What the agent does
The agent clones the repository into its own cloud sandbox and reads the Twitter calls first. In a small Flask app like Smerity/trending_arxiv, main.py builds tweepy.OAuthHandler at line 25 and tweepy.API(auth) at line 28 from keys in config.yaml, then fetch_timeline calls api.user_timeline(screen_name=username, count=200, page=0) at line 178, fetch_search calls api.search('@{} arxiv.org'.format(username), count=200) at line 183, and the /rate_limits route calls api.rate_limit_status() at line 213. add_tweet reads tweet.entities['urls'], tweet.user.id, tweet.user.screen_name, tweet._json and retweeted_status from each result. requirements.txt names tweepy with no version beside flask, Flask-SQLAlchemy, gevent, untangle and pyyaml, and run_dev_server.sh starts it with python -m flask run.
Then it reproduces both states with placeholder keys, which is enough because the break happens before any request is sent. A fresh install resolves Tweepy 4.17.0, and api.search(...) raises AttributeError: 'API' object has no attribute 'search'; search_tweets exists on the same object with the signature (q, **kwargs). On tweepy<4, which resolves 3.10.0, search exists and search_tweets does not. Two more v4 changes show up on the same object: OAuthHandler now prints DeprecationWarning: OAuthHandler is deprecated; use OAuth1UserHandler instead, and the page=0 argument on user_timeline logs WARNING:tweepy.api:Unexpected parameter: page before the call, because v4 no longer takes it. rate_limit_status is unchanged.
The fix is three lines. Line 183 becomes api.search_tweets(q='@{} arxiv.org'.format(username), count=200), with the query passed by keyword as the new signature requires; line 178 drops page=0, since the first page is the default and the name is ignored; line 25 constructs tweepy.OAuth1UserHandler with the same four values. add_tweet needs no change, because the Status objects v4 returns carry the same entities, user and _json attributes the function reads. The agent pins the floor with tweepy>=4 in requirements.txt, since the old method name no longer exists anywhere the new one does.
What the run cannot check, it says. With placeholder keys every call ends in 401 Unauthorized after the warnings, so the pull request shows the attribute and signature checks and the warnings, and leaves the live fetch to the owner's keys. It also notes that whether an account's API tier still allows the v1.1 search endpoint is a question of the account, not the library.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "This Flask app fails with \"AttributeError: 'API' object has no attribute 'search'\" after a tweepy upgrade. Read requirements.txt, config.yaml and every file that calls the tweepy API first. Install the current tweepy and, with placeholder credentials, build the API object the app builds and quote the AttributeError and every deprecation or unexpected-parameter warning the app's calls produce; then install tweepy 3 and show the same calls resolve. Move each removed call to its v4 name with the keyword arguments v4 requires, drop any parameter v4 ignores, replace deprecated auth classes, check that the attributes the app reads from each result still exist, and add the minimum tweepy version to requirements.txt. Say which calls could not be exercised without real credentials. Open a pull request that quotes the failing and passing checks.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
What comes back is a pull request touching main.py and requirements.txt: search renamed to search_tweets with q= on line 183, page=0 removed from line 178, OAuth1UserHandler on line 25, and tweepy>=4 in the requirements. The description quotes the AttributeError on 4.17.0, the Unexpected parameter: page warning and the OAuthHandler deprecation, the search attribute present on 3.10.0, the (q, **kwargs) signature of the replacement, and a line saying the live fetch ended 401 Unauthorized with placeholder keys.
A reviewer checks that the query string is unchanged apart from being passed by keyword, that no other api. call in the repository uses a name v4 removed, and that the fields add_tweet reads still exist on the result objects. The run's record holds each install and each call with its full output, so the version boundary can be read rather than taken on trust.
Run it on every Monday morning
Add an automation to the Project with this prompt and a schedule trigger, cron set to 0 9 * * 1 in your timezone. Each Monday one run installs the current tweepy, exercises the app's calls with placeholder keys and opens one pull request on its own branch when a name moved; a week where nothing changed opens nothing, and a week where a replacement cannot be found ends needs_review with the error attached.
The same body works from your own code with POST /o/{org}/sessions and an API key when a dependency bot's tweepy bump turns the deploy red; send a request_id that is a UUID v5 of the bot's pull request number so a retried start returns the same run, and the key's webhook_url receives run.completed or run.failed with the session URL when it ends.
The use case this job follows, with its automation and trigger: Upgrade dependencies weekly.
Why run it on Rush?
Renames like this one hide behind a network call, so a laptop with real keys finds them one failed request at a time. The agent's sandbox needs no keys for the part that broke: it builds the same objects the app builds, reads their attributes and signatures on the old and new release, and keeps every command and warning in the run's record, so the pull request names each change with the release that made it.
A library that talks to a paid API also raises a question the code cannot answer, which is whether the account still has access to the endpoint. The agent separates the two in the pull request: the renames it proved, and the live call it could not make, so the owner knows exactly what one run with real credentials still has to confirm.
Frequently asked questions
Why was search renamed?
Tweepy 4.0 renamed the v1.1 methods to match the endpoint names, so search became search_tweets, and several names that collided with other endpoints were dropped. The new method takes the query as q and keyword arguments only, which is why the positional string in the old call has to move.
What does the page parameter warning mean?
v4's user_timeline no longer takes page; it logs Unexpected parameter: page and sends the request without it. The call still works, so the agent removes the argument rather than mapping it, and notes that pagination in v4 goes through tweepy.Cursor if more than one page is ever needed.
Will the agent keep OAuthHandler?
It still works in 4.x as an alias, but it prints a DeprecationWarning on every start. The agent replaces it with OAuth1UserHandler, which takes the same consumer and access values, so the config file does not change.
Can it test the real fetch?
Not with placeholder credentials; every call after the attribute checks ends 401 Unauthorized. Put real keys in a Project secret and name the secret in the prompt if you want the run to fetch a timeline and show add_tweet storing it, and remember that access to the v1.1 search endpoint depends on the account's API tier.
More Tweepy jobs
More jobs like this: Upgrade dependencies weekly
- Fix "AttributeError: module 'ast' has no attribute 'Str'" on Python 3.14 (ast)
- Fix "Bad Argument Type. Starting with v5, Only the Object Form Is Allowed" After a TanStack Query Upgrade (TanStack Query)
- Fix the Browserslist caniuse-lite Is Outdated Warning (Browserslist)
- Fix Cannot Find Module @babel/core After a babel-loader Upgrade (babel-loader)
- Fix "Cannot Import Name 'Imputer' from 'sklearn.preprocessing'" After a scikit-learn Upgrade (scikit-learn)