BoundaryML/baml · error

No model type supported

Error message

No model type supported

What it means

`render_prompt` selects a chat or completion rendering path based on whether the prompt is compatible with each mode. When a prompt supports neither chat nor completion rendering (`(false, false)`), BAML bails with this terse message instead of rendering.

Source

Thrown at engine/baml-runtime/src/internal/llm_client/traits/mod.rs:327

                    ctx,
                    &chat,
                )
                .await?;
                RenderedPrompt::Chat(chat)
            }
        };

        let mut prompt = match (features.completion, features.chat) {
            (true, false) => {
                let options = self.completion_options(ctx)?;
                prompt.as_completion(&options)
            }
            (false, true) => {
                let options = self.chat_options(ctx)?;
                prompt.as_chat(&options)
            }
            (true, true) => prompt,
            (false, false) => anyhow::bail!("No model type supported"),
        };

        if features.max_one_system_prompt {
            // Do some more fixes.
            if let RenderedPrompt::Chat(chat) = &mut prompt {
                if chat.len() == 1 && chat[0].role == "system" {
                    // If there is only one message and it is a system message, change it to a user message,
                    // because these models always requires a user message.
                    chat[0].role = "user".into();
                } else {
                    // Otherwise, proceed with the existing logic for other messages.
                    chat.iter_mut().skip(1).for_each(|c| {
                        if c.role == "system" {
                            c.role = "user".into();
                        }
                    });
                }
            }

View on GitHub (pinned to bd85ce9dee)

Solutions

1. Ensure the function defines a non-empty prompt block with valid interpolation syntax.
2. Verify the prompt contains at least one chat message or completion text that the provider can render.
3. Fix template syntax errors (unclosed `{{ }}`) that can leave the prompt unrenderable.
4. Regenerate the client and re-test prompt rendering.

Example fix

// before
function F() {
  prompt #""#   // empty prompt
}

// after
function F() {
  prompt #"
    Answer the question: {{ question }}
    {{ ctx.output_format }}
  "#
}

// Ensure the prompt renders to at least one mode before calling
const rendered = fn.prompt.render(testCtx);
if (!rendered.chat.length && !rendered.completion) {
  throw new Error('Prompt is empty; add content so it can render as chat or completion');
}

function isRenderable(p) { return p && ((p.chat && p.chat.length > 0) || !!p.completion); }

try { await b.F(args); } catch (e) { if (String(e).includes('No model type supported')) console.error('Fix empty/malformed prompt block'); }

When it happens

Trigger: A `RenderedPrompt`/prompt AST that implements neither `as_chat` nor `as_completion` — e.g. a malformed prompt definition whose contents can't be classified as either chat messages or a completion string.

Common situations: An empty or invalid prompt block in a .baml function (e.g. a prompt with no renderable content), or programmatic prompt construction that produced an empty prompt, combined with a provider expecting one of the two modes.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

• {e}
• Completion prompts are not supported by this provider
• error while building HTTP request for function {function_nam
• compilation failed: {e:?}
• failed to serialize BAML bytecode: {e}

AI-assisted analysis of BoundaryML/baml@bd85ce9dee (2026-09-12). Data as JSON: /api/errors/5ea6205a7798266b. Report an issue: GitHub.