{"id":1257,"date":"2024-04-29T17:47:04","date_gmt":"2024-04-29T17:47:04","guid":{"rendered":"https:\/\/www.nicekj.com\/?p=1257"},"modified":"2024-04-29T17:47:26","modified_gmt":"2024-04-29T17:47:26","slug":"tongguoopenaihelangchaingoujianarxivlunwenzhaiyaotwitterjiqiren","status":"publish","type":"post","link":"https:\/\/www.nicekj.com\/tongguoopenaihelangchaingoujianarxivlunwenzhaiyaotwitterjiqiren.html","title":{"rendered":"\u901a\u8fc7 OpenAI \u548c Langchain \u6784\u5efa Arxiv \u8bba\u6587\u6458\u8981 Twitter \u673a\u5668\u4eba"},"content":{"rendered":"
\n

\u539f\u6587\u4f5c\u8005\uff1aYeyu Huang<\/p>\n

\u539f\u6587\u94fe\u63a5\uff1alevelup.gitconnected.com\/build-a-twi\u2026<\/a><\/p>\n<\/blockquote>\n

arXiv<\/a> \u662f\u4e00\u4e2a\u6536\u96c6\u7269\u7406\u5b66\u3001\u6570\u5b66\u3001\u8ba1\u7b97\u673a\u79d1\u5b66\u3001\u751f\u7269\u5b66\u4e0e\u6570\u7406\u7ecf\u6d4e\u5b66\u7684\u8bba\u6587\u9884\u5370\u672c\u7684\u7f51\u7ad9\uff0c\u6210\u7acb\u4e8e1991\u5e748\u670814\u65e5\u3002\u622a\u81f32008\u5e7410\u6708\uff0carXiv \u5df2\u6536\u96c6\u8d85\u8fc750\u4e07\u7bc7\u9884\u5370\u672c\uff1b\u81f32014\u5e74\u5e95\uff0c\u85cf\u91cf\u8fbe\u52301\u767e\u4e07\u7bc7\u3002\u622a\u81f32016\u5e7410\u6708\uff0c\u6bcf\u6708\u63d0\u4ea4\u91cf\u8d85\u8fc710,000\u7bc7\u3002<\/p>\n

arXiv \u7f51\u7ad9\u7684\u7814\u7a76\u8bba\u6587\u6bcf\u5929\u90fd\u5728\u66f4\u65b0\uff0c\u8ddf\u4e0a arXiv<\/a> \u6700\u65b0\u79d1\u5b66\u6210\u5c31\u7684\u8282\u594f\u5c31\u6210\u4e86\u4e00\u4ef6\u5177\u6709\u6311\u6218\u7684\u4e8b\u60c5\u3002\u751f\u5728\u8fd9\u4e2a\u5e74\u4ee3\u7684\u6211\u4eec\u662f\u5e78\u8fd0\u7684\uff0c\u8ddf\u4e0a AI \u7684\u6f6e\u6d41\uff0c\u901a\u8fc7\u4f7f\u7528\u5927\u578b\u8bed\u8a00\u6a21\u578b\u53ca\u5176\u652f\u6301\u5e94\u7528\u7a0b\u5e8f\u6846\u67b6\uff08\u5982 Langchain\uff09\uff0c\u53ef\u4ee5\u5b9e\u73b0\u81ea\u52a8\u8ddf\u8e2a arXiv \u8bba\u6587\u7684\u65b0\u7248\u672c\u3002\u5728\u672c\u6559\u7a0b\u4e2d\uff0c\u5c06\u4ecb\u7ecd\u5982\u4f55\u8bbe\u8ba1\u4e00\u4e2a Twitter \u673a\u5668\u4eba\uff0c\u53d1\u5e03\u65b0\u8bba\u6587\u6458\u8981\u7684\u63a8\u6587\uff0c\u8ba9\u7814\u7a76\u4eba\u5458\u548c\u4eba\u5de5\u667a\u80fd\u7231\u597d\u8005\u5feb\u901f\u4e86\u89e3\u6700\u6709\u8da3\u7684\u79d1\u5b66\u6210\u5c31\u6216\u53d1\u73b0\u3002<\/p>\n

1\u3001\u603b\u4f53\u8bbe\u8ba1<\/h3>\n

\u5148\u611f\u53d7\u4e00\u4e0b\u673a\u5668\u4eba\u53d1\u9001\u7684\u6f14\u793a\u63a8\u6587\uff1a<\/p>\n

