config.yaml

@@ -1,11 +1,7 @@
 project_id: bnt-orquestador-cognitivo-dev
 location: us-central1
-service_account: sa-cicd-gitlab@bnt-orquestador-cognitivo-dev.iam.gserviceaccount.com
 
 bucket: bnt_orquestador_cognitivo_gcs_configs_dev
-base_image: us-central1-docker.pkg.dev/bnt-orquestador-cognitivo-dev/dfcx-pipelines/ap01194-orq-cog-dfcx-agent-eval-in:v1.0.6-ft-anibal
-dialogflow_agent_id: 5590ff1d-1f66-4777-93f5-1a608f1900ac
-processing_image: us-central1-docker.pkg.dev/bnt-orquestador-cognitivo-dev/dfcx-pipelines/ap01194-orq-cog-rag-ops:v1.3.7
 
 agent_name: sigma
 agent_instructions: |
@@ -157,17 +153,10 @@ agent_embedding_model: gemini-embedding-001
 agent_thinking: 0
 
 index_name: si1
+index_deployed_id: si1_deployed
 index_endpoint: projects/1007577023101/locations/us-central1/indexEndpoints/76334694269976576
 index_dimensions: 3072
 index_machine_type: e2-standard-16
 index_origin: gs://bnt_orquestador_cognitivo_gcs_kb_dev/
 index_destination: gs://bnt_orquestador_cognitivo_gcs_configs_dev/
 index_chunk_limit: 3000
-
-bigquery_project_id: "bnt-lakehouse-innovacion-des"
-bigquery_dataset_id: "ds_orquestador_des"
-bigquery_table_ids:
-    synth_gen: "eval_questions"
-    search_eval: "search_results"
-    keypoint_eval: "keypoints_results"
-    red_team: "redteam_results"

pyproject.toml

@@ -29,6 +29,7 @@ build-backend = "uv_build"
 
 [dependency-groups]
 dev = [
+    "clai>=1.62.0",
     "pytest>=8.4.1",
     "ruff>=0.12.10",
     "ty>=0.0.1a19",

src/rag_eval/agent.py

@@ -52,7 +52,7 @@ async def conocimiento(ctx: RunContext[Deps], query: str) -> str:
     t_embed = time.perf_counter()
 
     search_results = await ctx.deps.vector_search.async_run_query(
-        deployed_index_id=settings.index_name,
+        deployed_index_id=settings.index_deployed_id,
         query=list(query_embedding.embeddings[0]),
         limit=5,
     )
@@ -82,3 +82,11 @@ async def conocimiento(ctx: RunContext[Deps], query: str) -> str:
         for i, result in enumerate(search_results, start=1)
     ]
     return "\n".join(formatted_results)
+
+
+if __name__ == "__main__":
+    deps = Deps(
+        vector_search=settings.vector_search,
+        embedder=settings.embedder,
+    )
+    agent.to_cli_sync(deps=deps)

src/rag_eval/config.py

@@ -22,7 +22,7 @@ class Settings(BaseSettings):
 
     project_id: str
     location: str
-    service_account: str
+    bucket: str
 
     agent_name: str
     agent_instructions: str
@@ -31,6 +31,7 @@ class Settings(BaseSettings):
     agent_thinking: int
 
     index_name: str
+    index_deployed_id: str
     index_endpoint: str
     index_dimensions: int
     index_machine_type: str = "e2-standard-16"
@@ -38,14 +39,6 @@ class Settings(BaseSettings):
     index_destination: str
     index_chunk_limit: int
 
-    bigquery_dataset_id: str
-    bigquery_project_id: str | None = None
-    bigquery_table_ids: dict[str, str]
-
-    bucket: str
-    base_image: str
-    dialogflow_agent_id: str
-    processing_image: str
 
     model_config = SettingsConfigDict(yaml_file=CONFIG_FILE_PATH)
 
@@ -75,12 +68,14 @@ class Settings(BaseSettings):
     @cached_property
     def vector_search(self) -> GoogleCloudVectorSearch:
         """Return a configured vector search client."""
-        return GoogleCloudVectorSearch(
+        vs = GoogleCloudVectorSearch(
             project_id=self.project_id,
             location=self.location,
             bucket=self.bucket,
             index_name=self.index_name,
         )
+        vs.load_index_endpoint(self.index_endpoint)
+        return vs
 
     @cached_property
     def embedder(self) -> Embedder: