Empty `messages` payload reaches agent loop and crashes with `IndexError`

[YizukiAme]

AI Disclosure

• This issue was written entirely by a human
• This issue was written with AI assistance (e.g. Copilot, ChatGPT, Claude) and reviewed and edited by a human
• I have read the AI Policy and agree to its terms

Human Verification

I have read the AI policy and I confirm this issue was reviewed by a human.

Describe the bug

LettaRequest.validate_input_or_messages only rejects messages is None; it does not reject messages=[]. All V2/V3 agent entrypoints call _prepare_in_context_messages_no_persist_async, which immediately dereferences input_messages[0].

Sending { "messages": [] } bypasses validation and crashes with IndexError, returning 500 instead of 4xx.

# letta/schemas/letta_request.py
if self.input is None and self.messages is None:
    raise ValueError("Must specify either 'input' or 'messages'.")
# messages=[] passes this check ↑

# letta/agents/helpers.py
if input_messages[0].type == "tool_return":  # ← IndexError

Fix: Reject empty arrays in the validator:

if self.messages is not None and len(self.messages) == 0:
    raise ValueError("'messages' must not be empty.")

How are you running Letta?

From source

Operating System

Linux

Letta Version

latest main

Model

No response

Steps to Reproduce

curl -X POST http://localhost:8283/v1/agents/{agent_id}/messages
-H "Content-Type: application/json"
-d '{"messages": []}'

→ 500 Internal Server Error (IndexError)

Expected: 400/422 Validation Error

Relevant Logs / Screenshots

Additional Context

No response

[SamSi0322]

I'll take this. Missing empty-list validation on messages — allows [] through to code that does input_messages[0]. PR incoming.

[YizukiAme]

ChatGPT spam

[HIJO790401]

I’m not interested in emotional reactions without structure.

If you think my post is incorrect, point to the causal chain, the technical issue, or the exact part that fails. If your reply is only “ChatGPT spam,” then there is nothing to discuss.

I may not write English perfectly, but I do write with structure. So my question is simple: what exactly is your issue?

If there is no concrete argument, no technical point, and no clear objection, then this is just noise, not discussion.

I’ve reported this comment.

[YizukiAme]

I’m not interested in emotional reactions without structure.

If you think my post is incorrect, point to the causal chain, the technical issue, or the exact part that fails. If your reply is only “ChatGPT spam,” then there is nothing to discuss.

I may not write English perfectly, but I do write with structure. So my question is simple: what exactly is your issue?

If there is no concrete argument, no technical point, and no clear objection, then this is just noise, not discussion.

I’ve reported this comment.

Oh, wonderful, yeah, "I'm not interested in emotional reactions without structure."
What a clear and logical reply. What kind of insight, exactly, were you hoping to get from me — the person who filed "Empty messages payload reaches agent loop and crashes with IndexError"?
Lavish praise for the brilliant design of your RAG system?

Do you genuinely believe there is anything to "discuss" about your spectacularly magnificent, exquisitely crafted, breathtakingly transcendent three-thousand-year project under this issue?
Tell me, what was your brilliant "causal chain, technical issue, or the exact part that fails"?
And what "concrete argument, technical point, or clear objection" did you contribute to this issue?
So would you say your own brilliant contribution here is "just noise," or "discussion"?

"I've reported this comment." Wow, terrifying. Nobody cares.

Take a look in the mirror.

[HIJO790401]

Understood. This issue is about a specific bug, and my previous comment was off-topic for that thread.

Your report is about validation behavior for "messages=[]", the resulting "IndexError", and returning 500 instead of a proper 4xx validation error. That is a concrete issue.

I’ve removed myself from the technical path of this bug thread. I won’t continue the off-topic discussion here.