Skip to content

Report an empty model response as itself, not as invalid JSON #272

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mostafaseyedan wants to merge 2 commits into
base: main
Choose a base branch
from
+13 −4
Open

Report an empty model response as itself, not as invalid JSON #272

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
33cba43
Report an empty model response as itself, not as invalid JSON
mostafaseyedan Sep 21, 2026
08a3acd
Merge branch 'main' into fix/llm-empty-response
mostafaseyedan Oct 21, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 13 additions & 4 deletions src/webview/panel-llm.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -345,6 +345,8 @@ async function selectModel(): Promise<vscode.LanguageModelChat> {
if (models.length > 0) return models[0];
}
const any = await vscode.lm.selectChatModels({});
runtimeDebug('panel-llm', 'model-catalog',
any.map(m => `${m.vendor}/${m.family}/${m.id} maxIn=${m.maxInputTokens}`).join(' ; ') || '(empty)');
if (any.length > 0) return any[0];
throw new Error('No language model available. Make sure GitHub Copilot is installed and signed in.');
}
Expand Down Expand Up @@ -411,6 +413,7 @@ export async function callLlmJson<T>(messages: vscode.LanguageModelChatMessage[]

let lastError: unknown;
let parseFailures = 0;
let emptyResponses = 0;
const retryMessages = [...redactMessages(messages)];

for (let attempt = 0; attempt <= LLM_MAX_RETRIES; attempt++) {
Expand All @@ -419,6 +422,7 @@ export async function callLlmJson<T>(messages: vscode.LanguageModelChatMessage[]
try {
const response = await model.sendRequest(retryMessages, options, cts.token);
for await (const chunk of response.text) text += chunk;
if (text.trim().length === 0) throw new Error('Model returned an empty response');
try {
return JSON.parse(text.trim()) as T;
} catch {
Expand All @@ -437,7 +441,9 @@ export async function callLlmJson<T>(messages: vscode.LanguageModelChatMessage[]
options.modelOptions = undefined;
}
// On parse failures, nudge the model to return valid JSON on the next attempt
if (lastError instanceof Error && /JSON|parse/i.test(lastError.message)) {
if (lastError instanceof Error && /empty response/i.test(lastError.message)) {
emptyResponses++;
} else if (lastError instanceof Error && /JSON|parse/i.test(lastError.message)) {
parseFailures++;
if (retryMessages.length === messages.length) {
retryMessages.push(vscode.LanguageModelChatMessage.User(
Expand All @@ -450,8 +456,11 @@ export async function callLlmJson<T>(messages: vscode.LanguageModelChatMessage[]
}
}

const label = parseFailures > 0
? `LLM returned invalid JSON after ${LLM_MAX_RETRIES + 1} attempts. Please try again.`
: (lastError instanceof Error ? lastError.message : 'LLM request failed after retries');
const label = emptyResponses > 0
? `The language model (${model.id}) returned an empty response ${emptyResponses} time(s). ` +
'This usually means the request quota is exhausted or the model declined the request.'
: parseFailures > 0
? `LLM returned invalid JSON after ${LLM_MAX_RETRIES + 1} attempts. Please try again.`
: (lastError instanceof Error ? lastError.message : 'LLM request failed after retries');
throw new Error(label);
}