM and OpenAI embedding model, so you can run it directly. + +Here we will show you how to deploy a high availability cluster of DB-GPT with docker compose. + +First, build the docker image just include openai dependencies: + +```bash +bash ./docker/base/build_proxy_image.sh --pip-index-url https://pypi.tuna.tsinghua.edu.cn/simple +``` + +Then, run the following command to start the high availability cluster: + +```bash +OPENAI_API_KEY="{your api key}" OPENAI_API_BASE="https://api.openai.com/v1" \ +docker compose -f ha-cluster-docker-compose.yml up -d +``` + +## QA + +### It will support more model registry types in the future? +Yes. We will support more model registry types in the future, such as `etcd`, `consul`, etc. + +### How to deploy the high availability cluster with Kubernetes? +We will provide a Helm chart to deploy the high availability cluster with Kubernetes in the future. \ No newline at end of file diff --git a/docs/sidebars.js b/docs/sidebars.js index 164b662ac..e4fe838bf 100755 --- a/docs/sidebars.js +++ b/docs/sidebars.js @@ -227,6 +227,10 @@ const sidebars = { type: 'doc', id: 'installation/model_service/cluster', }, + { + type: 'doc', + id: 'installation/model_service/cluster_ha', + }, ], }, { diff --git a/examples/notebook/agent_auto_plan_dialogue_example.ipynb b/examples/notebook/agent_auto_plan_dialogue_example.ipynb index 550cbbd91..5c861e558 100644 --- a/examples/notebook/agent_auto_plan_dialogue_example.ipynb +++ b/examples/notebook/agent_auto_plan_dialogue_example.ipynb @@ -23,8 +23,8 @@ "\n", "nest_asyncio.apply()" ], - "outputs": [], - "execution_count": 7 + "execution_count": 7, + "outputs": [] }, { "cell_type": "code", @@ -32,12 +32,12 @@ "metadata": { "is_executing": true }, - "outputs": [], "source": [ "# Set your api key and api base url\n", "# os.environ[\"OPENAI_API_KEY\"] = \"Your API\"\n", "# os.environ[\"OPENAI_API_BASE\"] = \"https://api.openai.com/v1\"" - ] + ], + "outputs": [] }, { "cell_type": "code", @@ -84,666 +84,8 @@ " # message=\"find papers on LLM applications from arxiv in the last month, create a markdown table of different domains.\",\n", ")" ], - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mUser\u001B[0m (to PlanManager)-[]:\n", - "\n", - "\"Obtain simple information about issues in the repository 'eosphoros-ai/DB-GPT' in the past three days and analyze the data. Create a Markdown table grouped by day and status.\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: [\n", - " {\n", - " \"serial_number\": \"1\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Use Github API to retrieve issues from the repository 'eosphoros-ai/DB-GPT' in the past three days.\",\n", - " \"rely\": \"\"\n", - " },\n", - " {\n", - " \"serial_number\": \"2\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Analyze the retrieved issues data, extract the necessary information such as issue title, status, and created date.\",\n", - " \"rely\": \"1\"\n", - " },\n", - " {\n", - " \"serial_number\": \"3\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Group the extracted data by day and status.\",\n", - " \"rely\": \"2\"\n", - " },\n", - " {\n", - " \"serial_number\": \"4\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Create a Markdown table with the grouped data.\",\n", - " \"rely\": \"3\"\n", - " }\n", - "]\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanner\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"[\\n {\\n \\\"serial_number\\\": \\\"1\\\",\\n \\\"agent\\\": \\\"CodeEngineer\\\",\\n \\\"content\\\": \\\"Use Github API to retrieve issues from the repository 'eosphoros-ai/DB-GPT' in the past three days.\\\",\\n \\\"rely\\\": \\\"\\\"\\n },\\n {\\n \\\"serial_number\\\": \\\"2\\\",\\n \\\"agent\\\": \\\"CodeEngineer\\\",\\n \\\"content\\\": \\\"Analyze the retrieved issues data, extract the necessary information such as issue title, status, and created date.\\\",\\n \\\"rely\\\": \\\"1\\\"\\n },\\n {\\n \\\"serial_number\\\": \\\"3\\\",\\n \\\"agent\\\": \\\"CodeEngineer\\\",\\n \\\"content\\\": \\\"Group the extracted data by day and status.\\\",\\n \\\"rely\\\": \\\"2\\\"\\n },\\n {\\n \\\"serial_number\\\": \\\"4\\\",\\n \\\"agent\\\": \\\"CodeEngineer\\\",\\n \\\"content\\\": \\\"Create a Markdown table with the grouped data.\\\",\\n \\\"rely\\\": \\\"3\\\"\\n }\\n]\"\n", - "\u001B[32m>>>>>>>>Planner Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Planner Action report: \n", - "execution succeeded,\n", - "[\n", - " {\n", - " \"serial_number\": \"1\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Use Github API to retrieve issues from the repository 'eosphoros-ai/DB-GPT' in the past three days.\",\n", - " \"rely\": \"\"\n", - " },\n", - " {\n", - " \"serial_number\": \"2\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Analyze the retrieved issues data, extract the necessary information such as issue title, status, and created date.