LangSmith tracing wired into all five configs and eval harness

requirements.txt

@@ -7,3 +7,4 @@ psycopg2-binary==2.9.10
 python-dotenv==1.1.0
 requests==2.34.2
 sentence-transformers==3.4.1
+langsmith

src/configs/grounded.py

@@ -8,6 +8,7 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+from langsmith import traceable
 from openai import OpenAI
 
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
@@ -42,6 +43,7 @@ SYSTEM_PROMPT = (
 )
 
 
+@traceable(name="grounded")
 def run(query: str, source_filter: list[str] | None = None) -> dict:
     query_embedding = embed_query(query)
     retrieved_chunks = rrf_search(query, query_embedding, source_filter=source_filter, top_k=5)

src/configs/hybrid.py

@@ -7,6 +7,7 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+from langsmith import traceable
 from openai import OpenAI
 
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
@@ -33,6 +34,7 @@ SYSTEM_PROMPT = (
 )
 
 
+@traceable(name="hybrid")
 def run(query: str, source_filter: list[str] | None = None) -> dict:
     query_embedding = embed_query(query)
     retrieved_chunks = rrf_search(query, query_embedding, source_filter=source_filter, top_k=5)

src/configs/kg_augmented.py

@@ -12,6 +12,7 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+from langsmith import traceable
 from openai import OpenAI
 
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
@@ -181,6 +182,7 @@ def _format_chunk(chunk: dict) -> str:
     return f"[{chunk['source']} | {chunk['paragraph_id']}]\n{chunk['text']}\n"
 
 
+@traceable(name="kg_augmented")
 def run(query: str, source_filter: list[str] | None = None) -> dict:
     query_embedding = embed_query(query)
     retrieved_chunks = rrf_search(query, query_embedding, source_filter=source_filter, top_k=5)

src/configs/naive.py

@@ -7,6 +7,7 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+from langsmith import traceable
 from openai import OpenAI
 
 sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
@@ -33,6 +34,7 @@ SYSTEM_PROMPT = (
 )
 
 
+@traceable(name="naive")
 def run(query: str, source_filter: list[str] | None = None) -> dict:
     query_embedding = embed_query(query)
     retrieved_chunks = dense_search(query_embedding, source_filter=source_filter, top_k=5)

src/configs/rerank.py

@@ -7,6 +7,7 @@ import sys
 from pathlib import Path
 
 from dotenv import load_dotenv
+from langsmith import traceable
 from openai import OpenAI
 from sentence_transformers import CrossEncoder
 
@@ -36,6 +37,7 @@ SYSTEM_PROMPT = (
 )
 
 
+@traceable(name="rerank")
 def run(query: str, source_filter: list[str] | None = None) -> dict:
     query_embedding = embed_query(query)
     candidates = rrf_search(query, query_embedding, source_filter=source_filter, top_k=20)

src/eval.py

@@ -6,6 +6,8 @@ results for comparison. No scoring/taxonomy logic here yet.
 import sys
 from pathlib import Path
 
+from langsmith import traceable
+
 sys.path.insert(0, str(Path(__file__).resolve().parent))
 
 from configs.grounded import run as run_grounded
@@ -23,6 +25,7 @@ CONFIG_RUNNERS = {
 }
 
 
+@traceable(name="evaluate_query")
 def evaluate_query(query: str, source_filter: list[str] | None = None) -> dict:
     results = {}