\n

\"image.png\"<\/figure>\n<\/p>\n

Twitter \u673a\u5668\u4eba\u5728\u529f\u80fd\u4e0a\u79f0\u4e3a arXiv \u8bba\u6587\u6458\u8981\u5668<\/code>\uff0c\u7531\u51e0\u4e2a\u65e0\u7f1d\u534f\u4f5c\u7684\u5173\u952e\u7ec4\u4ef6\u7ec4\u6210\u3002\u8fd9\u4e9b\u7ec4\u4ef6\u5305\u62ec\u7528\u4e8e\u68c0\u7d22\u65b0\u8bba\u6587\u53d1\u5e03\u7684 arXiv API<\/code>\u3001\u7528\u4e8e\u6458\u8981\u7684 LLMChain<\/code> \u4ee5\u53ca\u5e26\u6709\u81ea\u5b9a\u4e49\u63a8\u6587\u5de5\u5177\u7684 Langchain<\/code> \u4ee3\u7406\uff0c\u7528\u4e8e\u6267\u884c\u6574\u4e2a\u53d1\u5e03\u8fc7\u7a0b\u3002<\/p>\n

\u4e3a\u4e86\u66f4\u597d\u5730\u7406\u89e3\u8be5\u673a\u5668\u4eba\u5185\u7684\u4fe1\u606f\u6d41\u548c\u64cd\u4f5c\uff0c\u4e0b\u9762\u662f\u6574\u4f53\u7684\u67b6\u6784\u56fe\uff1a<\/p>\n

\n

\"image.png\"<\/figure>\n<\/p>\n

\u5728\u6b64\u6f14\u793a\u5e94\u7528\u7a0b\u5e8f\u4e2d\uff0c\u5c06\u8bed\u8a00\u6a21\u578b\u5728\u4e0d\u540c\u6b65\u9aa4\u4e2d\u8fd0\u884c\u4e24\u6b21\u3002\u7b2c\u4e00\u6b21\u8fd0\u884c\u662f\u4f7f\u7528\u666e\u901a\u94fe\u6839\u636e\u8bba\u6587\u6458\u8981\u751f\u6210\u6458\u8981\uff0c\u7b2c\u4e8c\u6b21\u8fd0\u884c\u662f\u8981\u6c42\u4ee3\u7406\u6267\u884c\u63a8\u6587\u5e16\u5b50\u3002\u663e\u7136\uff0c\u7b2c\u4e8c\u6b21\u8fd0\u884c\u53ef\u4ee5\u7528\u6ca1\u6709\u4efb\u4f55 AI \u53c2\u4e0e\u7684\u7eaf\u7a0b\u5e8f\u4ee3\u7801\u4ee3\u66ff\uff0c\u4f46\u662f\uff0c\u4e3a\u4e86\u6f14\u793a\u5982\u4f55\u6700\u597d\u5730\u5229\u7528\u751f\u6210\u5f0f AI \u6765\u5b8c\u6210\u672a\u6765\u66f4\u590d\u6742\u7684\u4efb\u52a1\uff0c\u503c\u5f97\u5c06\u6b64\u4efb\u52a1\u5305\u88c5\u5230 Langchain \u4ee3\u7406\u4e2d\u3002<\/p>\n

2\u3001\u4ee3\u7801\u6a21\u5757<\/h3>\n

\u73b0\u5728\uff0c\u5c31\u6765\u9010\u6b65\u63a2\u7d22\u4f7f\u8fd9\u4e2a\u63a8\u6587\u673a\u5668\u4eba\u7684\u4ee3\u7801\u5de5\u4f5c\u6d41\u7a0b\u3002\u5c06\u6df1\u5165\u7814\u7a76\u6bcf\u4e2a\u6b65\u9aa4\uff0c\u63d0\u4f9b\u4ee3\u7801\u7247\u6bb5\u548c\u89e3\u91ca\uff0c\u4ee5\u786e\u4fdd\u987a\u5229\u5b9e\u65bd\u3002<\/p>\n

