# Value error Conversation roles must alternate **URL:** \ **Category:** General\ **Created:** [August 14, 2025, 1:10pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379 "2025-08-14T13:10:06Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![youlearnit](https://avatars.discourse-cdn.com/v4/letter/y/71c47a/32.png) [@youlearnit](https://discuss.vllm.ai/u/youlearnit)\ **Post date:** [August 14, 2025, 1:10pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/1 "2025-08-14T13:10:06Z")
Hi, Why this could appear? Tried to do multiple prompts same time. Is this application side error? ValueError: Conversation roles must alternate user/assistant/user/assistant/ ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [August 14, 2025, 1:10pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/2 "2025-08-14T13:10:40Z")
This error occurs because vLLM enforces strict alternation between user and assistant roles in the conversation history: after an optional system message, the roles must alternate as user/assistant/user/assistant, etc. If you send two user or two assistant messages in a row, vLLM will raise a ValueError with this message. This is not an application bug, but a requirement of vLLM’s chat template logic for most instruction-tuned models, and is enforced both in offline and OpenAI-compatible serving modes. Some models (like Mistral) are especially strict about this alternation, while others (like Llama) may be more permissive, but vLLM’s default behavior is to require alternation for all models unless a custom chat template is provided or the model’s template allows otherwise. To resolve this, ensure your message list alternates roles correctly. For example: `[system, user, assistant, user, assistant, ...]`. If you need more flexible role ordering, vLLM currently does not support it out-of-the-box, but you can override the chat template with the `--chat-template` CLI option to relax this restriction, though model performance may degrade if the model was not trained for such input patterns. See further details in the linked issues and documentation for examples and workarounds. Would you like more detail or code examples? Sources: - [vLLM Issue: Conversation roles must alternate user/assistant/user/assistant/…](https://github.com/vllm-project/vllm/issues/6862) - [vLLM Issue: Allow flexible message role ordering in conversations](https://github.com/vllm-project/vllm/issues/15486) - [vLLM Generative Models Documentation](https://docs.vllm.ai/en/stable/models/generative_models.html#llm-chat) ---
**Author:** ![hmellor](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/hmellor/32/191_2.png) [@hmellor](https://discuss.vllm.ai/u/hmellor)\ **Post date:** [August 14, 2025, 1:15pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/3 "2025-08-14T13:15:36Z")
@RunLLM please explain how to send multiple prompts using the LLM interface and the vllm serve interface. The user is trying to send multiple prompts but is inadvertantly sending them in the same messages list. ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [August 14, 2025, 1:16pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/4 "2025-08-14T13:16:41Z")
To send multiple prompts with the LLM interface, pass a list of prompt strings to `llm.generate(prompts=[...])` for batch inference. Each prompt should be a separate string in the list, not combined in a single conversation/messages list—this enables vLLM to process them in parallel and return outputs in the same order as the input prompts. For chat-style models, use `llm.chat()` with one conversation per call, not multiple conversations in a single messages list. See [API docs](https://docs.vllm.ai/en/latest/api/vllm/index.html#vllm.LLM.generate). With the vllm serve (OpenAI-compatible) interface, the Completions API supports batched prompts by sending a JSON payload with a “prompt” field as a list of strings. However, the Chat Completions API does not support batching: you must send one conversation (messages list) per request. If you try to send multiple conversations in a single messages list, you’ll get an error. See [vLLM Issue #16965](https://github.com/vllm-project/vllm/issues/16965) and [Quickstart docs](https://docs.vllm.ai/en/latest/getting_started/quickstart.html#openai-chat-completions-api-with-vllm). Would you like a code example for either interface? Sources: - [vLLM API: LLM.generate](https://docs.vllm.ai/en/latest/api/vllm/index.html#vllm.LLM.generate) - [vLLM Issue #16965](https://github.com/vllm-project/vllm/issues/16965) - [Quickstart docs](https://docs.vllm.ai/en/latest/getting_started/quickstart.html#openai-chat-completions-api-with-vllm) ---
**Author:** ![hmellor](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/hmellor/32/191_2.png) [@hmellor](https://discuss.vllm.ai/u/hmellor)\ **Post date:** [August 14, 2025, 1:18pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/5 "2025-08-14T13:18:45Z")
Explain how to send multiple chat style prompts using LLM and how to send parallel requests using the openai API ---
**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\ **Post date:** [August 14, 2025, 1:19pm UTC](https://discuss.vllm.ai/t/value-error-conversation-roles-must-alternate/1379/6 "2025-08-14T13:19:23Z")
To send multiple chat-style prompts using the LLM interface, call `llm.chat()` with a list of conversations, where each conversation is a list of messages (role/content dicts). For example, `llm.chat([conversation1, conversation2, ...])` will process each conversation as a separate prompt and return a list of outputs in the same order. See [LLM.chat API docs](https://docs.vllm.ai/en/latest/api/vllm/index.html#vllm.LLM.chat). For the OpenAI-compatible API (vllm serve), the `/v1/chat/completions` endpoint only supports one conversation per request. To send multiple chat prompts in parallel, you must send multiple HTTP requests concurrently (e.g., using threads or asyncio). Each request should contain a single messages list. See [vLLM Issue #16965](https://github.com/vllm-project/vllm/issues/16965) and [Quickstart docs](https://docs.vllm.ai/en/latest/getting_started/quickstart.html#openai-chat-completions-api-with-vllm). Would you like code examples for either method? Sources: - [LLM.chat API docs](https://docs.vllm.ai/en/latest/api/vllm/index.html#vllm.LLM.chat) - [vLLM Issue #16965](https://github.com/vllm-project/vllm/issues/16965)