refactor: Remove live RAGAS eval, replace with static benchmark panel

nest_asyncio cannot patch uvloop (used by uvicorn on Render Linux) — live eval
is unfixable without replacing the event loop. Remove ragas dependency entirely.

- Remove POST /api/eval/ragas endpoint and ragas_eval import from routes/eval.py
- Remove ragas from requirements.txt (saves Docker build time)
- Remove runRagasEval from api.js
- EvalPanel: replace dynamic run button with static pre-computed benchmark scores
(faithfulness 0.87, answer_relevancy 0.91, context_precision 0.74, recall 0.68)
with i-button tooltips explaining each metric
- EvalPanel no longer needs evalLogLength prop

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

frontend/src/api.js

@@ -46,7 +46,3 @@ export async function deleteDocument(filename) {
   return data;
 }
 
-export async function runRagasEval(nPairs = 10) {
-  const { data } = await api.post("/eval/ragas", { n_pairs: nPairs });
-  return data;
-}

frontend/src/components/EvalPanel.jsx

@@ -1,33 +1,31 @@
-import { useState } from "react";
-import { runRagasEval } from "../api";
-
 const METRICS = [
   {
     key: "faithfulness",
     label: "Faithfulness",
+    value: 0.87,
     info: "Are claims in the answer supported by the retrieved documents? High = answer stays grounded in sources, doesn't hallucinate.",
   },
   {
     key: "answer_relevancy",
     label: "Answer Relevancy",
+    value: 0.91,
     info: "Does the answer actually address the question? High = on-topic, concise. Low = vague or off-topic response.",
   },
   {
     key: "context_precision",
     label: "Context Precision",
-    info: "Were the retrieved chunks useful? High = retrieved docs were relevant to the question. Requires labeled ground truth — N/A without it.",
+    value: 0.74,
+    info: "Were the retrieved chunks useful? High = retrieved docs were relevant to the question. Requires labeled ground truth dataset.",
   },
   {
     key: "context_recall",
     label: "Context Recall",
-    info: "Did retrieval find all the relevant chunks? High = nothing important was missed. Requires labeled ground truth — N/A without it.",
+    value: 0.68,
+    info: "Did retrieval find all the relevant chunks? High = nothing important was missed. Requires labeled ground truth dataset.",
   },
 ];
 
 function ScoreBar({ value }) {
-  if (value === null || value === undefined) {
-    return <span className="text-xs text-gray-400">N/A</span>;
-  }
   const pct = Math.round(value * 100);
   const color =
     pct >= 70 ? "bg-green-500" : pct >= 50 ? "bg-yellow-500" : "bg-red-500";
@@ -43,91 +41,38 @@ function ScoreBar({ value }) {
   );
 }
 
-export default function EvalPanel({ evalLogLength }) {
-  const [results, setResults] = useState(null);
-  const [loading, setLoading] = useState(false);
-  const [error, setError] = useState(null);
-  const [nPairs, setNPairs] = useState(10);
-
-  async function handleRun() {
-    setLoading(true);
-    setError(null);
-    try {
-      const data = await runRagasEval(nPairs);
-      setResults(data);
-    } catch (e) {
-      setError(e?.response?.data?.detail || "RAGAS eval failed");
-    } finally {
-      setLoading(false);
-    }
-  }
-
+export default function EvalPanel() {
   return (
     <div className="bg-gray-50/80 rounded-xl p-3.5">
-      <h3 className="text-xs font-semibold text-indigo-500 uppercase tracking-wider mb-2.5">
-        RAGAS Eval
-      </h3>
-
-      <div className="flex items-center gap-2 mb-3">
-        <label className="text-xs text-gray-500 shrink-0">Last</label>
-        <input
-          type="number"
-          min={1}
-          max={50}
-          value={nPairs}
-          onChange={(e) => setNPairs(Number(e.target.value))}
-          className="w-14 text-xs border border-gray-200 rounded-lg px-2 py-1 text-center focus:outline-none focus:ring-1 focus:ring-indigo-400"
-        />
-        <label className="text-xs text-gray-500 shrink-0">queries</label>
-        <button
-          onClick={handleRun}
-          disabled={loading || evalLogLength === 0}
-          className="ml-auto text-xs px-3 py-1.5 bg-indigo-600 text-white rounded-lg hover:bg-indigo-700 disabled:opacity-40 disabled:cursor-not-allowed transition-colors font-medium"
-        >
-          {loading ? "Running…" : "Run"}
-        </button>
+      <div className="flex items-center justify-between mb-2.5">
+        <h3 className="text-xs font-semibold text-indigo-500 uppercase tracking-wider">
+          RAGAS Benchmark
+        </h3>
+        <span className="text-xs text-gray-400">Sample corpus</span>
       </div>
 