\",\n", - " \"rely\": \"1\"\n", - " },\n", - " {\n", - " \"serial_number\": \"3\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Group the extracted data by day and status.\",\n", - " \"rely\": \"2\"\n", - " },\n", - " {\n", - " \"serial_number\": \"4\",\n", - " \"agent\": \"CodeEngineer\",\n", - " \"content\": \"Create a Markdown table with the grouped data.\",\n", - " \"rely\": \"3\"\n", - " }\n", - "]\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "GroupChat select_speaker failed to resolve the next speaker's name. This is because the speaker selection OAI call returned:\n", - "CodeEngineer\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"Use Github API to retrieve issues from the repository 'eosphoros-ai/DB-GPT' in the past three days.\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: To retrieve issues from the GitHub repository 'eosphoros-ai/DB-GPT' in the past three days, we can use the GitHub REST API. Here is a Python script that accomplishes this. This script uses the 'requests' library to send a GET request to the GitHub API. The script then parses the JSON response to filter out issues created in the last three days.\n", - "\n", - "```python\n", - "# python\n", - "import requests\n", - "from datetime import datetime, timedelta\n", - "import pytz\n", - "\n", - "# Get the current date and time\n", - "now = datetime.now()\n", - "\n", - "# Calculate the date 3 days ago\n", - "three_days_ago = now - timedelta(days=3)\n", - "\n", - "# Convert the date to ISO 8601 format\n", - "three_days_ago_iso = three_days_ago.isoformat()\n", - "\n", - "# Specify the URL of the GitHub API\n", - "url = \"https://api.github.com/repos/eosphoros-ai/DB-GPT/issues\"\n", - "\n", - "# Send a GET request to the GitHub API\n", - "response = requests.get(url)\n", - "\n", - "# Check if the request was successful\n", - "if response.status_code == 200:\n", - " # Parse the JSON response\n", - " issues = response.json()\n", - "\n", - " # Filter out issues created in the last three days\n", - " recent_issues = [issue for issue in issues if issue['created_at'] > three_days_ago_iso]\n", - "\n", - " # Print the recent issues\n", - " for issue in recent_issues:\n", - " print(\"Issue ID: \", issue['id'])\n", - " print(\"Issue Title: \", issue['title'])\n", - " print(\"Issue URL: \", issue['html_url'])\n", - " print(\"Issue Created At: \", issue['created_at'])\n", - " print(\"\\n\")\n", - "else:\n", - " print(\"Failed to retrieve issues. HTTP Status Code: \", response.status_code)\n", - "```\n", - "\n", - "Please note that this script doesn't handle pagination. The GitHub API returns a maximum of 30 items per page. If there are more than 30 issues, you will need to send multiple requests to retrieve all issues. Also, the GitHub API has a rate limit. For unauthenticated requests, the rate limit is 60 requests per hour. For authenticated requests, the rate limit is 5000 requests per hour.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...\u001B[0m\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "execute_code was called without specifying a value for use_docker. Since the python docker package is not available, code will be run natively. Note: this fallback behavior is subject to change\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "un_stream ai response: True\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"To retrieve issues from the GitHub repository 'eosphoros-ai/DB-GPT' in the past three days, we can use the GitHub REST API. Here is a Python script that accomplishes this. This script uses the 'requests' library to send a GET request to the GitHub API. The script then parses the JSON response to filter out issues created in the last three days.\\n\\n```python\\n# python\\nimport requests\\nfrom datetime import datetime, timedelta\\nimport pytz\\n\\n# Get the current date and time\\nnow = datetime.now()\\n\\n# Calculate the date 3 days ago\\nthree_days_ago = now - timedelta(days=3)\\n\\n# Convert the date to ISO 8601 format\\nthree_days_ago_iso = three_days_ago.isoformat()\\n\\n# Specify the URL of the GitHub API\\nurl = \\\"https://api.github.com/repos/eosphoros-ai/DB-GPT/issues\\\"\\n\\n# Send a GET request to the GitHub API\\nresponse = requests.get(url)\\n\\n# Check if the request was successful\\nif response.status_code == 200:\\n # Parse the JSON response\\n issues = response.json()\\n\\n # Filter out issues created in the last three days\\n recent_issues = [issue for issue in issues if issue['created_at'] > three_days_ago_iso]\\n\\n # Print the recent issues\\n for issue in recent_issues:\\n print(\\\"Issue ID: \\\", issue['id'])\\n print(\\\"Issue Title: \\\", issue['title'])\\n print(\\\"Issue URL: \\\", issue['html_url'])\\n print(\\\"Issue Created At: \\\", issue['created_at'])\\n print(\\\"\\\\n\\\")\\nelse:\\n print(\\\"Failed to retrieve issues. HTTP Status Code: \\\", response.status_code)\\n```\\n\\nPlease note that this script doesn't handle pagination. The GitHub API returns a maximum of 30 items per page. If there are more than 30 issues, you will need to send multiple requests to retrieve all issues. Also, the GitHub API has a rate limit. For unauthenticated requests, the rate limit is 60 requests per hour. For authenticated requests, the rate limit is 5000 requests per hour.