arXiv API<\/h4>\n

arXiv \u662f\u5eb7\u5948\u5c14\u5927\u5b66\u56fe\u4e66\u9986\u7684\u4e00\u4e2a\u6d41\u884c\u7684\u5728\u7ebf\u5b58\u50a8\u5e93\uff0c\u5b58\u50a8\u7269\u7406\u3001\u6570\u5b66\u3001\u8ba1\u7b97\u673a\u79d1\u5b66\u7b49\u5404\u4e2a\u9886\u57df\u7684\u79d1\u5b66\u8bba\u6587\u3002\u5b83\u4e3a\u7814\u7a76\u4eba\u5458\u5728\u6b63\u5f0f\u540c\u884c\u8bc4\u5ba1\u548c\u5728\u4f20\u7edf\u79d1\u5b66\u671f\u520a\u4e0a\u53d1\u8868\u4e4b\u524d\u5206\u4eab\u4ed6\u4eec\u7684\u5de5\u4f5c\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5e73\u53f0\u3002arXiv \u4e0a\u6258\u7ba1\u7684\u8bba\u6587\u53ef\u4ee5\u514d\u8d39\u5411\u516c\u4f17\u5f00\u653e\uff0c\u8ba9\u7814\u7a76\u4eba\u5458\u548c\u6280\u672f\u7231\u597d\u8005\u80fd\u591f\u53ca\u65f6\u4e86\u89e3\u6700\u65b0\u7684\u7814\u7a76\u8fdb\u5c55\u3002<\/p>\n

arXiv \u8fd8\u63d0\u4f9b\u5bf9\u5176\u8d44\u6e90\u7684 API \u8bbf\u95ee\uff0c\u53ef\u4ee5\u901a\u8fc7 Python \u8c03\u7528API\u5b9a\u671f\u4e0b\u8f7d\u6240\u9700\u7684\u8bba\u6587\u53ca\u5176\u5173\u952e\u6570\u636e\u3002<\/p>\n

