fix(rerank): derive the doc cap from tokens (measured floor 1.47), cap the query (#9)

rag/retrieval.py

@@ -25,6 +25,32 @@ BM25_DB = Path(os.environ.get("BM25_DB",
                str(REPO_ROOT / "bm25" / "crop_chem_docs.db")))
 COLLECTION = f"{os.environ.get('PRODUCT_NAME', 'crop_chem')}_docs"
 
+# --- Reranker input budget (see docs-mcp-template + retrieval-lore.md) ------
+# jina-reranker-v2 is a BERT cross-encoder with bert.context_length=1024 and
+# LEARNED ABSOLUTE position embeddings: 1024 is a HARD ceiling, not something a
+# bigger llama.cpp --ubatch-size can lift. llama.cpp 500s the ENTIRE batch if
+# any (query, doc) pair exceeds it, so ONE oversized chunk silently drops that
+# whole query to fused order — the same ~90%->~62% P@1 cliff as the sidecar
+# being off the network, and just as quiet.
+#
+# A char cap is only a PROXY for tokens. Derive it from this corpus's measured
+# chars-per-token FLOOR (tokenise real chunks via {RERANK_URL}/tokenize and
+# take the min). Measured floor for this corpus: 1.47 (EPA/Bayer label text).
+RERANK_CTX_TOKENS = int(os.environ.get("RERANK_CTX_TOKENS", "1024"))
+RERANK_CHARS_PER_TOKEN = float(os.environ.get("RERANK_CHARS_PER_TOKEN", "1.45"))
+RERANK_QUERY_MAX_CHARS = int(os.environ.get("RERANK_QUERY_MAX_CHARS", "300"))
+# Margin covers [CLS]/[SEP] framing plus slack, because chars-per-token is an
+# ESTIMATE from a sample. Budget to ~94%, never to exactly 1024.
+_RERANK_MARGIN_TOKENS = int(os.environ.get("RERANK_MARGIN_TOKENS", "64"))
+_RERANK_QUERY_TOKENS = int(RERANK_QUERY_MAX_CHARS / RERANK_CHARS_PER_TOKEN) + 1
+RERANK_DOC_MAX_CHARS = max(
+    256,
+    int(
+        (RERANK_CTX_TOKENS - _RERANK_QUERY_TOKENS - _RERANK_MARGIN_TOKENS)
+        * RERANK_CHARS_PER_TOKEN
+    ),
+)
+
 
 class Retriever(Protocol):
     name: str
@@ -230,7 +256,7 @@ class RerankedRetriever:
             if not d:
                 continue
             # Truncate to keep under the reranker's per-pair context limit
-            docs.append(d[:2000])
+            docs.append(d[:RERANK_DOC_MAX_CHARS])
             kept_pages.append((source, source_key))
 
         if not docs:
@@ -240,7 +266,8 @@ class RerankedRetriever:
         try:
             r = httpx.post(
                 f"{self.rerank_url}/v1/rerank",
-                json={"query": query, "documents": docs},
+                json={"query": query[:RERANK_QUERY_MAX_CHARS],
+                      "documents": docs},
                 timeout=self.timeout,
             )
             r.raise_for_status()