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution succeeded,\n", - "\n", - "Issue ID: 2234707865\n", - "Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/issues/1398\n", - "Issue Created At: 2024-04-10T03:57:27Z\n", - "\n", - "\n", - "Issue ID: 2234675367\n", - "Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/issues/1397\n", - "Issue Created At: 2024-04-10T03:12:34Z\n", - "\n", - "\n", - "Issue ID: 2234172353\n", - "Issue Title: feat(agent): Release agent SDK\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/pull/1396\n", - "Issue Created At: 2024-04-09T19:35:04Z\n", - "\n", - "\n", - "Issue ID: 2232992034\n", - "Issue Title: feat(RAG):add metadata properties filters\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/pull/1395\n", - "Issue Created At: 2024-04-09T09:19:53Z\n", - "\n", - "\n", - "Issue ID: 2232526474\n", - "Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/issues/1389\n", - "Issue Created At: 2024-04-09T03:18:41Z\n", - "\n", - "\n", - "Issue ID: 2230167238\n", - "Issue Title: [Feature][RAG] Metadata Properties Filter\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/issues/1387\n", - "Issue Created At: 2024-04-08T03:15:31Z\n", - "\n", - "\n", - "Issue ID: 2229746443\n", - "Issue Title: Feat: support oceanbase monaco plugin\n", - "Issue URL: https://github.com/eosphoros-ai/DB-GPT/pull/1386\n", - "Issue Created At: 2024-04-07T12:29:13Z\n", - "\n", - "\n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "GroupChat select_speaker failed to resolve the next speaker's name. This is because the speaker selection OAI call returned:\n", - "CodeEngineer\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"Read the result data of the dependent steps in the above historical message to complete the current goal:Analyze the retrieved issues data, extract the necessary information such as issue title, status, and created date.\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: From the historical message, we have the following issues data:\n", - "\n", - "1. Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\n", - " Status: Not specified\n", - " Created Date: 2024-04-10\n", - "\n", - "2. Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\n", - " Status: Not specified\n", - " Created Date: 2024-04-10\n", - "\n", - "3. Issue Title: feat(agent): Release agent SDK\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "4. Issue Title: feat(RAG):add metadata properties filters\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "5. Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "6. Issue Title: [Feature][RAG] Metadata Properties Filter\n", - " Status: Not specified\n", - " Created Date: 2024-04-08\n", - "\n", - "7. Issue Title: Feat: support oceanbase monaco plugin\n", - " Status: Not specified\n", - " Created Date: 2024-04-07\n", - "\n", - "Note: The status of the issues is not provided in the historical message. It could be open, closed, or in review. Further API calls would be needed to retrieve the current status of each issue.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is unknown)...\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"From the historical message, we have the following issues data:\\n\\n1. Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\\n Status: Not specified\\n Created Date: 2024-04-10\\n\\n2. Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\\n Status: Not specified\\n Created Date: 2024-04-10\\n\\n3. Issue Title: feat(agent): Release agent SDK\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n4. Issue Title: feat(RAG):add metadata properties filters\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n5. Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n6. Issue Title: [Feature][RAG] Metadata Properties Filter\\n Status: Not specified\\n Created Date: 2024-04-08\\n\\n7. Issue Title: Feat: support oceanbase monaco plugin\\n Status: Not specified\\n Created Date: 2024-04-07\\n\\nNote: The status of the issues is not provided in the historical message. It could be open, closed, or in review. Further API calls would be needed to retrieve the current status of each issue.