\u5b89\u88c5\u8f6f\u4ef6\u5305\uff1a<\/p>\n

<\/div>
<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>pip install arxiv<\/span>\n<\/code><\/pre>\n
\u641c\u7d22\u5173\u4e8e language model<\/code> \u7684\u6700\u65b0\u8bba\u6587\u3002<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>import arxiv<\/span>\ndef arxiv_search():<\/span>\n    summary<\/span> = ''<\/span><\/span>\n    search<\/span> = arxiv.Search(<\/span>\n      query<\/span> = \"language model\"<\/span>,<\/span>\n      max_results<\/span> = 1<\/span>,<\/span>\n      sort_by<\/span> = arxiv.SortCriterion.SubmittedDate<\/span>\n    )<\/span>\n    for result in search.results():<\/span>\n        summary<\/span> = f\"Title: {result.title}, Abstract: {result.summary}, URL: {result.entry_id}\"<\/span><\/span>\n<\/span>\n    return summary<\/span>\n<\/code><\/pre>\n
\u4ee3\u7801\u901a\u8fc7 API \u5728 arXiv \u6570\u636e\u5e93\u4e0a\u6267\u884c search()<\/code> \u68c0\u7d22\u4e0e\u67e5\u8be2\u5339\u914d\u5173\u4e8e language model<\/code> \u7684\u6700\u65b0\u8bba\u6587\uff0c\u7ed3\u679c\u6309\u63d0\u4ea4\u65e5\u671f\u6392\u5e8f\u3002<\/p>\n
\u7136\u540e\u8fed\u4ee3\u641c\u7d22\u7ed3\u679c\uff0c\u5bf9\u4e8e\u6bcf\u4e2a\u7ed3\u679c\uff0c\u5b83\u5c06\u8bba\u6587\u7684\u6807\u9898\u3001\u6458\u8981\u548c URL \u9644\u52a0\u5230\u5c06\u7528\u4e8e\u540e\u7eed\u6587\u672c\u6458\u8981\u6b65\u9aa4\u7684 summary<\/code> \u5b57\u7b26\u4e32\u3002<\/p>\n
\n
\u6709\u5173 arXiv API<\/a> \u7684\u66f4\u8be6\u7ec6\u7528\u6cd5\uff0c\u53ef\u4ee5\u5728\u8fd9\u91cc\u627e\u5230\u5b83\u4eec\u3002<\/p>\n<\/blockquote>\n
\u4f7f\u7528 LLMChain \u8fdb\u884c\u6458\u8981<\/h4>\n
\u73b0\u5728\uff0c\u5c06\u5b9e\u73b0\u4e00\u4e2a\u7b80\u5355\u7684 LLM \u94fe\u6765\u81ea\u52a8\u5904\u7406\u52a8\u6001\u63d0\u793a\uff0c\u8981\u6c42 gpt-3.5-turbo<\/code> \u6a21\u578b\u5728 arXiv<\/code> \u62bd\u8c61\u4e0a\u4e0b\u6587\u4e0a\u751f\u6210\u6458\u8981\u3002<\/p>\n
\u5b89\u88c5\u8f6f\u4ef6\u5305\uff1a<\/p>\n
<\/div>
<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>pip install langchain openai<\/span>\n<\/code><\/pre>\n
\u5c06 OpenAI API \u5bc6\u94a5\u8bbe\u7f6e\u4e3a\u73af\u5883\u3002<\/p>\n
<\/div>
lua<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>os<\/span>.environ[\"OPENAI_API_KEY\"<\/span>] = \"{Your_API_Key}\"<\/span><\/span>\n<\/code><\/pre>\n
\u521b\u5efa\u4e00\u4e2a\u597d\u7684\u63d0\u793a\u6a21\u677f\uff08\u59cb\u7ec8\u662f\u6700\u91cd\u8981\u7684\uff09\u3002<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>prompt_tweet_template<\/span> = \"\"\"<\/span><\/span>\nYou are a AI technology blogger who write regular tweets for shortly summarizing new papers on Arxiv about language<\/span>\nmodel topics. You write summary for tweets based on text with content of Title, Abstract, URL: ##start {text} ##end<\/span>\nYour write a 50-word quick summary including key information from the Title and Abstract, then, attach the URL after summary.<\/span>\nSummary: <\/span>\n\"\"\"<\/span>\n<\/code><\/pre>\n
\u4f7f\u7528\u6a21\u578b gpt-3.5-turbo<\/code> \u548c\u63d0\u793a\u6a21\u677f PROMPT_TWEET<\/code> \u521b\u5efa LLMChain<\/code>\u3002<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>from langchain import PromptTemplate, LLMChain<\/span>\nfrom langchain.chat_models import ChatOpenAI<\/span>\n<\/span>\nPROMPT_TWEET<\/span> = PromptTemplate(template=prompt_tweet_template, input_variables=[\"text\"<\/span>])<\/span>\nllm<\/span> = ChatOpenAI(temperature=0.5<\/span>)<\/span>\ntweet_chain<\/span> = LLMChain(llm=llm, prompt=PROMPT_TWEET,verbose=True<\/span>)<\/span>\n<\/code><\/pre>\n
