vllm-project/vllm · error · Error

text request `{request_id}` must contain at least one prompt

Error message

text request `{request_id}` must contain at least one prompt token ID

What it means

Error variant raised in rust/src/text/src/request.rs:228 when a text request supplies prompt_token_ids that is empty. The frontend requires at least one prompt token before it can lower the request to the engine. is_request_validation_error() classifies it as an invalid user request (400-class), so the HTTP layer maps it to a client error, not a server fault.

Source

Thrown at rust/src/text/src/error.rs:16

// SPDX-License-Identifier: Apache-2.0
// SPDX-FileCopyrightText: Copyright contributors to the vLLM project

use thiserror::Error;
use vllm_engine_core_client::Error as EngineCoreError;
use vllm_llm::Error as LlmError;

pub use crate::lower::logprobs::LogprobsError;
pub use crate::lower::sampling::SamplingParamsError;
pub use crate::lower::token_ids::TokenIdsError;

#[derive(Debug, Error)]
pub enum Error {
    #[error("tokenizer error: {0}")]
    Tokenizer(String),
    #[error("text request `{request_id}` must contain at least one prompt token ID")]
    EmptyPromptTokenIds { request_id: String },
    #[error("text request `{request_id}` stop strings cannot be empty")]
    EmptyStopString { request_id: String },
    #[error(
        "this model's maximum context length is {max_model_len} tokens, \
         but the prompt contains {prompt_len} input tokens"
    )]
    PromptTooLong { max_model_len: u32, prompt_len: u32 },
    #[error(transparent)]
    Logprobs(#[from] LogprobsError),
    #[error(transparent)]
    TokenIds(#[from] TokenIdsError),
    #[error(transparent)]
    SamplingParams(#[from] SamplingParamsError),
    #[error(
        "`min_tokens` must be less than or equal to `max_tokens`, \
         got min_tokens={min_tokens}, max_tokens={max_tokens}"
    )]

View on GitHub (pinned to c794754062)

Solutions

1. Send a non-empty prompt, or at least one token ID when using prompt_token_ids.
2. Guard in client code: skip or reject empty inputs before calling the API.
3. If the prompt is legitimately whitespace-only, check what your tokenizer yields and send minimal content instead.

Example fix

// before
request.prompt_token_ids = Some(vec![]);

// after
request.prompt_token_ids = Some(vec![1]);
// or better: reject empty input client-side
if tokens.is_empty() { return Err("empty prompt"); }

if using_token_ids && prompt_token_ids.is_empty() {
    return Err("prompt must contain at least one token");
}

fn prompt_is_sendable(token_ids: &[u32]) -> bool {
    !token_ids.is_empty()
}

match result {
    Err(e @ vllm_text::Error::EmptyPromptTokenIds { .. }) => {
        respond_400(e); // client error: bad request body, not a server fault
    }
    other => other,
}

When it happens

Trigger: POST /v1/completions (or the internal text API) with "prompt_token_ids": [] and no text prompt; a client sending an empty string prompt that tokenizes to zero tokens; a gateway forwarding a tokenized request built from empty input.

Common situations: Programmatic clients building token-id requests from a variable that ended up empty; upstream chunking/splitting logic producing zero-length segments; prompt preprocessing that strips everything (e.g. whitespace-only prompt with a tokenizer that yields no tokens).

Related errors

• this model's maximum context length is {max_model_len} token
• text request `{request_id}` stop strings cannot be empty
• allowed_token_ids should not be empty
• token_id(s) {token_ids:?} in {parameter} are out of vocabula
• generate request `{request_id}` has an empty prompt_token_id

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/73f1298295faa930. Report an issue: GitHub.