feat: adding cached llm #66
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| h.Write([]byte(prompt)) | ||
| for _, f := range files { | ||
| h.Write([]byte(f.Name)) | ||
| h.Write([]byte(f.MIME)) | ||
| h.Write(f.Data) |
There was a problem hiding this comment.
The cache key for GenerateWithFiles is derived solely from the prompt and file contents (lines 91‑95), so when files is nil or empty the key becomes exactly the same sha256(prompt) used by Generate (line 72). After a plain Generate call caches a response, a subsequent GenerateWithFiles call with no attachments will return that cached plain-response without executing the file-aware path, even though many implementations produce different outputs for multi-line prompts or do extra wrapping when invoked via GenerateWithFiles. This can surface incorrect responses whenever callers sometimes pass files and sometimes don’t for the same prompt string.
Useful? React with 👍 / 👎.
No description provided.