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution failed,\n", - "exitcode: 1 (execution failed)\n", - " \n", - "unknown language unknown\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"exitcode: 1 (execution failed)\\n \\nunknown language unknown\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: I apologize for the confusion. The information provided was not meant to be executed as a code but was an analysis of the issue data from the historical message. \n", - "\n", - "Here's a summary again:\n", - "\n", - "1. Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\n", - " Status: Not specified\n", - " Created Date: 2024-04-10\n", - "\n", - "2. Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\n", - " Status: Not specified\n", - " Created Date: 2024-04-10\n", - "\n", - "3. Issue Title: feat(agent): Release agent SDK\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "4. Issue Title: feat(RAG):add metadata properties filters\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "5. Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\n", - " Status: Not specified\n", - " Created Date: 2024-04-09\n", - "\n", - "6. Issue Title: [Feature][RAG] Metadata Properties Filter\n", - " Status: Not specified\n", - " Created Date: 2024-04-08\n", - "\n", - "7. Issue Title: Feat: support oceanbase monaco plugin\n", - " Status: Not specified\n", - " Created Date: 2024-04-07\n", - "\n", - "Please note that this information is a summary and analysis of the issue data retrieved from the historical message and does not need to be executed as code.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is unknown)...\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"I apologize for the confusion. The information provided was not meant to be executed as a code but was an analysis of the issue data from the historical message. \\n\\nHere's a summary again:\\n\\n1. Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\\n Status: Not specified\\n Created Date: 2024-04-10\\n\\n2. Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\\n Status: Not specified\\n Created Date: 2024-04-10\\n\\n3. Issue Title: feat(agent): Release agent SDK\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n4. Issue Title: feat(RAG):add metadata properties filters\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n5. Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\\n Status: Not specified\\n Created Date: 2024-04-09\\n\\n6. Issue Title: [Feature][RAG] Metadata Properties Filter\\n Status: Not specified\\n Created Date: 2024-04-08\\n\\n7. Issue Title: Feat: support oceanbase monaco plugin\\n Status: Not specified\\n Created Date: 2024-04-07\\n\\nPlease note that this information is a summary and analysis of the issue data retrieved from the historical message and does not need to be executed as code.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution failed,\n", - "exitcode: 1 (execution failed)\n", - " \n", - "unknown language unknown\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"exitcode: 1 (execution failed)\\n \\nunknown language unknown\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: I apologize for the confusion, it seems like there was a misunderstanding. The text provided was an analysis and not a code to be executed. \n", - "\n", - "However, if you want to format this information in a more structured way using Python code, you can use the following code:\n", - "\n", - "```python\n", - "# python code\n", - "issues_data = [\n", - " {\"title\": \"[Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\", \"status\": \"Not specified\", \"created_date\": \"2024-04-10\"},\n", - " {\"title\": \"[Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\", \"status\": \"Not specified\", \"created_date\": \"2024-04-10\"},\n", - " {\"title\": \"feat(agent): Release agent SDK\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"feat(RAG):add metadata properties filters\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"[Bug] [datasource] Unable to modify database table structure after executing query\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"[Feature][RAG] Metadata Properties Filter\", \"status\": \"Not specified\", \"created_date\": \"2024-04-08\"},\n", - " {\"title\": \"Feat: support oceanbase monaco plugin\", \"status\": \"Not specified\", \"created_date\": \"2024-04-07\"}\n", - "]\n", - "\n", - "for issue in issues_data:\n", - " print(f\"Issue Title: {issue['title']}\\nStatus: {issue['status']}\\nCreated Date: {issue['created_date']}\\n\")\n", - "```\n", - "\n", - "This code creates a list of dictionaries where each dictionary represents an issue with its title, status, and created date. The code then iterates over this list and prints the information for each issue.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...