Accept long sequences; truncate to ESM2 max len with visible note

Previously the hard 1022-aa cap in validate_protein rejected common
CRISPR effectors (SpCas9 = 1368 aa, FnCas12a = 1300 aa, LshCas13a = 1389
aa) on both Compare and Distance tabs. Twin's custom tower and its ESM
tokenizer already truncate internally; only fair-esm's batch_converter
in embed_esm2 needed explicit truncation.

Changes:
- validate_protein no longer errors on sequences > 1022 aa
- embed_esm2 truncates to ESM2_MAX_LEN = 1022 before batch_converter
- Compare tab: blockquote note above the stats table when truncated
- Distance tab: orange-bordered notice at the top of the result when
either Protein A or B exceeded 1022 aa

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

app.py

@@ -138,8 +138,12 @@ def get_device():
 AMINO_ACIDS = set("ACDEFGHIKLMNPQRSTVWY")
 AMINO_ACIDS_EXTENDED = AMINO_ACIDS | set("XBZJOU")  # Include ambiguous
 
+ESM2_MAX_LEN = 1022  # ESM2 position embedding limit; longer sequences are truncated
+
+
 def validate_protein(sequence):
-    """Validate protein sequence."""
+    """Validate protein sequence. Does NOT reject long sequences — both embedders
+    truncate internally; we surface a note in the output instead."""
     if not sequence or len(sequence.strip()) == 0:
         return False, "Sequence is empty"
     sequence = sequence.upper().replace(" ", "").replace("\n", "")
@@ -148,8 +152,6 @@ def validate_protein(sequence):
         return False, f"Invalid characters: {invalid}"
     if len(sequence) < 10:
         return False, f"Sequence too short: {len(sequence)} < 10 aa"
-    if len(sequence) > 1022:
-        return False, f"Sequence too long: {len(sequence)} > 1022 aa (ESM2 limit)"
     return True, ""
 
 def strip_fasta_header(text):
@@ -159,11 +161,15 @@ def strip_fasta_header(text):
 
 @torch.no_grad()
 def embed_esm2(sequence):
-    """Compute ESM2 embedding (mean-pooled)."""
+    """Compute ESM2 embedding (mean-pooled). Truncates to ESM2_MAX_LEN."""
     model, alphabet = get_esm2()
     batch_converter = alphabet.get_batch_converter()
     device = get_device()
 
+    # ESM2 position embeddings cap at 1022; longer sequences must be truncated.
+    if len(sequence) > ESM2_MAX_LEN:
+        sequence = sequence[:ESM2_MAX_LEN]
+
     data = [("protein", sequence)]
     _, _, batch_tokens = batch_converter(data)
     batch_tokens = batch_tokens.to(device)
@@ -491,8 +497,12 @@ def process(sequence: str, top_k: int = 10, twin_aspect: str = "BP"):
             "uniprot": str(e).splitlines()[0][:200],
         }])
 
-    summary = f"""### Results
+    trunc_note = (f"\n> ⚠️ Sequence truncated from {len(sequence)} to {ESM2_MAX_LEN} aa "
+                  f"(ESM2 position-embedding limit). Twin also truncates internally.\n"
+                  if len(sequence) > ESM2_MAX_LEN else "")
 
+    summary = f"""### Results
+{trunc_note}
 | | ESM2 | Twin ({twin_aspect}) |
 |---|---|---|
 | Dimension | {esm2_stats['dim']} | {twin_stats['dim']} |
@@ -577,6 +587,14 @@ with gr.Blocks(
                 valid, err = validate_protein(seq)
                 if not valid:
                     return f"<div style='color:#dc2626;font-weight:600;'>Error in sequence {name}: {err}</div>"
+            trunc_notes = []
+            for name, seq in (("A", seq_a), ("B", seq_b)):
+                if len(seq) > ESM2_MAX_LEN:
+                    trunc_notes.append(f"Protein {name} truncated from {len(seq)} → {ESM2_MAX_LEN} aa")
+            trunc_html = (f"<div style='background:#fff7ed;border-left:3px solid #f97316;"
+                          f"padding:8px 12px;margin:8px 0;font-size:12px;color:#9a3412;'>"
+                          f"⚠️ {'; '.join(trunc_notes)} (ESM2 position-embedding limit).</div>"
+                          if trunc_notes else "")
             d = compute_distance(seq_a, seq_b, aspect)
             # Green = similar, red = dissimilar
             l2_bar = _distance_bar(
@@ -595,6 +613,7 @@ with gr.Blocks(
             )
             return (
                 f"<h3 style='margin-top:0;'>Twin/{aspect} distance</h3>"
+                f"{trunc_html}"
                 f"{l2_bar}{cos_bar}"
                 f"<p style='font-size:12px;color:#666;margin-top:16px;'>"
                 f"cosine similarity = {d['cos_sim']:+.4f}&nbsp; · &nbsp;"