fix(textGeneration): finalize empty truncated generations

Gate the completion handling on `generated_text != null` instead of truthiness
so a terminal output with empty text is still finalized. A model can hit
finish_reason "length" before emitting any visible content (hidden-reasoning
models, very small budgets); previously that output was skipped (the special
token was ignored) and the truncation warning never fired. Now the truncation
is flagged + logged for the empty case too.

Note: a genuinely empty body still surfaces as a "no output" error downstream
(content equals the initial empty content), which is the appropriate UX; this
change is about observability and a consistent interrupted flag, not changing
that outcome.

Adds an adapter test for the empty length-terminated stream.

Author: gary149

src/lib/server/endpoints/openai/openAIChatToTextGenerationStream.spec.ts

@@ -39,6 +39,21 @@ describe("openAIChatToTextGenerationStream truncation handling", () => {
 		expect(final?.generated_text).toContain("<think>");
 	});
 
+	it("emits a terminal truncated output even when no content was produced", async () => {
+		// Hidden-reasoning models can burn the whole budget without surfacing any tokens,
+		// then terminate with finish_reason "length". The adapter must still yield a terminal
+		// output (generated_text "" rather than null) carrying truncated:true so the pipeline
+		// can finalize/flag it instead of silently dropping the empty completion.
+		const chunks = [{ choices: [{ index: 0, delta: {}, finish_reason: "length" }] }];
+
+		const outputs = await collect(openAIChatToTextGenerationStream(mockStream(chunks)));
+		const final = outputs.find((o) => o.generated_text !== null);
+
+		expect(final).toBeDefined();
+		expect(final?.generated_text).toBe("");
+		expect(final?.truncated).toBe(true);
+	});
+
 	it("does not flag truncated on a normal finish_reason 'stop'", async () => {
 		const chunks = [
 			{ choices: [{ index: 0, delta: { content: "Hello" } }] },

src/lib/server/textGeneration/generate.ts

@@ -86,8 +86,11 @@ export async function* generate(
 				continue;
 			}
 		}
-		// text generation completed
-		if (output.generated_text) {
+		// text generation completed. Use `!= null` (not truthiness) so a terminal output
+		// with empty text is still finalized: a model can hit finish_reason "length" before
+		// emitting any visible content (hidden-reasoning models, tiny budgets), and that
+		// truncation must still be flagged + logged rather than silently dropped.
+		if (output.generated_text != null) {
 			// If an abort happened just before final output, stop here and let
 			// the caller emit an interrupted final answer with partial text.
 			const abortTime = AbortedGenerations.getInstance().getAbortTime(conv._id.toString());