\u001B[0m\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "execute_code was called without specifying a value for use_docker. Since the python docker package is not available, code will be run natively. Note: this fallback behavior is subject to change\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "un_stream ai response: True\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"I apologize for the confusion, it seems like there was a misunderstanding. The text provided was an analysis and not a code to be executed. \\n\\nHowever, if you want to format this information in a more structured way using Python code, you can use the following code:\\n\\n```python\\n# python code\\nissues_data = [\\n {\\\"title\\\": \\\"[Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-10\\\"},\\n {\\\"title\\\": \\\"[Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-10\\\"},\\n {\\\"title\\\": \\\"feat(agent): Release agent SDK\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"feat(RAG):add metadata properties filters\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"[Bug] [datasource] Unable to modify database table structure after executing query\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"[Feature][RAG] Metadata Properties Filter\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-08\\\"},\\n {\\\"title\\\": \\\"Feat: support oceanbase monaco plugin\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-07\\\"}\\n]\\n\\nfor issue in issues_data:\\n print(f\\\"Issue Title: {issue['title']}\\\\nStatus: {issue['status']}\\\\nCreated Date: {issue['created_date']}\\\\n\\\")\\n```\\n\\nThis code creates a list of dictionaries where each dictionary represents an issue with its title, status, and created date. The code then iterates over this list and prints the information for each issue.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution succeeded,\n", - "\n", - "Issue Title: [Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\n", - "Status: Not specified\n", - "Created Date: 2024-04-10\n", - "\n", - "Issue Title: [Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\n", - "Status: Not specified\n", - "Created Date: 2024-04-10\n", - "\n", - "Issue Title: feat(agent): Release agent SDK\n", - "Status: Not specified\n", - "Created Date: 2024-04-09\n", - "\n", - "Issue Title: feat(RAG):add metadata properties filters\n", - "Status: Not specified\n", - "Created Date: 2024-04-09\n", - "\n", - "Issue Title: [Bug] [datasource] Unable to modify database table structure after executing query\n", - "Status: Not specified\n", - "Created Date: 2024-04-09\n", - "\n", - "Issue Title: [Feature][RAG] Metadata Properties Filter\n", - "Status: Not specified\n", - "Created Date: 2024-04-08\n", - "\n", - "Issue Title: Feat: support oceanbase monaco plugin\n", - "Status: Not specified\n", - "Created Date: 2024-04-07\n", - "\n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "GroupChat select_speaker failed to resolve the next speaker's name. This is because the speaker selection OAI call returned:\n", - "CodeEngineer\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"Read the result data of the dependent steps in the above historical message to complete the current goal:Group the extracted data by day and status.\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: ```python\n", - "# filename: group_issues.py\n", - "\n", - "import pandas as pd\n", - "from datetime import datetime\n", - "\n", - "# The issue data from the historical message\n", - "issue_data = [\n", - " {\"title\": \"[Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\", \"status\": \"Not specified\", \"created_date\": \"2024-04-10\"},\n", - " {\"title\": \"[Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\", \"status\": \"Not specified\", \"created_date\": \"2024-04-10\"},\n", - " {\"title\": \"feat(agent): Release agent SDK\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"feat(RAG):add metadata properties filters\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"[Bug] [datasource] Unable to modify database table structure after executing query\", \"status\": \"Not specified\", \"created_date\": \"2024-04-09\"},\n", - " {\"title\": \"[Feature][RAG] Metadata Properties Filter\", \"status\": \"Not specified\", \"created_date\": \"2024-04-08\"},\n", - " {\"title\": \"Feat: support oceanbase monaco plugin\", \"status\": \"Not specified\", \"created_date\": \"2024-04-07\"},\n", - "]\n", - "\n", - "# Convert the issue data into a DataFrame\n", - "df = pd.DataFrame(issue_data)\n", - "\n", - "# Convert the 'created_date' column to datetime\n", - "df['created_date'] = pd.to_datetime(df['created_date'])\n", - "\n", - "# Group the data by date and status\n", - "grouped = df.groupby([df['created_date'].dt.date, 'status']).size()\n", - "\n", - "# Print the grouped data\n", - "print(grouped)\n", - "```\n", - "\n", - "Please save this Python script as `group_issues.py` and execute it. It will group the extracted issue data by day and status and print the result.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...