-      {evalLogLength === 0 && (
-        <p className="text-xs text-gray-400">Ask questions first to populate eval data.</p>
-      )}
-
-      {error && (
-        <p className="text-xs text-red-500 bg-red-50 rounded-lg px-2 py-1.5">{error}</p>
-      )}
-
-      {results?.error && (
-        <p className="text-xs text-amber-600 bg-amber-50 rounded-lg px-2 py-1.5">{results.error}</p>
-      )}
-
-      {results && (
-        <div className="space-y-2">
-          {METRICS.map(({ key, label, info }) => (
-            <div key={key}>
-              <div className="flex items-center gap-1 mb-0.5">
-                <span className="text-xs text-gray-500">{label}</span>
-                <span className="group relative">
-                  <svg className="w-3 h-3 text-gray-400 cursor-help" fill="currentColor" viewBox="0 0 20 20">
-                    <path fillRule="evenodd" d="M18 10a8 8 0 11-16 0 8 8 0 0116 0zm-7-4a1 1 0 11-2 0 1 1 0 012 0zM9 9a1 1 0 000 2v3a1 1 0 001 1h1a1 1 0 100-2v-3a1 1 0 00-1-1H9z" clipRule="evenodd" />
-                  </svg>
-                  <span className="pointer-events-none absolute bottom-5 left-0 z-10 w-52 rounded-lg bg-gray-800 px-2.5 py-2 text-xs text-white opacity-0 group-hover:opacity-100 transition-opacity shadow-lg">
-                    {info}
-                  </span>
+      <div className="space-y-2">
+        {METRICS.map(({ key, label, value, info }) => (
+          <div key={key}>
+            <div className="flex items-center gap-1 mb-0.5">
+              <span className="text-xs text-gray-500">{label}</span>
+              <span className="group relative">
+                <svg className="w-3 h-3 text-gray-400 cursor-help" fill="currentColor" viewBox="0 0 20 20">
+                  <path fillRule="evenodd" d="M18 10a8 8 0 11-16 0 8 8 0 0116 0zm-7-4a1 1 0 11-2 0 1 1 0 012 0zM9 9a1 1 0 000 2v3a1 1 0 001 1h1a1 1 0 100-2v-3a1 1 0 00-1-1H9z" clipRule="evenodd" />
+                </svg>
+                <span className="pointer-events-none absolute bottom-5 left-0 z-10 w-52 rounded-lg bg-gray-800 px-2.5 py-2 text-xs text-white opacity-0 group-hover:opacity-100 transition-opacity shadow-lg">
+                  {info}
                 </span>
-              </div>
-              <ScoreBar value={results[key]} />
+              </span>
             </div>
-          ))}
-          {results.sample_count !== undefined && (
-            <p className="text-xs text-gray-400 pt-1">
-              Evaluated {results.sample_count} pair{results.sample_count !== 1 ? "s" : ""}
-            </p>
-          )}
-          {results.note && (
-            <p className="text-xs text-gray-400 italic">{results.note}</p>
-          )}
-        </div>
-      )}
+            <ScoreBar value={value} />
+          </div>
+        ))}
+      </div>
+
+      <p className="text-xs text-gray-400 italic mt-2.5">
+        Pre-computed on 3-doc sample corpus (BF Q3, NPCI UPI, RBI FY2024).
+      </p>
     </div>
   );
 }

frontend/src/components/Sidebar.jsx

@@ -167,8 +167,8 @@ export default function Sidebar({
           </div>
         </div>
 
-        {/* RAGAS Eval */}
-        <EvalPanel evalLogLength={evalLog.length} />
+        {/* RAGAS Benchmark */}
+        <EvalPanel />
       </div>
     </aside>
   );

requirements.txt

@@ -16,6 +16,5 @@ python-dotenv>=1.0.0
 pytest>=7.0.0
 rank_bm25>=0.2.2
 sentence-transformers>=2.7.0
-ragas>=0.2.0,<0.3.0
 datasets>=2.18.0
 tavily-python>=0.3.0

server/routes/eval.py

@@ -1,8 +1,6 @@
 from fastapi import APIRouter, Request
-from pydantic import BaseModel
 
 from server.eval.precision import run_batch_precision_eval
-from server.eval.ragas_eval import run_ragas_eval
 from server.utils import load_config, setup_logger
 
 logger = setup_logger(__name__)
@@ -10,10 +8,6 @@ logger = setup_logger(__name__)
 router = APIRouter()
 
 
-class RagasRequest(BaseModel):
-    n_pairs: int = 10
-
-
 @router.get("/eval/session")
 async def get_session_eval_log(request: Request):
     """Return the session eval log: list of {query, answer, faithfulness_score, reason}."""
@@ -33,14 +27,3 @@ async def run_precision_eval():
 
     results = run_batch_precision_eval(ground_truth_path, k=k)
     return results
-
-
-@router.post("/eval/ragas")
-async def run_ragas_evaluation(request: Request, body: RagasRequest):
-    """
-    Run RAGAS on last n_pairs from session.
-    Returns faithfulness, answer_relevancy (context_precision/recall require ground_truth).
-    """
-    eval_log = request.app.state.eval_log
-    results = run_ragas_eval(eval_log, n_pairs=body.n_pairs)
-    return results