fix(docs): update README.md to reflect the new threshold value for cosine similarity in the ask command

fix(cli.py): change default threshold value in the ask command from 0.25 to 0.001 for consistency
fix(config.py): update threshold value from 0.25 to 0.001 to align with new requirements
test(config.py): update test case to check for the new threshold value of 0.001 instead of 0.25

Author: ariel-frischer

README.md

@@ -125,7 +125,7 @@ sources = [
 # embed_dims = 1536
 # chat_model = "gpt-4o-mini"
 # max_chunk_chars = 2000
-# threshold = 0.25             # minimum cosine similarity to include a result (also --threshold flag)
+# threshold = 0.001             # minimum cosine similarity to include a result (also --threshold flag)
 # rerank_fetch_k = 20
 # rerank_top_k = 5
 # index_code = false       # set true to also index source code files

src/kb/cli.py

@@ -43,7 +43,7 @@
   kb index [DIR...] [--no-size-limit]  Index sources (skip files > max_file_size_mb)
   kb allow <file>                Whitelist a large file for indexing
   kb search "query" [k]          Hybrid semantic + keyword search (default k=5)
-  kb ask "question" [k] [--threshold N]  RAG: search + rerank + answer (default k=8, threshold=0.25)
+  kb ask "question" [k] [--threshold N]  RAG: search + rerank + answer (default k=8, threshold=0.001)
   kb similar <file> [k]          Find similar documents (no API call, default k=10)
   kb tag <file> tag1 [tag2...]   Add tags to a document
   kb untag <file> tag1 [tag2...]  Remove tags from a document

src/kb/config.py

@@ -39,7 +39,7 @@
 # min_chunk_chars = 50
 
 # Search
-# threshold = 0.25        # minimum cosine similarity to include a result (0.0–1.0)
+# threshold = 0.001        # minimum cosine similarity to include a result (0.0–1.0)
 # rrf_k = 60.0            # RRF smoothing constant
 # rerank_fetch_k = 20     # candidates to fetch for LLM rerank
 # rerank_top_k = 5        # how many to keep after rerank
@@ -91,7 +91,7 @@ class Config:
     chat_model: str = "gpt-4o-mini"
     max_chunk_chars: int = 2000
     min_chunk_chars: int = 50
-    threshold: float = 0.25
+    threshold: float = 0.001
     rrf_k: float = 60.0
     rerank_fetch_k: int = 20
     rerank_top_k: int = 5

tests/test_config.py

@@ -22,7 +22,7 @@ def test_defaults(self):
         assert cfg.chat_model == "gpt-4o-mini"
         assert cfg.max_chunk_chars == 2000
         assert cfg.min_chunk_chars == 50
-        assert cfg.threshold == 0.25
+        assert cfg.threshold == 0.001
         assert cfg.rrf_k == 60.0
         assert cfg.rerank_fetch_k == 20
         assert cfg.rerank_top_k == 5