\u001B[0m\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "execute_code was called without specifying a value for use_docker. Since the python docker package is not available, code will be run natively. Note: this fallback behavior is subject to change\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "un_stream ai response: True\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"```python\\n# filename: group_issues.py\\n\\nimport pandas as pd\\nfrom datetime import datetime\\n\\n# The issue data from the historical message\\nissue_data = [\\n {\\\"title\\\": \\\"[Bug] [remote_worker] 使用model-controller作为集群模式时，通过web进行知识库问答，会将所有模型请求一遍\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-10\\\"},\\n {\\\"title\\\": \\\"[Bug] LLMServer Generate Error, Please CheckErrorInfo.: CUDA error\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-10\\\"},\\n {\\\"title\\\": \\\"feat(agent): Release agent SDK\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"feat(RAG):add metadata properties filters\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"[Bug] [datasource] Unable to modify database table structure after executing query\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-09\\\"},\\n {\\\"title\\\": \\\"[Feature][RAG] Metadata Properties Filter\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-08\\\"},\\n {\\\"title\\\": \\\"Feat: support oceanbase monaco plugin\\\", \\\"status\\\": \\\"Not specified\\\", \\\"created_date\\\": \\\"2024-04-07\\\"},\\n]\\n\\n# Convert the issue data into a DataFrame\\ndf = pd.DataFrame(issue_data)\\n\\n# Convert the 'created_date' column to datetime\\ndf['created_date'] = pd.to_datetime(df['created_date'])\\n\\n# Group the data by date and status\\ngrouped = df.groupby([df['created_date'].dt.date, 'status']).size()\\n\\n# Print the grouped data\\nprint(grouped)\\n```\\n\\nPlease save this Python script as `group_issues.py` and execute it. It will group the extracted issue data by day and status and print the result.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution succeeded,\n", - "\n", - "created_date status \n", - "2024-04-07 Not specified 1\n", - "2024-04-08 Not specified 1\n", - "2024-04-09 Not specified 3\n", - "2024-04-10 Not specified 2\n", - "dtype: int64\n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "GroupChat select_speaker failed to resolve the next speaker's name. This is because the speaker selection OAI call returned:\n", - "CodeEngineer\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"Read the result data of the dependent steps in the above historical message to complete the current goal:Create a Markdown table with the grouped data.\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: To create a Markdown table with the grouped data, we can write a Python function that converts the data into a Markdown table format.\n", - "\n", - "```python\n", - "# python code\n", - "def create_markdown_table(data):\n", - " header = \"| \" + \" | \".join(data.columns) + \" |\"\n", - " separator = \"| \" + \" | \".join(\"---\" for _ in data.columns) + \" |\"\n", - " rows = [\"| \" + \" | \".join(str(x) for x in row) + \" |\" for row in data.values]\n", - " return \"\\n\".join([header, separator] + rows)\n", - "\n", - "# Assuming the data is stored in a pandas DataFrame named 'df'\n", - "print(create_markdown_table(df))\n", - "```\n", - "\n", - "Please replace the `df` with the actual DataFrame that contains your data. This function will output a Markdown table as a string. The table can be copied and pasted into any Markdown file or Markdown-supported text field.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...\u001B[0m\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "execute_code was called without specifying a value for use_docker. Since the python docker package is not available, code will be run natively. Note: this fallback behavior is subject to change\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"To create a Markdown table with the grouped data, we can write a Python function that converts the data into a Markdown table format.\\n\\n```python\\n# python code\\ndef create_markdown_table(data):\\n header = \\\"| \\\" + \\\" | \\\".join(data.columns) + \\\" |\\\"\\n separator = \\\"| \\\" + \\\" | \\\".join(\\\"---\\\" for _ in data.columns) + \\\" |\\\"\\n rows = [\\\"| \\\" + \\\" | \\\".join(str(x) for x in row) + \\\" |\\\" for row in data.values]\\n return \\\"\\\\n\\\".join([header, separator] + rows)\\n\\n# Assuming the data is stored in a pandas DataFrame named 'df'\\nprint(create_markdown_table(df))\\n```\\n\\nPlease replace the `df` with the actual DataFrame that contains your data. This function will output a Markdown table as a string. The table can be copied and pasted into any Markdown file or Markdown-supported text field.