\u4ece arxiv_search()<\/code> \u63d0\u4f9b\u6700\u65b0 arXiv \u8bba\u6587\u7684\u6458\u8981\u5b57\u7b26\u4e32\u3002<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>new_paper<\/span> = arxiv_search()<\/span>\ntweet_body<\/span> = tweet_chain.run(new_paper)<\/span>\n<\/code><\/pre>\n
\u73b0\u5728\u6709\u4e86\u4e00\u6761\u4e0d\u9519\u7684\u63a8\u6587\uff0c\u5176\u4e2d\u5305\u542b arXiv<\/code> \u8bba\u6587\u6458\u8981 tweet_body<\/code> \uff0c\u63a5\u4e0b\u6765\u5c31\u51c6\u5907\u53d1\u9001\u3002<\/p>\n
Twitter API<\/h4>\n
\u8981\u83b7\u5f97 Twitter API \u7684\u8bb8\u53ef\uff0c\u5fc5\u987b\u767b\u5f55 Twitter \u5f00\u53d1\u8005\u5e73\u53f0\u521b\u5efa\u4e00\u4e2a\u9879\u76ee\u53ca\u5176\u5e94\u7528\u7a0b\u5e8f\uff0c\u4ee5\u65b9\u4fbf\u673a\u5668\u4eba\u7a0d\u540e\u8c03\u7528\u5b83\u4eec\u7684 API\u3002<\/p>\n
\u5f00\u53d1\u8005\u5e73\u53f0\u7684 URL\uff1adeveloper.twitter.com\/en\/portal\/<\/a>\u3002<\/p>\n
\n
\"image.png\"<\/figure>\n<\/p>\n
\u5982\u679c\u5f00\u53d1\u8005\u514d\u8d39\u8d26\u6237\uff0c\u5219\u53ea\u80fd\u521b\u5efa\u5305\u542b\u4e00\u4e2a\u5e94\u7528\u7a0b\u5e8f\u7684\u9879\u76ee\u3002\u91cd\u547d\u540d\u9879\u76ee\u5e94\u7528\u7a0b\u5e8f\u540e\uff0c\u627e\u5230\u94a5\u5319<\/code>\u56fe\u6807\uff0c\u7136\u540e\u5355\u51fb\uff0c\u5c06\u91cd\u5b9a\u5411\u5230\u8be5\u5e94\u7528\u7a0b\u5e8f\u7684Keys and tokens<\/code>\u9875\u9762\u3002<\/p>\n
\n
\"image.png\"<\/figure>\n<\/p>\n
\u5355\u51fb Generate<\/code> \u6216 Regenerate<\/code> \u6309\u94ae\u751f\u6210\u5bc6\u94a5\uff0c\u7136\u540e\u5c06 consumer_key<\/code> \u548c consumer_secret<\/code> \u590d\u5236\u5230\u5c06\u5728\u540e\u7eed\u6b65\u9aa4\u4e2d\u4f7f\u7528\u7684\u5b89\u5168\u4f4d\u7f6e\u3002<\/p>\n
\u4ee5\u4e0b\u4ee3\u7801\u6f14\u793a\u4e86\u5982\u4f55\u901a\u8fc7\u5bc6\u94a5\u548c secret<\/code> \u6388\u6743\u83b7\u5f97\u5bf9 Twitter API \u7684\u5b8c\u5168\u8bbf\u95ee\u6743\u9650\u3002<\/p>\n
\u5b89\u88c5\u8f6f\u4ef6\u5305<\/p>\n
<\/div>
<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>pip install requests requests_oauthlib<\/span>\n<\/code><\/pre>\n
\u901a\u8fc7\u6388\u6743\u6765\u521d\u59cb\u5316 Twitter \u4f1a\u8bdd\u3002<\/p>\n
<\/div>
ini<\/span>
\u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>def tweet_init():<\/span>\n    consumer_key<\/span> = \"{Your_consumer_key}\"<\/span><\/span>\n    consumer_secret<\/span> = \"{Your_consumer_secret}\"<\/span><\/span>\n<\/span>\n    # Get request token<\/span><\/span>\n    request_token_url<\/span> = \"https:\/\/api.twitter.com\/oauth\/request_token?oauth_callback=oob&x_auth_access_type=write\"<\/span><\/span>\n    oauth<\/span> = OAuth1Session(consumer_key, client_secret=consumer_secret)<\/span>\n    print(oauth)<\/span>\n    try:<\/span>\n        fetch_response<\/span> = oauth.fetch_request_token(request_token_url)<\/span>\n    except ValueError:<\/span>\n        print(<\/span>\n            \"There may have been an issue with the consumer_key or consumer_secret you entered.