✨ phoenix tracing added for llm

backend/.env.template

@@ -83,5 +83,10 @@ POSTHOG_API_KEY=
 POSTHOG_HOST=
 
 # QDRANT API Configuration
+## need to be set before running the server
 QDRANT_API_KEY=
-QDRANT_TOP_K=
+QDRANT_TOP_K=
+
+# Phoenix Configuration
+## default values set in the code for local development
+PHOENIX_API_ENDPOINT=

backend/app/config.py

@@ -135,4 +135,7 @@
 # Dspy Integration Configuration
 CLINICAL_TRIAL_SQL_PROGRAM: str  = "app/dspy_integration/dspy_programs/clinical_trials_sql_generation.json"
 CLINICAL_TRIALS_RESPONSE_REFINEMENT_PROGRAM: str = "app/dspy_integration/dspy_programs/clinical_trials_response_refinement.json"
-ORCHESRATOR_ROUTER_PROMPT_PROGRAM: str = "app/dspy_integration/dspy_programs/orchestrator_router_prompt.json"
+ORCHESRATOR_ROUTER_PROMPT_PROGRAM: str = "app/dspy_integration/dspy_programs/orchestrator_router_prompt.json"
+
+# Phoenix Configuration
+PHOENIX_API_ENDPOINT: str = config("PHOENIX_API_ENDPOINT", default="http://127.0.0.1:6007/v1/traces")

backend/app/main.py

@@ -1,7 +1,10 @@
+# setup tracing
+from app.services.tracing import setup_tracing
+setup_tracing()
+
 from fastapi import FastAPI, HTTPException
 from fastapi.responses import RedirectResponse
 
-# from authx import AuthX, AuthXConfig
 from app.database.redis import Redis
 
 from app import config
@@ -11,9 +14,6 @@
 
 from app.middleware.process_time import ProcessTimeHeaderMiddleware
 
-from app.services.tracing import setup_tracing
-
-
 def get_application() -> FastAPI:
     application = FastAPI(
         title=config.PROJECT_NAME, debug=config.DEBUG, version=config.VERSION
@@ -27,9 +27,6 @@ def redirect_to_docs() -> RedirectResponse:  # pylint: disable=W0612
     async def startup():  # pylint: disable=W0612
         print()
 
-        # setup tracing
-        setup_tracing()
-
         # connect to redis
         cache = Redis()
         await cache.connect()

backend/app/services/tracing.py

@@ -5,7 +5,14 @@
 from sentry_sdk.integrations.starlette import StarletteIntegration
 from sentry_sdk.integrations.fastapi import FastApiIntegration
 
-from app.config import SENTRY_DSN, SENTRY_ENABLE_TRACING
+from openinference.instrumentation.llama_index import LlamaIndexInstrumentor
+from opentelemetry import trace as trace_api
+from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
+from opentelemetry.sdk import trace as trace_sdk
+from opentelemetry.sdk.resources import Resource
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor
+
+from app.config import SENTRY_DSN, SENTRY_ENABLE_TRACING, PHOENIX_API_ENDPOINT, ENVIRONMENT
 
 def setup_tracing():
     sentry_sdk.init(
@@ -24,4 +31,14 @@ def setup_tracing():
                 event_level=logging.WARNING
             ),
         ],
-    )
+    )
+
+
+    if ENVIRONMENT == 'production':
+        resource = Resource(attributes={})
+        tracer_provider = trace_sdk.TracerProvider(resource=resource)
+        span_exporter = OTLPSpanExporter(endpoint=PHOENIX_API_ENDPOINT)
+        span_processor = SimpleSpanProcessor(span_exporter=span_exporter)
+        tracer_provider.add_span_processor(span_processor=span_processor)
+        trace_api.set_tracer_provider(tracer_provider=tracer_provider)
+        LlamaIndexInstrumentor().instrument()

backend/pyproject.toml

@@ -31,6 +31,9 @@ llama-index-llms-together = "^0.1.3"
 llama-index-postprocessor-cohere-rerank = "^0.1.2"
 together = "^0.2.11"
 llama-index-vector-stores-qdrant = "^0.1.4"
+openinference-semantic-conventions = "^0.1.5"
+openinference-instrumentation-llama-index = "^1.2.1"
+opentelemetry-exporter-otlp = "^1.23.0"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^8.1.1"