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution failed,\n", - "exitcode: 1 (execution failed)\n", - " \n", - "Traceback (most recent call last):\n", - " File \"\", line 9, in \n", - " print(create_markdown_table(df))\n", - "NameError: name 'df' is not defined\n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to Turing)-[]:\n", - "\n", - "\"exitcode: 1 (execution failed)\\n \\nTraceback (most recent call last):\\n File \\\"\\\", line 9, in \\n print(create_markdown_table(df))\\nNameError: name 'df' is not defined\\n\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: Apologies for the confusion. It seems there was a misunderstanding. The 'df' was a placeholder for your actual data. However, since there's no data provided in the context, I'll assume the data based on the previous responses and create a markdown table. \n", - "\n", - "```python\n", - "# python code\n", - "import pandas as pd\n", - "\n", - "# Assuming the data is like this:\n", - "data = {\n", - " 'created_date': ['2024-04-07', '2024-04-08', '2024-04-09', '2024-04-10'],\n", - " 'status': ['Not specified', 'Not specified', 'Not specified', 'Not specified'],\n", - " 'count': [1, 1, 3, 2]\n", - "}\n", - "\n", - "df = pd.DataFrame(data)\n", - "\n", - "def create_markdown_table(data):\n", - " header = \"| \" + \" | \".join(data.columns) + \" |\"\n", - " separator = \"| \" + \" | \".join(\"---\" for _ in data.columns) + \" |\"\n", - " rows = [\"| \" + \" | \".join(str(x) for x in row) + \" |\" for row in data.values]\n", - " return \"\\n\".join([header, separator] + rows)\n", - "\n", - "print(create_markdown_table(df))\n", - "```\n", - "\n", - "Please replace the 'data' dictionary with your actual data in the same format. This script will create a markdown table from the given data.\n", - "\u001B[31m\n", - ">>>>>>>> EXECUTING CODE BLOCK 0 (inferred language is python)...\u001B[0m\n" - ] - }, - { - "name": "stderr", - "output_type": "stream", - "text": [ - "execute_code was called without specifying a value for use_docker. Since the python docker package is not available, code will be run natively. Note: this fallback behavior is subject to change\n" - ] - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "un_stream ai response: True\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mTuring\u001B[0m (to PlanManager)-[gpt-4]:\n", - "\n", - "\"Apologies for the confusion. It seems there was a misunderstanding. The 'df' was a placeholder for your actual data. However, since there's no data provided in the context, I'll assume the data based on the previous responses and create a markdown table. \\n\\n```python\\n# python code\\nimport pandas as pd\\n\\n# Assuming the data is like this:\\ndata = {\\n 'created_date': ['2024-04-07', '2024-04-08', '2024-04-09', '2024-04-10'],\\n 'status': ['Not specified', 'Not specified', 'Not specified', 'Not specified'],\\n 'count': [1, 1, 3, 2]\\n}\\n\\ndf = pd.DataFrame(data)\\n\\ndef create_markdown_table(data):\\n header = \\\"| \\\" + \\\" | \\\".join(data.columns) + \\\" |\\\"\\n separator = \\\"| \\\" + \\\" | \\\".join(\\\"---\\\" for _ in data.columns) + \\\" |\\\"\\n rows = [\\\"| \\\" + \\\" | \\\".join(str(x) for x in row) + \\\" |\\\" for row in data.values]\\n return \\\"\\\\n\\\".join([header, separator] + rows)\\n\\nprint(create_markdown_table(df))\\n```\\n\\nPlease replace the 'data' dictionary with your actual data in the same format. This script will create a markdown table from the given data.\"\n", - "\u001B[32m>>>>>>>>Turing Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>Turing Action report: \n", - "execution succeeded,\n", - "\n", - "| created_date | status | count |\n", - "| --- | --- | --- |\n", - "| 2024-04-07 | Not specified | 1 |\n", - "| 2024-04-08 | Not specified | 1 |\n", - "| 2024-04-09 | Not specified | 3 |\n", - "| 2024-04-10 | Not specified | 2 |\n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mPlanManager\u001B[0m (to User)-[]:\n", - "\n", - "\"Obtain simple information about issues in the repository 'eosphoros-ai/DB-GPT' in the past three days and analyze the data. Create a Markdown table grouped by day and status.\"\n", - "\u001B[32m>>>>>>>>PlanManager Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>PlanManager Action report: \n", - "execution succeeded,\n", - "```vis-code\n", - "{\"exit_success\": true, \"language\": \"python\", \"code\": [[\"python\", \"# python code\\nimport pandas as pd\\n\\n# Assuming the data is like this:\\ndata = {\\n 'created_date': ['2024-04-07', '2024-04-08', '2024-04-09', '2024-04-10'],\\n 'status': ['Not specified', 'Not specified', 'Not specified', 'Not specified'],\\n 'count': [1, 1, 3, 2]\\n}\\n\\ndf = pd.DataFrame(data)\\n\\ndef create_markdown_table(data):\\n header = \\\"| \\\" + \\\" | \\\".join(data.columns) + \\\" |\\\"\\n separator = \\\"| \\\" + \\\" | \\\".join(\\\"---\\\" for _ in data.columns) + \\\" |\\\"\\n