\"<\/span>\n        )<\/span>\n<\/span>\n    resource_owner_key<\/span> = fetch_response.get(\"oauth_token\"<\/span>)<\/span>\n    resource_owner_secret<\/span> = fetch_response.get(\"oauth_token_secret\"<\/span>)<\/span>\n    print(\"Got OAuth token: %s\" % resource_owner_key)<\/span>\n<\/span>\n    # Get authorization<\/span><\/span>\n    base_authorization_url<\/span> = \"https:\/\/api.twitter.com\/oauth\/authorize\"<\/span><\/span>\n    authorization_url<\/span> = oauth.authorization_url(base_authorization_url)<\/span>\n    print(\"Please go here and authorize: %s\" % authorization_url)<\/span>\n    verifier<\/span> = input(\"Paste the PIN here: \"<\/span>)<\/span>\n<\/span>\n    # Get the access token<\/span><\/span>\n    access_token_url<\/span> = \"https:\/\/api.twitter.com\/oauth\/access_token\"<\/span><\/span>\n    oauth<\/span> = OAuth1Session(<\/span>\n        consumer_key,<\/span>\n        client_secret<\/span>=consumer_secret,<\/span>\n        resource_owner_key<\/span>=resource_owner_key,<\/span>\n        resource_owner_secret<\/span>=resource_owner_secret,<\/span>\n        verifier<\/span>=verifier,<\/span>\n    )<\/span>\n    oauth_tokens<\/span> = oauth.fetch_access_token(access_token_url)<\/span>\n<\/span>\n    access_token<\/span> = oauth_tokens[\"oauth_token\"<\/span>]<\/span>\n    access_token_secret<\/span> = oauth_tokens[\"oauth_token_secret\"<\/span>]<\/span>\n<\/span>\n    oauth<\/span> = OAuth1Session(<\/span>\n        consumer_key,<\/span>\n        client_secret<\/span>=consumer_secret,<\/span>\n        resource_owner_key<\/span>=access_token,<\/span>\n        resource_owner_secret<\/span>=access_token_secret,<\/span>\n    )<\/span>\n    return oauth<\/span>\n<\/span>\nmytweet<\/span> = tweet_init()<\/span>\n<\/code><\/pre>\n
\u6b64\u4ee3\u7801\u7247\u6bb5\u8bbe\u7f6e\u4f7f\u7528 OAuth 1.0<\/code> \u8bbf\u95ee Twitter API \u7684\u8eab\u4efd\u9a8c\u8bc1\u8fc7\u7a0b\u3002\u5b83\u6d89\u53ca\u83b7\u53d6\u6388\u6743\u6240\u9700\u4ee4\u724c\u7684\u51e0\u4e2a\u6b65\u9aa4\u3002<\/p>\n
    \n
  • \u9996\u5148\u4f7f\u7528 fetch_request_token()<\/code> \u65b9\u6cd5\u4ece Twitter \u7684 API \u8bf7\u6c42\u4e34\u65f6\u8bf7\u6c42\u4ee4\u724c resources_owner<\/code> token<\/code>\u548csecret<\/code>\u3002<\/li>\n
  • \u751f\u6210\u4e00\u4e2a\u4e34\u65f6 URL\uff0c\u7528\u4e8e\u63d0\u4f9b PIN \u7801\uff0c\u5e0c\u671b\u7528\u6237\u5728\u7ec8\u7aef\u63d0\u793a\u4e2d\u8f93\u5165\u5982\u4e0b\u6240\u793a\u7684\u5185\u5bb9\uff1a<\/li>\n<\/ul>\n
    \n
    \u8bf7\u524d\u5f80\u6b64\u5904\u6388\u6743\uff1ahttps:\/\/api.twitter.com\/oauth\/authorize?oauth_token=jvPdHgAAAAABofwiAAABiRDFU_M<\/code> \u5c06 PIN<\/code> \u7801\u7c98\u8d34\u5230\u6b64\u5904\uff1a{Input the PIN display on above link}<\/code><\/p>\n<\/blockquote>\n
      \n
    • \u4f7f\u7528 PIN \u7801\u4f5c\u4e3a verifier<\/code>\uff0c\u4e0e\u5176\u4ed6\u4e24\u4e2a token\/secret <\/code> \u5bf9\u4e00\u8d77\u8bf7\u6c42\u4e00\u5bf9\u8bbf\u95ee token\/secret <\/code>\u3002<\/li>\n
    • \u6700\u540e\uff0c\u4f7f\u7528\u83b7\u5f97\u7684 access token<\/code> \u548c secret<\/code> \u91cd\u65b0\u914d\u7f6e OAuth1Session<\/code> \u5b9e\u4f8b\u3002\u6b64\u91cd\u65b0\u914d\u7f6e\u5141\u8bb8\u4f7f\u7528 Twitter \u5e10\u6237\u5bf9 Twitter API \u7684\u540e\u7eed\u8bf7\u6c42\u8fdb\u884c\u8eab\u4efd\u9a8c\u8bc1\u3002<\/li>\n<\/ul>\n
      \u5728\u5e94\u7528\u7a0b\u5e8f\u88ab\u6388\u4e88\u5bf9 API \u7684\u8bbf\u95ee\u6743\u9650\u540e\uff0c\u53ea\u9700\u7f16\u5199\u4e00\u4e2a\u51fd\u6570 tweet_send()<\/code> \u5373\u53ef\u901a\u8fc7\u6388\u6743\u5b9e\u4f8b mytweet<\/code> \u53d1\u5e03\u5305\u542b\u6709\u6548\u8d1f\u8f7d\u5185\u5bb9\u7684\u63a8\u6587\u3002<\/p>\n
      <\/div>
