Add second check for prompt length before autocomplete

[lwang2070]

Description

There has been reports showing that continue will truncate prefix when sending request to LLM for autocompletion, See #3372 and #6003. The reason is that the renderPrompt function does not strictly follow the contextLength restriction, which causes the prompt to be truncated before sending to LLM.

To fix the issue, I

1. add checks to ensure the final prompt length is less then (contextLength - reservedTokens - safetyBuffer)
2. If exceeded, truncate the prefix and suffix proportion to their length.

and that's it!

Checklist

• I've read the contributing guide
• The relevant docs, if any, have been updated or created
• [] The relevant tests, if any, have been updated or created

My first PR, not sure how to test the code here. Would be happy to update if any guidance are provided :)

---

Summary by cubic

Added a second check to ensure prompt length stays within model limits before sending autocomplete requests, preventing prompt truncation.

• Bug Fixes
  • Truncates prefix and suffix proportionally if the prompt is too long, so the full prompt fits within allowed context.

[netlify]

👷 Deploy request for continuedev pending review.

Visit the deploys page to approve it

Name	Link
🔨 Latest commit	103c8ad

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[lwang2070]

recheck

[lwang2070]

I have read the CLA Document and I hereby sign the CLA

[Patrick-Erichsen]

Passing this to @jpoly1219 who is working on autocomplete things at the moment

[Patrick-Erichsen]

Can we add unit tests for this new function?

[jpoly1219]

The core logic of this is a duplicate of renderPrompt. I think a good next step to take here is to refactor renderPrompt using buildPrompt. Also +1 to adding unit tests.

[jpoly1219]

It looks good overall. I think reducing repetitive logic and adding unit tests would be greatly appreciated!