Agent Generates Tweets But Never Publishes to Twitter

[Clinksys]

I’m using an AI agent that is supposed to post tweets automatically, but it only generates text and logs “Saving Tweet” without ever actually publishing. I’ve verified that TWITTER_DRY_RUN=false in the .env, and the Twitter login appears to succeed, yet there’s no sign of an actual API call to post the tweet (no “Posting…” log, and nothing appears on the timeline). I’ve tried adjusting intervals (e.g., TWITTER_POLL_INTERVAL) and verified that the generated text is valid, but the agent remains stuck in a “generate/store” loop without calling the Twitter API. Any guidance on diagnosing why it never proceeds to post (or how to confirm the agent is truly attempting the Twitter API call) would be greatly appreciated.

[github-actions]

Hello @Clinksys! Welcome to the ai16z community. Thank you for opening your first issue; we appreciate your contribution. You are now a ai16z contributor!

[tcm390]

Hello! Could you kindly share a screenshot of your terminal or try asking in the Discord server for more assistance? Without enough context, it’s a bit challenging to debug your situation. 😊

[tommycp3]

yes, I was having the same issue, I was using anthropic and was getting this error.

["፧ Received response from Anthropic model."] 

Bot would respond to tweet 1877410567498907791 with: BUILD's agent ecosystem aligns perfectly with decentralized development principles. Excited to see their community-driven approach.
Error sending response post: Error: {"errors":[{"message":"NumericString value expected. Received The most interesting and relevant tweet for Ruby to reply to is 1877410567498907791.","extensions":{"name":"MalformedVariablesError","source":"Client","code":366,"kind":"Validation","tracing":{"trace_id":"2fbd23677782b4f2"}},"code":366,"kind":"Validation","name":"MalformedVariablesError","source":"Client","tracing":{"trace_id":"2fbd23677782b4f2"}}]}

and I noticed the core logic was also trying to use the openai api even though I specified Claude so I decided to try out the open ai api and it saved me for the time being working out why it was failing but it does work i with openai api.

So my first question to you is, what model are you using? and can you share the logs?

[AIFlowML]

@tommycp3 pointing out the OAI model issue is due to some internal embedding that are stil running on the OAI directly.
So i guess @Clinksys is probably having that embedding issue.
@tcm390 Can you have a look and check if there is something hardcoded that prevent the post if no OAI model ?

[tcm390]

@tommycp3 pointing out the OAI model issue is due to some internal embedding that are stil running on the OAI directly. So i guess @Clinksys is probably having that embedding issue. @tcm390 Can you have a look and check if there is something hardcoded that prevent the post if no OAI model ?

Yes I will check that

[tcm390]

yes, I was having the same issue, I was using anthropic and was getting this error.

["፧ Received response from Anthropic model."] 

Bot would respond to tweet 1877410567498907791 with: BUILD's agent ecosystem aligns perfectly with decentralized development principles. Excited to see their community-driven approach.
Error sending response post: Error: {"errors":[{"message":"NumericString value expected. Received The most interesting and relevant tweet for Ruby to reply to is 1877410567498907791.","extensions":{"name":"MalformedVariablesError","source":"Client","code":366,"kind":"Validation","tracing":{"trace_id":"2fbd23677782b4f2"}},"code":366,"kind":"Validation","name":"MalformedVariablesError","source":"Client","tracing":{"trace_id":"2fbd23677782b4f2"}}]}

and I noticed the core logic was also trying to use the openai api even though I specified Claude so I decided to try out the open ai api and it saved me for the time being working out why it was failing but it does work i with openai api.

So my first question to you is, what model are you using? and can you share the logs?

Hello @tommycp3, the fix is in this PR: #2430. Please check if it works when you have a moment

[tcm390]

fixed in #2430