      python<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>def<\/span> tweet_send<\/span>(payload<\/span>):<\/span>\n    <\/span>\n    payload_json = {\"text\"<\/span>: payload}<\/span>\n    # Making the request<\/span><\/span>\n    response = mytweet.post(<\/span>\n        \"https:\/\/api.twitter.com\/2\/tweets\"<\/span>,<\/span>\n        json=payload_json,<\/span>\n    )<\/span>\n<\/span>\n    if<\/span> response.status_code != 201<\/span>:<\/span>\n        raise<\/span> Exception(<\/span>\n            \"Request returned an error: {} {}\"<\/span>.format<\/span>(response.status_code, response.text)<\/span>\n        )<\/span>\n<\/span>\n    return<\/span>  response.status_code<\/span>\n<\/code><\/pre>\n
      Langchain \u4ee3\u7406\u521b\u5efa<\/h4>\n
      \u6700\u540e\u4e00\u6b65\u662f\u521b\u5efa\u4e00\u4e2a\u8fd0\u884c ReAct \u6765\u53d1\u5e03\u63a8\u6587\u7684 Langchain \u4ee3\u7406\u3002<\/p>\n
      \u5bfc\u5165\u5fc5\u8981\u7684\u6a21\u5757\u3002<\/p>\n
      <\/div>
      python<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>from<\/span> langchain.agents import<\/span> AgentType, initialize_agent<\/span>\nfrom<\/span> langchain.tools import<\/span> BaseTool, StructuredTool, Tool, tool<\/span>\nfrom<\/span> pydantic import<\/span> BaseModel, Field<\/span>\nfrom<\/span> typing import<\/span> Optional<\/span>, Type<\/span><\/span>\n<\/span>\nfrom<\/span> langchain.callbacks.manager import<\/span> (<\/span>\n    AsyncCallbackManagerForToolRun,<\/span>\n    CallbackManagerForToolRun,<\/span>\n)<\/span>\n<\/code><\/pre>\n
      \u521b\u5efa\u4ece BaseTool \u6d3e\u751f\u7684\u540d\u4e3a TweetSendTool<\/code> \u7684\u5de5\u5177\u6765\u8c03\u7528\u5916\u90e8\u51fd\u6570 tweet_send(payload)<\/code>\u3002<\/p>\n
      <\/div>
      python<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>class<\/span> TweetSendInput<\/span>(BaseModel<\/span>):<\/span>\n    payload: str<\/span> = Field(..., description=\"This is the content of tweet to be sent.\"<\/span>)<\/span>\n<\/span>\nclass<\/span> TweetSendTool<\/span>(BaseTool<\/span>):<\/span>\n    name = \"Tweet Send\"<\/span><\/span>\n    description = \"useful for when you need to send a tweet\"<\/span><\/span>\n<\/span>\n    def<\/span> _run<\/span>(<\/span><\/span>\n        self, payload: str<\/span>, run_manager: Optional<\/span>[CallbackManagerForToolRun] = None<\/span><\/span>\n    ) -> str<\/span>:<\/span>\n        \"\"\"Use the tool.\"\"\"<\/span><\/span>\n        return<\/span> tweet_send(payload)<\/span>\n<\/span>\n    async<\/span> def<\/span> _arun<\/span>(<\/span><\/span>\n        self, payload: str<\/span>, run_manager: Optional<\/span>[AsyncCallbackManagerForToolRun] = None<\/span><\/span>\n    ) -> str<\/span>:<\/span>\n        \"\"\"Use the tool asynchronously.\"\"\"<\/span><\/span>\n        raise<\/span> NotImplementedError(\"Tweet send does not support async\"<\/span>)<\/span>\n<\/span>\n    args_schema: Type<\/span>[BaseModel] = TweetSendInput<\/span>\n<\/code><\/pre>\n
      \u4ece\u5728\u524d\u9762\u7684\u6b65\u9aa4\u4e2d\u521b\u5efa\u7684\u73b0\u6709\u5de5\u5177\u548c llm \u521b\u5efa\u4ee3\u7406\u3002<\/p>\n
      <\/div>
      ini<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>tools<\/span> = [TweetSendTool()]<\/span>\nagent<\/span> = initialize_agent(<\/span>\n    tools, llm, agent<\/span>=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True<\/span><\/span>\n)<\/span>\n<\/code><\/pre>\n