rows = [\\\"| \\\" + \\\" | \\\".join(str(x) for x in row) + \\\" |\\\" for row in data.values]\\n return \\\"\\\\n\\\".join([header, separator] + rows)\\n\\nprint(create_markdown_table(df))\"]], \"log\": \"\\n| created_date | status | count |\\n| --- | --- | --- |\\n| 2024-04-07 | Not specified | 1 |\\n| 2024-04-08 | Not specified | 1 |\\n| 2024-04-09 | Not specified | 3 |\\n| 2024-04-10 | Not specified | 2 |\\n\"}\n", - "```\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - } - ], - "execution_count": 4 + "execution_count": 4, + "outputs": [] } ], "metadata": { diff --git a/examples/notebook/agent_awel_layout_dialogue_example.ipynb b/examples/notebook/agent_awel_layout_dialogue_example.ipynb index 267d3a19a..d7b583ab6 100644 --- a/examples/notebook/agent_awel_layout_dialogue_example.ipynb +++ b/examples/notebook/agent_awel_layout_dialogue_example.ipynb @@ -31,8 +31,8 @@ "nest_asyncio.apply()\n", "test_plugin_dir = os.path.join(ROOT_PATH, \"examples/test_files/plugins\")" ], - "outputs": [], - "execution_count": 11 + "execution_count": 11, + "outputs": [] }, { "cell_type": "code", @@ -99,73 +99,16 @@ " # message=\"查询今天的最新热点财经新闻\",\n", ")" ], - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mWrappedAWELLayoutManager\u001B[0m (to LuBan)-[]:\n", - "\n", - "\"查询成都今天天气\"\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: {\n", - " \"tool_name\": \"baidu_search\",\n", - " \"args\": {\n", - " \"query\": \"成都今天天气\"\n", - " },\n", - " \"thought\": \"I have selected the 'baidu_search' tool with the query parameter set to '成都今天天气' to search for the weather in Chengdu today.\"\n", - "}\n", - "{'query': '成都今天天气'}\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mLuBan\u001B[0m (to Aristotle)-[gpt-3.5-turbo]:\n", - "\n", - "\"{\\n \\\"tool_name\\\": \\\"baidu_search\\\",\\n \\\"args\\\": {\\n \\\"query\\\": \\\"成都今天天气\\\"\\n },\\n \\\"thought\\\": \\\"I have selected the 'baidu_search' tool with the query parameter set to '成都今天天气' to search for the weather in Chengdu today.\\\"\\n}\"\n", - "\u001B[32m>>>>>>>>LuBan Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>LuBan Action report: \n", - "execution succeeded,\n", - "### [...天气预报一周_成都天气预报7天、15天、40天天查询_中国...](http://www.baidu.com/link?url=nSNTTnrxEUFL7oMRAYqg98BfeXkWtwHUaYN7WrTjaxBpSy0blKc4jIZ9m34mP97fFARfXJStjbRoBN6U0s0BDq) \n", - " \n", - "### [成都天气_成都天气预报一周_成都天气预报15天](http://www.baidu.com/link?url=Fmp4cnf8Cqqd8N06PpAe3Mn6Esp5q39Scfsnfr7ALxqB5XfoWu9-wY5UjS4n-95Y) \n", - " \n", - "### [【成都天气】成都天气预报,蓝天,蓝天预报,雾霾,雾霾消散,...](http://www.baidu.com/link?url=BQF3cexr1Z6hqkdOjRO2pq8YnOuruBV8nBFY0LE7FJJl8_TCcO806skK-aWkmC8UAZ23K-v3SvoXO58Ayze7Da) \n", - " \n", - "### [...天气预报一周_成都天气预报7天、15天、40天天查询_中国...](http://www.baidu.com/link?url=rt26_NNSBBWHLr0rAX2RPUbBhVjfr4m3Cd21RG7MOe4gsirRquQyp5fMLbSfeU1iC2b1ZhNVjUzlex39iYN_wq) \n", - " \n", - "### [【成都天气预报15天_成都天气预报15天查询】-中国天气网](http://www.baidu.com/link?url=vnZ3GlUxqllZ7Lenc94cImrur2AixgD6dkSOxfNc63PTewisg-RXg3sKzLpBEuPgCWXLr9VnR9gsSZetfPA_94HdTG0It_uAvZpLdUiGmY_) \n", - " \n", - "\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n", - "un_stream ai response: Did not find the information you want.\n", - "\n", - "--------------------------------------------------------------------------------\n", - "\u001B[33mWrappedAWELLayoutManager\u001B[0m (to User)-[]:\n", - "\n", - "\"查询成都今天天气\"\n", - "\u001B[32m>>>>>>>>WrappedAWELLayoutManager Review info: \n", - "Pass(None)\u001B[0m\n", - "\u001B[34m>>>>>>>>WrappedAWELLayoutManager Action report: \n", - "execution succeeded,\n", - "Did not find the information you want.\u001B[0m\n", - "\n", - "--------------------------------------------------------------------------------\n" - ] - } - ], - "execution_count": 12 + "execution_count": 12, + "outputs": [] }, { "cell_type": "code", "execution_count": null, "id": "7ded4107", "metadata": {}, - "outputs": [], - "source": [] + "source": [], + "outputs": [] } ], "metadata": { diff --git a/requirements/dev-requirements.txt b/requirements/dev-requirements.txt index c67f77c5d..364ce239c 100644 --- a/requirements/dev-requirements.txt +++ b/requirements/dev-requirements.txt @@ -14,4 +14,4 @@ aioresponses pre-commit # Type checking mypy==1.7.0 -httpx==0.26.0 \ No newline at end of file +httpx \ No newline at end of file diff --git a/setup.py b/setup.py index a6051cdc6..4702554e1 100644 --- a/setup.py +++ b/setup.py @@ -552,7 +552,7 @@ def quantization_requires(): # 1. Compute Capability 7.5 (sm75). Turing and later architectures are supported. # 2. CUDA Toolkit 11.8 and later. cuda_version = get_cuda_version() - autoawq_latest_version = get_latest_version("autoawq", "", "0.2.4") + # autoawq_latest_version = get_latest_version("autoawq", "", "0.2.4") if cuda_version is None or cuda_version == "12.1": quantization_pkgs.extend(["autoawq", _build_autoawq_requires(), "optimum"]) else: