Bot Generates Multiple Replies Due to DEFAULT_MAX_TWEET_LENGTH

[naiveai-dev]

Eliza Version
latest v0.1.9

Issue Summary
When testing a bot that replies to its followers, I noticed that it sometimes generates multiple replies to a single tweet. The API logs indicate that the bot only makes two calls per response (one for the medium model and one for the large model). However, due to DEFAULT_MAX_TWEET_LENGTH, the generated response is split into multiple tweets to fit within the 280-character limit.

This is usually a minor compliance issue, but in some cases, the agent detects these split responses as incomplete sentences and attempts to generate additional replies repeatedly. As a result, the API sometimes makes 4 to 6 calls instead of 2, leading to excessive LLM token usage and increased costs.

Steps to Reproduce

1. Enable the bot to reply to follower tweets.
2. Observe API logs, which show only two calls being made.
3. Due to DEFAULT_MAX_TWEET_LENGTH, the reply is split into multiple tweets.
4. In some cases, the agent perceives the split content as incomplete and generates additional responses.
5. The bot may end up making 4–6 API calls per tweet, leading to unnecessary token consumption.

Expected Behavior

• The bot should generate only one response per tweet and appropriately split it if needed, without triggering unnecessary additional responses.

Actual Behavior

• The bot sometimes misinterprets split responses as incomplete and keeps generating additional replies, leading to multiple API calls (4–6 instead of 2).

Impact

• High token usage, leading to increased API costs.
• Potential rate limits or compliance issues with excessive replies to a single tweet.

Temporary Fixes

• set DEFAULT_MAX_TWEET_LENGTH to much higher number(like 500)

I think this is mainly from interation.ts and continue.ts.

Screenshots