      \u6784\u5efa\u4e00\u4e2a\u50cf\u6837\u7684\u63d0\u793a\u6a21\u677f\uff0c\u7528\u4e8e\u4ece\u5907\u7528 arXiv \u6458\u8981 tweet_body<\/code> \u5230\u63a8\u6587\u53d1\u5e03\u65b9\u6cd5\u7684\u8fde\u63a5\u3002<\/p>\n
      <\/div>
      ini<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>instructions<\/span> = f\"\"\"<\/span><\/span>\n    Post a tweet with the summary of Arxiv paper, use the common Twitter expression style, and with '#AI #arXiv', and provide URL at the end.<\/span>\n    The summary: {tweet_body}<\/span>\n\"\"\"<\/span>\n<\/code><\/pre>\n
      \u8fd0\u884c\u4ee3\u7406\u3002<\/p>\n
      <\/div>
      scss<\/span>
      \u590d\u5236\u4ee3\u7801<\/div><\/div><\/div>agent.run<\/span>(instructions)<\/span>\n<\/code><\/pre>\n
      \u6bcf\u6b21\u4ee3\u7406\u8fd0\u884c\u65f6\uff0cTwitter \u4e0a\u90fd\u4f1a\u53d1\u5e03\u4e00\u6761\u63a8\u6587\uff0c\u5176\u4e2d\u603b\u7ed3\u4e86\u6700\u65b0 arXiv \u8bba\u6587\u4e2d\u6709\u5173\u8bed\u8a00\u6a21\u578b\u4e3b\u9898\u7684\u5173\u952e\u4fe1\u606f\u3002\u4e0d\u8db3\u7684\u662f\u6bcf\u6b21\u90fd\u9700\u8981\u6388\u6743 PIN \u7801\uff0c\u514d\u8d39\u7684 Twitter API \u6ca1\u529e\u6cd5\u5b9e\u73b0\u81ea\u52a8\u6267\u884c\u53d1\u5e03\u3002<\/p>","protected":false},"excerpt":{"rendered":"
      \u8ddf\u4e0a AI \u7684\u6f6e\u6d41\uff0c\u901a\u8fc7\u4f7f\u7528\u5927\u578b\u8bed\u8a00\u6a21\u578b\u53ca\u5176\u652f\u6301\u5e94\u7528\u7a0b\u5e8f\u6846\u67b6\uff08\u5982 Langchain\uff09\uff0c\u53ef\u4ee5\u5b9e\u73b0\u81ea\u52a8\u8ddf\u8e2a arXiv \u8bba\u6587\u7684\u65b0\u7248\u672c\u3002\u5728\u672c\u6559\u7a0b\u4e2d\uff0c\u5c06\u4ecb\u7ecd\u5982\u4f55\u8bbe\u8ba1\u4e00\u4e2a Twitter \u673a\u5668\u4eba\uff0c\u53d1\u5e03\u65b0\u8bba\u6587\u6458<\/p>\n","protected":false},"author":1,"featured_media":2962,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"rank_math_title":"AI\u8bba\u6587\u6458\u8981\u673a\u5668\u4eba\uff1aLangchain\u6253\u9020Twitter\u5b66\u672f\u52a9\u624b - \u5948\u65af\u79d1\u6280\u793e\u533a","rank_math_description":"\u672c\u6587\u4ecb\u7ecd\u5982\u4f55\u4f7f\u7528 Langchain \u548c arXiv API \u6784\u5efa\u4e00\u4e2a Twitter \u673a\u5668\u4eba\uff0c\u81ea\u52a8\u53d1\u5e03\u6700\u65b0 AI \u8bba\u6587\u6458\u8981\u7684\u63a8\u6587\uff0c\u5e2e\u52a9\u7814\u7a76\u4eba\u5458\u548c\u7231\u597d\u8005\u8ffd\u8e2a\u79d1\u6280\u524d\u6cbf\u52a8\u6001\u3002","rank_math_focus_keyword":"Langchain, arXiv, \u63a8\u6587\u673a\u5668\u4eba, AI\u8bba\u6587\u6458\u8981, \u81ea\u7136\u8bed\u8a00\u5904\u7406","views":"23","footnotes":""},"categories":[3],"tags":[126,127,128,129,136],"collection":[],"class_list":["post-1257","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-fenlei2","tag-gpt","tag-ai","tag-128","tag-129","tag-136"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts\/1257","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/comments?post=1257"}],"version-history":[{"count":0,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/posts\/1257\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/media\/2962"}],"wp:attachment":[{"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/media?parent=1257"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/categories?post=1257"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/tags?post=1257"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.nicekj.com\/nicekj2024\/wp\/v2\/collection?post=1257"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}