fix(demo): keep answers cited when the free chain falls to a weak model

gemma-4-26b started 502ing upstream, so every /demo/chat answer fell
through to nemotron-nano-12b-vl, which ignores the system-prompt
citation rule entirely: answers rendered with no sources. Pinning that
model with the production prompt reproduced it - zero bracket citations
until the rule is restated next to the question.

- restate the citation rule at the end of the user message; takes
  nemotron-nano from zero to per-claim citations across repeated probes
- add two free vision models that cite reliably (gemma-4-31b,
  nex-n2-pro) to the chain ahead of the nemotron fallback

Author: NorthernLightx

src/config/settings.py

@@ -59,7 +59,10 @@ class Settings(BaseSettings):
     # time as a misconfiguration guard. Env-overridable (RAG_DEMO_MODELS) so
     # the chain can be re-pointed without an image rebuild when the free pool
     # shifts. Models must be vision-capable: the demo client sends page images.
-    demo_models: str = "google/gemma-4-26b-a4b-it:free,nvidia/nemotron-nano-12b-v2-vl:free"
+    demo_models: str = (
+        "google/gemma-4-26b-a4b-it:free,google/gemma-4-31b-it:free,"
+        "nex-agi/nex-n2-pro:free,nvidia/nemotron-nano-12b-v2-vl:free"
+    )
 
     top_k: int = Field(default=5, ge=1)
     rerank_top_k: int = Field(default=50, ge=1)

tests/unit/test_demo_route.py

@@ -119,7 +119,8 @@ def handler(body: dict[str, Any]) -> httpx.Response:
     assert res.status_code == 200
     assert [b["model"] for b in seen] == [
         "google/gemma-4-26b-a4b-it:free",
-        "nvidia/nemotron-nano-12b-v2-vl:free",
+        "google/gemma-4-31b-it:free",
+        "nex-agi/nex-n2-pro:free",
     ]
 
 

web/app/api.js

@@ -328,7 +328,15 @@
         content.push({ type: "image_url", image_url: { url: dataUrl } });
       }
     }
-    content.push({ type: "text", text: `\nQuestion: ${latestUserText}` });
+    // The citation rule lives in the system prompt, but the small free models
+    // the demo chain falls back to drop it there: nemotron-nano-12b emits zero
+    // bracket citations until the rule is restated next to the question.
+    content.push({
+      type: "text",
+      text:
+        `\nQuestion: ${latestUserText}` +
+        "\n(Reminder: every factual claim taken from the context must cite its supporting chunk id in square brackets, e.g. [2604.22753v1::p5::c24] — an answer that states facts without bracket citations is rejected.)",
+    });
     messages.push({ role: "user", content });
     // `injected` rides along so the UI can show this evidence in the panel —
     // it is context the model saw, but it is not a retrieval result.