Vector database integrations

Auto-instrument ChromaDB, Pinecone, Qdrant, Milvus, Astra DB, and PostgreSQL vector operations with OpenTelemetry

OpenLIT instruments vector database clients to trace every query, upsert, delete, and index operation your application performs. You get full visibility into retrieval latency, collection names, operation types, and result counts — without changing your database code.

Get started

1. Install OpenLIT

   pip install openlit
   
2. Install your vector database client

   # ChromaDB
   pip install chromadb
   
   # Pinecone
   pip install pinecone
   
   # Qdrant
   pip install qdrant-client
   
   # Milvus
   pip install pymilvus
   
   # Astra DB
   pip install astrapy
   
3. Initialize OpenLIT

   Call openlit.init() before any vector database operations.

   ChromaDB example

   import openlit
   import chromadb
   
   openlit.init(otlp_endpoint="http://127.0.0.1:4318")
   
   client = chromadb.Client()
   collection = client.create_collection("documents")
   
   collection.add(
       documents=["OpenLIT makes AI observability easy"],
       ids=["doc1"]
   )
   
   results = collection.query(
       query_texts=["AI observability"],
       n_results=1
   )
   print(results)
   

   import openlit
   from pinecone import Pinecone
   
   openlit.init(otlp_endpoint="http://127.0.0.1:4318")
   
   pc = Pinecone(api_key="YOUR_API_KEY")
   index = pc.Index("my-index")
   
   index.upsert(
       vectors=[("vec1", [0.1, 0.2, 0.3])]
   )
   
   results = index.query(
       vector=[0.1, 0.2, 0.3],
       top_k=5
   )
   print(results)
   

Supported vector databases

Database	Python	TypeScript	Instrumentor key
ChromaDB	✅ chromadb	✅	chroma
Pinecone	✅ pinecone	✅	pinecone
Qdrant	✅ qdrant-client	✅	qdrant
Milvus	✅ pymilvus	—	milvus
Astra DB	✅ astrapy	—	astra
PostgreSQL (psycopg3)	✅ psycopg	—	psycopg

Span attributes captured

Vector database spans include the following attributes:

Attribute	Description
db.system	Database system name (e.g., chromadb, pinecone)
db.operation	Operation type (e.g., query, upsert, delete)
db.collection.name	Collection or index name
db.vector.query.top_k	Number of results requested
db.vector.dimension	Vector dimension
server.address	Database host
server.port	Database port

PostgreSQL (psycopg3) instrumentation traces SQL queries executed against pgvector-enabled tables using the same span attributes as other vector operations, plus standard database span attributes like db.statement.

Disabling specific databases

import openlit

# Instrument everything except Pinecone
openlit.init(
    otlp_endpoint="http://127.0.0.1:4318",
    disabled_instrumentors=["pinecone"]
)

---

LLM provider integrations

Trace LLM calls with token counts, cost, and latency

AI framework integrations

Trace agent execution, tool calls, and chains

Destinations

Send telemetry to Datadog, Grafana, New Relic, and other observability stacks