update

backend/config.py

@@ -1,3 +1,4 @@
+import os
 from pathlib import Path
 
 PROJECT_ROOT = Path(__file__).resolve().parent.parent
@@ -41,11 +42,17 @@ EMBEDDING_BLOCK_CAP = 25      # max. titles per block (keep the LLM list short/s
 # block context — from this cosine on two are the same statement (checked on aak: ≥0.88 are
 # without exception true duplicates). Conservative 0.90 so different aspects (∈NP ≠ NP-hard) stay separate.
 EMBEDDING_SUB_DUP = 0.90
+# Cosine: two judge-subblocks state the SAME point → one cluster in the clarify majority vote.
+# Lower than _DUP because it must merge paraphrases (not just typo-variants). Empirically 0.80 keeps
+# distinct aspects (∈NP vs NP-hard) apart while clustering re-wordings of the same fact.
+EMBEDDING_SUB_SAME = 0.80
 
-# Cap for concurrent CLI agent processes (across all generations).
-# Own lane for interactive calls (chat, elements) so they don't hang behind
-# running writers in the queue.
-MAX_CONCURRENT_AGENTS = 10
+# Caps for concurrent CLI agent processes (env-overridable). Two nested limits, both always active:
+# a per-topic cap and a global cap across all topics. Defaults 10/10 = previous behavior (global
+# dominates). Locally raise the global cap to actually parallelize across topics (per-topic stays 10).
+# Own lane for interactive calls (chat, elements) so they don't hang behind running writers.
+MAX_CONCURRENT_AGENTS = int(os.getenv("MAX_CONCURRENT_AGENTS", "10"))                      # global, all topics
+MAX_CONCURRENT_AGENTS_PER_TOPIC = int(os.getenv("MAX_CONCURRENT_AGENTS_PER_TOPIC", "10"))  # per topic
 MAX_CONCURRENT_INTERACTIVE = 8
 
 # Grace window of the consensus races (blocks, guide, OnePager): after the first